Which type of parking is best?

Which type of parking is best?

Futuristic autonomous car flying in the sky. Original design. 3D rendering image.

Local laws promoting Clearways, prohibit on-street parking during certain time periods to increase the number of lanes available for vehicular movement. While this benefits long-distance travelers, it increases the cost of searching for parking and walking to the final destination. This present study sought to quantify the negative impact of clearways and determine the optimized clearway ratio to minimize the total travel time (TTT). It redefined the TTT as a combination of the travel time from the origin to the destination (TOD), the additional vehicle cruising time for parking after reaching the final destination (Tcruising) and the parkers’ walking time from the parked vehicle to the final destination (Twalking). 

A parking searching model (PSM) was built in Python based on the parking distribution and the employment rate in the City of Sydney, Australia. The model simulated the vehicle cruising behavior for a parking bay after reaching the final destination in a three-hour time frame. The simulation was repeated for different parking demands (Q) and clearway ratios (ρ). The output Cruising and Walking were subsequently analyzed by adopting three different approaches. Firstly, the Cruising and Walking were plotted against a three-hour time frame to evaluate their tendency. Secondly, they were combined to investigate the average additional time expended for each Q and ρ combination. Thirdly, these factors were further evaluated with the collected TOD to explore the optimized ρ for each Q

It is discovered that all clearway levels exhibit a longer Cruising and Walking than the on-street parking situation. The later a parker enters the simulation system, the greater the impact of the clearway. As the Q increases, the range of Cruising and Walking becomes bigger, and the extent to which they rise will also increase. The TTT increases with an increase in ρ when the Q is low. The most optimized TTT occurs at smaller ρ when the Q is higher. Ultimately, it will be concluded that the presence of a clearway does not always benefit the traffic network and mitigate the TTT

1.1 Research Background 

On-street parking provides easy and convenient access to stored cars for residents, workers, and customers. From an economic perspective, it saves a considerable amount of urban space by reducing cities’ demand for car parks, if the streets would otherwise be unused. However, when traffic management is taken into consideration, the provision of on-street parking adversely impacts the traffic capacity and driving speed on the road, which results in increased congestion. 

Existing studies (Guo et al., 2012; Portilla et al., 2009; Yousif and Purnawan, 1999) have quantitatively demonstrated the benefits of creating clearway by removing on-street parking. Weant and Levinson (1990) claimed that removing on-street parking would result in a gain of 50% of traffic capacity. In Australia, a clearway is a special road on which vehicles, except for taxis and buses, are generally not permitted to stop at a particular time of day. However, the overall travel time may not be reduced as a result of the clearway. This is primarily because those who previously would have parked near to their final destination, now face additional in-vehicle travel time to search for and find a new parking location (on- or off-street) when the nearby on-street parking is prohibited, as well as the extra walking time from the carpark location to their final destination. 

1.2 Research Question 

The research focuses on the combined travel time (Tlocal) of cruising a parking bay after the vehicle has reached its destination (Tcruising) and walking from the parked vehicle to the destination (Twalking). These additional impacts will be discussed in conjunction with the inherent traffic capacity loss, in terms of the origin to destination travel time (TOD), that is caused by on-street parking. This will ensure the efficiency of this research in its efforts to re-evaluate whether on-street parking will reduce total travel time (TTT) and the need for a clearway. The following questions have been raised to guide the present study: 

  • 1. During the peak hour (period), which scenarios consume more time when travelers cruise for a 1
  • parking bay: clearways or on-street parking? 
  • 2. During the peak hour (period), what is the impact of different clearway ratio (ρ) on Cruising and Walking of parkers under various demands (Q)? 
  • 3. Which ρ in a city’s transport network can optimize Parker’s TTT
  • 4. What is the parker’s TTT when Cruising and Walking for on-street and off-street parking are considered? 

1.3 Objectives 

The aim of the present study is to investigate the significance of on-street parking in the traffic network, when compared to the clearway. Thus, the objectives of the present study are as follows: 

  • 1. With regards to the different Q, summarizing the trend of Tlocal when the amount of on-street parking locations are reduced. 
  • 2. Determine the relationship between the ρ in the system and Cruising and Walking, respectively. 
  • 3. Re-evaluate the necessity of on-street parking according to the factors of travel time and road capacity. 
  • 4. Quantitatively determine the vehicle TTT for both clearway and on-street parking situations with varying Q and ρ. 

1.4 Significance 

Nobel Prize Winning Economist & Stanford Professor Paul Romer on Hyperinflation & Protecting Science

The literature review refers to the studies which have explored the influence of on-street parking and clearways. It was discovered that a considerable body of research has evaluated the impact of on-street parking on traffic with most studies concluding that on-street parking will increase travel time and reduce road capacity. There are also quite a few papers that discuss the positive effects of clearways. In contrast, there has been limited focus on the additional Cruising and Walking expended after parking at a farther location, which is also a significant amount of travel time. Accordingly, the present study will investigate the surplus Cruising and Walking when determining the average travel time for clearway and on-street parking. As such, this research will fulfill a significant deficiency in this field. 

2 Literature Review 

The provision of on-street parking imposes negative impacts on travel time and reduces the traffic capacity, (Cao et al., 2015; Guo et al., 2012; Portilla et al., 2009; Wijayaratna, 2015; Yousif and Purnawan, 1999) yet it still benefits the urban space in various ways. (Marshall et al., 2008) As asserted by many researchers, clearways facilitate a safe and smooth flow of traffic with less travel time. (Zhang and Excel, 2011) However, the Cruising for a bay being available for parking is significant for both on- and off-street parking (Belloche, 2015; Geroliminis, 2015; Ommeren et al., 2012; Shoup, 2006) Nonetheless, research on the parking and travel time issue overlooks the additional time required after arriving at the destination. Therefore, the present study focuses on determining the effect of on-street parking versus clearway on travel time based on the current spatial distribution of parking. (Young et al., 2010) 

2.1 Impacts of on-street parking on traffic 

On-street parking can impact road capacity in two ways. The reduction in the available lanes of a road to accommodate on-street parking is the primary factor that reduces road capacity. Consequently, on-street parking can cause extensive delay, especially during peak hours (Wijayaratna, 2015). Secondly, on-street parking causes delays as parkers search for and attempt to acquire on-street parking, stopping, and doing a reverse or parallel parking maneuver. 

2.1.1 The influence of on-street parking maneuvers 

Portilla et al. (2009) quantified the influence of on-street parking maneuvers on the average travel time. It is achieved through applying the M/M/∞ queuing model to register the parking delay. Portilla et al. (2009) found that for each parking and unparking maneuver, the adjacent lane is affected. Meanwhile, the following vehicle either needs to slow down or change lanes to avoid the obstacle until the movement is completed. The simulated result suggested that the lane changing reduces the free lane’s traffic flow under specific traffic intensity. 

Guo et al. (2012) applied the hazard-based duration model to analyze the influential factors of on-street parking and its corresponding impact on travel time. The researcher found that the parking maneuver is one of the influential factors related to on-street parking.(Guo et al., 2012) Oncoming vehicles tend to queue at blockages, where the parking maneuver occurs. 

In 2012, Guo et al. (2012) investigated the interactions between parking/unparking maneuvers and through vehicles by conducting a Monte Carlo Simulation. According to the simulated results, parking maneuvers lead to a capacity reduction of 35% if the parking/unparking maneuver in the system increases from 0% to 35%. (Guo et al., 2012) The longer the journey duration, the more the road will be congested for two reasons: vehicles need to slow down to find parking, which causes the following vehicles to slow down as a result, generating a spillback, and the blockage on the road reduces traffic flow. 

Not only do parking maneuvers impact on the travel time, but also the type of parking maneuvers can affect the traffic. Box (2004) found that parallel parking is more desirable than the angle on street parking. Segal (1972) also found that the probability of an accident occurring while parallel parking is only 20% that of angle parking. Both studies proposed that the application of angle parking should be minimized. Yousif and Purnawan (1999) analyzed the time taken for arrival and departure during parking, discovering that entering the parking bay requires more time than departing. Parallel parking should be used to avoid angle parking and maximize the utilized road space. (Yousif and Purnawan, 1999) These findings are consistent with the suggestions made by Segal (1972) and Box (2004). 

The aforementioned existing studies all highlighted the negative influence of on-street parking on travel time from the perspective of parking maneuvers as well as the effect of different types of parking maneuvers. Accordingly, the present study counts the Tcruising in TTT in addition to the TOD which has been evaluated by numerous authors when determining the advantages for on-street parking. 

2.1.2 The effect of on-street parking on vehicular flow 

Many scholars have suggested that the presence of parked vehicles on the street influences the vehicular flow (Goolsby, 1971; Gordon et al., 2005; Smith et al., 2003). Guo et al. (2012) found an efficient lane width exhibits a significant positive relationship with vehicular flow. The main reason is that the presence of parked vehicles can easily cause vehicles to queue when the traffic flow remains constant but the road capacity decreases. This conclusion strongly resonated with Goolsby (1971)’s results in which blocking lanes generates different levels of capacity reduction. For example, blocking one out of three lanes can potentially lead to a capacity reduction of 50 percent. Other articles have analyzed the effect of street-parked vehicles on vehicular flow.

Most studies identified the negative effects of on-street parking without exception. The removal of street parking zones offers significant benefits to traffic flow. Conversely, Marshall et al. (2008) argued that “On-street parking should be more commonly used but especially in the situations in which the road is part of the destination and the intent is to cause drivers to slow down.” 

2.1.3 Marginal effects of on-street parking 

Interestingly, there are many articles, such as Portilla et al. (2009) and Wilbur and Associates (1965) study the benefits of on-street parking as against it. Marshall et al. (2008) did not challenge the impact of on-street parking on traffic capacity, but also stressed the benefits that on-street parking offers the city. The author concluded that on-street parking promotes land use efficiency. (Marshall et al., 2008) Consequently, it facilitates a higher density of commercial development in a region where the parking bay is sufficient.

On-street parking does not require additional access lands or driveways. Moreover, when on-street parking is prohibited, the convenience factor of travel is diminished and commercial activities near the on-street parking zone would inevitably be affected. (Marshall et al., 2008) In contrast, a study reveals that not only did the deduction of on-street parking increase road capacity, but also that off-street parking in the central business district (CBD) enhances retail activity. (Wilbur and Associates, 1965) The finding of both Marshall et al. (2008) and Wilbur and Associates (1965) seem to be a different concept.

Nonetheless, they were in agreement that commercial activities near parking zones are easier to gain benefits if they are to rely solely on the nearby parking lots to meet all parking needs. Statistically, Marshall et al. (2008) analyzed the parking demand based on a study of six town centers. Surprisingly, the experimental results reveal that on-street parking is the most favorable parking location amongst off-street surface, off-street garage, and on-street parking.

On-street parking is consistently the most in-demand amongst three opinions. Marshall et al. (2008) claimed that it is even true when the parking meter charges are higher for the on-street parking. The researcher explained that the shortest time allotment is the principal reason, which maintains the high demand in these areas. These data and conclusions further demonstrated the importance of on-street parking from the perspective of efficiency in land use. 

According to Byrd and Sisiopiku (2006), ‘on-street parking is also widely considered to have a significant effect on the pedestrian environment.’ They treated the on-street parking as a buffer zone between pedestrians and through traffic as it prevents people from vehicles’ sudden collisions whenever an accident occurs.(Byrd and Sisiopiku, 2006) However, according to the data published by HRB (1971), 16% of the crashes in American cities were directly related to vehicles parked on the street. Many studies have indicated a reduction in traffic collisions with the prohibition of on-street parking. Desjardins (1977) conducted research on the city of Hamilton, Ontario, where it was discovered that the non-intersection collision rate fell by 37% for a six-lane road after on-street parking was removed. The researchers discussed, as well as many other scholars have concluded that on-street parking is prone to traffic collisions, which is not safe. 

Nevertheless, Marshall et al. (2008) found an extraordinary phenomenon through his statistical analysis, namely that “the low-speed street with on-street parking had by far the lowest rate per mile of fatal and severe crashes.” It can be observed that the low-speed streets with parking only recorded 4% serious injuries amongst all the crashes, which is 6% lower than the low-speed streets without parking. (Marshall et al., 2008) It is less than half the low-speed street without parking. This indicates that on-street parking promotes the safety factor if the street is a low-speed environment. 

2.2 Influences of clearway 

The previous section examined traffic performance under the influence of on-street parking. The removal of on-street parking locations and creation of a clearway tends to increase the road capacity significantly, and the corresponding driving time will thus be reduced. The following sections evaluate the actual vehicle performance when the road is clearway.

2.2.1 Quantifying the influence of clearway 

Zhang and Excel (2011) used the micro-simulation approach to quantify the benefits of clearway. The model indicates that the travel demand is positively correlated with the travel time-saving. In addition, the shorter road sections are more sensitive to on-street parking compared with the long ones.(Zhang and Excel, 2011) The generated effects of the clearway are in the line with the results recorded by some other studies, as the clearway tends to mitigate the travel time. Models proposed by Weant and Levinson (1990) and Humphreys et al. (1978) eliminated all street parking zones and the simulated results indicated that the removal of on-street parking could bring significant benefits to the traffic flow.

2.2.2 Marginal effect of clearway 

Clearway provides smoother traffic conditions.(Zhang and Excel, 2011) With clearway, vehicles do not need to circumvent parked vehicles on the street. Fewer lane changes consequently occur. Moreover, since there is not much deceleration and acceleration, vehicles are more evenly distributed on the road. From a safety perspective, the researcher believes that clearway will offer drivers a broader view.(Zhang and Excel, 2011) However, clearway could also bring certain negative impacts, such as inconvenience for residual property owners and the loss of the business near the on-street parking zone. 

2.3 Cruising time for parking and the additional walking time 

Numerous articles (Guo et al., 2012; Portilla et al., 2009; Wijayaratna, 2015) criticize the on-street parking as it significantly increases the travel time and has been quantified various benefits after removing the on-street parking. However, when vehicles arrive at the destination, both on-street and off-street parking require extra Cruising and Walking. Therefore, when on-street parking is banned, travelers can only park their vehicles farther and possibly spend more Cruising and Walking

Back To News

2.3.1 Cruising time for parking 

Cruising is a driver’s searching time for a parking location. (Ommeren et al., 2012) It has a unique pattern over the day with the existence of rush hours in the morning and evening. (Ommeren et al., 2012) Within the twentieth century, travelers usually cruise for 3.5 to 14 minutes to find an on-street parking bay. In addition, 8% to 74% of the traffic was cruising for parking. (Shoup, 2006) In Shoup (2005)’s book of High Cost of Free Parking, the author mentioned the average search time before finding an on-street parking space in downtown Sydney is 6.5 minutes.

Moreover, drivers are willing to pay 3.5 times their wage rate to reduce cruising time.(Shoup, 2005) In contrast, a study that focused on the parking issue in the Netherlands revealed that the Tcruising was solely 36 seconds compared to the average 20-minute travel time. (Ommeren et al., 2012) This is primarily because on- and off-street parking in the Netherlands are approximately the same price in contrast to Sydney, hence, less Tcruising is required. However Shoup (2006) emphasized that even a limited amount of Tcruising can cause a considerable amount of traffic. Axhausen and Polak (1991) found that, in the UK and German, the ratio of TOD to Tcruising is 2 to 1. 

2.3.2 Additional congestion caused by cruising 

Shoup (2006) discovered that cruising behavior leads to a mobile queue that is waiting for vacant parking bays and is a key factor causing congestion. Meanwhile, Ommeren et al. (2012) reported that cruising could cause a negative effect on the traffic flow as it causes the following cars to slow down. Figure 2.1 was plotted by Geroliminis (2015) based on the macroscopic simulation. It highlights the relationship between the number of available parking bays (Np) and the total delay. It can be observed that the delay in travel time negatively correlates with the number of available parking bays.

When the number of available parking bays decreases, the corresponding total delay due to cruising phenomenon increases. It is because parkers need to travel longer to find an available parking space. This phenomenon further challenges the rationality of the clearway, because if all the on-street parking is removed, individuals will have to drive 

Figure. 2.1. Delays for different values of Np (Geroliminis, 2015) 

further, and the pressure on the traffic will be greater. 

2.3.3 Walking time 

The additional Walking is a considerable amount of cost for parking. Thompson and J. (1998) states the native cost associated with an off-street parking, involve both the monetary cost of parking fee and expected fine as well as the Twalking from the off-street parking to final destination. Arnott et al. (1999) also emphasize the Cruising and the Walking can be an appreciable fraction of TTT. In particular, in CBD, employed commuters are more likely to walk from where they park.(Arnott et al., 1999) 

More frequent parkers, who can be described as those with knowledge related to local regions, face the trade-off between the search time, parking price and location.(Bonsall and Palmer, 2004) For example, if a parker parks a vehicle at some distance farther from the destination. It can induce additional time consuming for walking and could also be further delayed by traffic lights. 

It is worth noting that different classes of parkers have different parking behaviors, people with higher income are less likely to choose a parking location, which would result in a long walking distance to their destinations.(Bonsall and Palmer, 2004) The phenomenon is also mentioned by Arnott et al. (1999), executives are willing to pay the premium to a parking location closer to their work. In contrast, middle-class travelers are more concerned about the parking price rather than the distance. 

In addition, Shoup (2006) stressed the importance of Walking from the parked vehicles to final 10

destinations as the parking locations are not always ideally located in relation to destinations. Shoup (2006) proposed that the off-street parking is less convenient than the on-street parking as the additional Cruising and Walking needed to access an above ground garage may be significant. Benenson et al. (2008) explored the relationship between the walking distance and air distance (the straight line distance between two points), and proposed that the actual walking distance between two points, which are located several hundred meters apart is 1.3-1.4 times larger than the air distance. 

Surprisingly, Benenson et al. (2008)’s study discovered that adding additional off-street parking locations in an area lacking parking bays (adding 200 parking bays in an area lacking 1200 parking bays) could lead to a minor improvement, namely 7% reduction in searching time and 12% in distance. Meanwhile, the author also emphasized that such an improvement is limited and short-lived, claiming that ’A new parking lot will hardly change the average resident’s perception of the parking situation in the area.’ This is consistent with the findings of Shoup (2006), who asserted that if the supply and demand ratio is greater than or equal to 1, the average Cruising and Walking would barely be improved. 

Marshall et al. (2008) examined more than 250 road segments in New English and discovered that on-street parking constituted approximately 11% at the traditional sites and 1.1% at contemporary locations. The traditional sites were fee-based, whereas the contemporary sites were free. The researcher discovered that all sites provided less parking than was required. However, the peak demand was still less than 80% of the provided parking bays.(Marshall et al., 2008) The findings were consistent with the case study conducted by Benenson et al. (2008) in Tel Aviv, Israel, which observed that approximately 20.8% of the parking bays were unoccupied during the peak hours. 

3 Methodology 

The investigation of travelers’ TTT consists of two parts. The first part is the travel time for a vehicle traveling from its origin to the destination (TOD). This journey is referred to as the ‘OD Trip’. The second part is referred to as the ‘Local Trip’ (Tlocal). It consists of the Cruising for a parking bay after reaching the final destination, and Walking from the parked vehicle to their destinations. The primary focus of this research is to quantify the negative impact of a clearway (Tlocal) when a portion of on-street parking locations are prohibited to create a clearway because researchers often overlook the additional time parkers spend after reaching their destination. The TTTs of both clearway and on-street parking situations are modeled and compared by applying the formula Equation 1. 

TTT = TO + Cruising + Walking (1) 

A parking searching model (PSM) was constructed using the Quantum Geographic Information System (QGIS) and Python 3 to model parkers searching for a parking bay in the ‘Local Trip’. Modeling for this process was based on imported GeoJSON files with both on-street and off-street parking distributions in Sydney. TOD was collected from the New South Wales Roads and Maritime Services (2015),Transport for NSW (2019) and KPMG (2017). 

Throughout the model, a parking bay is defined as a space in which a single vehicle can be parked and a parking location is an area that is composed of parking bays. The parking location can be either on-street or off-street. 

3.1 Data description 

The dataset used for the modeling process is a combination of three GeoJSON files, namely on- and off-street parking, floor space and employment block data. All three datasets were obtained from the City of Sydney Open Data Portal. GeoJSON is known as an open standard geospatial data interchange format that represents simple geographic features and their nonspatial attributes. (ArcGIS, 2019) The investigation of Tlocal for the Local Trip focused on the City of Sydney Region.(Figure 3.3a) The model is derived from the distribution of parking locations (Figure 3.1 and 3.2) and the employment rate within

(a) Heatmap of on-street parking distribution (b) Spatial distribution of on-street parking 

Figure. 3.1. Geographical distribution of on-street parking locations 

the extracted Sydney region. 

• On-street parking 

The on-street parking dataset comprises 1,361 on-street parking locations within the City of Sydney. Each parking location has between 1 and 30 available bays with a mean of 6.77. The file also includes attributes, such as payment methods, streets and suburbs for each parking location.(City of Sydney Data hub, 2019) Each location has a geographical latitude and longitude coordinates that can be used to pinpoint the location of these parking locations. 

In the spatial distribution map in Figure 3.1, the City of Sydney is encircled with a yellow line and data points congregate in the northern region. The first reason for this phenomenon is that the provided data only accounts for the parking locations that have a parking meter. Various sites in the lower half of the region that allow on-street parking have not been recorded because there are no meters present. The second reason is that the northern region is closer to the CBD. As demonstrated in the on-street parking heatmap (Figure 3.1a), the districts of Surrey Hill, Pyrmont, and Dawes Point have the most on-street parking locations available. 

• Off-street parking 


(a) Heatmap of off-street parking distribution (b) Spatial distribution of off-street parking 

Figure. 3.2. Geographical distribution of off-street parking locations 

The off-street parking dataset comprises 80 parking locations that are available for service and delivery drivers in Sydney’s CBD.(City of Sydney Data hub, 2019) The number of available parking bays for each location range from 14 to 1,153 with a mean of 277. Although the number of off-street parking locations is marginal in comparison to the other locations, the 80 off-street parking locations constitute 80% of the total available parking bays in the system. In addition, information about each parking point is expressed as a geometric coordinate and various features, such as the address. However, the provided dataset is incomplete; it excludes off-street parking locations that are not open to the public such as private parking locations at universities or hospitals. The spatial distribution of public off-street parking locations clusters in Sydney CBD is presented in Figure 3.2. 

• Floor space and employment block data 

The 2017 floor space and employment block data are captured by the Geographic Information System (GIS). This provides visualization and spatial analysis of the city’s floor usage and the employee numbers by block in the local government area. Blocks in GIS are expressed as polygons as coordinate members in GIS geometry are arrays of LineString coordinates in which the first and last points are equivalent. 

The GeoJSON database consists of 1,103 blocks. It contains information about 26,203 buildings, 23,511 businesses and approximately 883,000 spatial records.(City of Sydney Data hub, 2019) Figure 3.3 


(a) The region of employment survey blocks (b) The employee rate in each block 

Figure. 3.3. Floor space and employment survey block 

depicts the employee distribution in each block. Blocks within the Sydney CBD have the highest number of employees. This particular dataset enables researchers to estimate the parking demand for work trips in each block. 

3.2 Data processing 

Data processing includes transforming the collected parking and block data to a GeoJSON format that is suitable for Python 3 whilst ensuring data consistency and validity. The flowchart in Figure 3.4 presents the data processing procedure for constructing the PSM. Details of the model are outlined in section 3.3. 

3.2.1 Data Cleaning 

Data cleaning is performed using QGIS and Python 3. Each feature layer in GeoJSON contains several attributes. Attributes in the collected GeoJSON file that do not contribute to the study purpose are omitted as their presence may cause a disturbance in the simulation. The on- and off-street parking datasets allow definition of the parking location distribution and the corresponding number of parking bays. Attributes other than the parking ID and the number of parking bays are removed.

Figure. 3.4. Data processing procedures 

Moreover, the two parking layers of the same geometry type are combined into a single GeoJSON file through QGIS’s ‘merge vector layers’ function. However, QGIS only superimposes two files‘ feature layers and does not link attributes with the same meaning. For example, the number of parking bays is shown as both ‘total_number_of_bay’ and ‘approxPaybay’ for each parking location. The data is therefore further integrated in Python to store values with the same meaning under the same attribute. 

The floor space and employee survey data divides the City of Sydney into blocks and lists the number of employees in each block. Therefore, the block ID, the number of part time employees and the total number of employees are retained.(Figure 3.5) For each block, its centroid is pre-determined by the ’centroid’ function in QGIS. 

Figure. 3.5. Extracting attributes 

The block coverage area differs from the parking’s coverage area. The former one covers the entire City of Sydney whereas the latter one clusters around the regions near the CBD, where parking meters are present. The parts of employee blocks that overlap with the parking locations are extracted for use as the model study area. As mentioned previously, on-street parking without a parking meter and the private off-street parking are not recorded in the system. In order to minimize potential biases caused by unrecorded data, the modeling region was reduced to include only surrounding areas of the Sydney CBD. (Figure 3.6) In total, there are 260 blocks and 1,024 parking points, including both on-street and off-street parking. 

3.2.2 Data transformation 

Once the data for GIS files have been cleaned, the subsequent data transformation step is performed to derive the required data. This step involves calculating the block employee rate, assigning parking and establishing destinations in each block. These derived results are used to determine Parker’s potential destinations in the model. The shapely. geometry, NumPy and GeoPandas Python packages were applied to read GeoJSON format files, which enabled further analysis.

Figure. 3.6. The modeling area. The black and blue dots represent the on-street parking and off-street parking, respectively. The orange area is the simulation region 

Calculating employee rate The employee rate (Remployee) for each block was calculated by dividing the given number of employees in each block (Nemployeei) by the total number of employees in the system. The formula is presented in Equation 2: 

Employee =Employee 



Assigning parking locations For each block, the LineString points that define block boundaries are initially extracted to construct a polygon for each block ID using the shapely.geometry package. Subsequently, looping through all on- and off- parking locations and checking if any are in a block polygon using the ‘within()’ function. The parking location will be assigned the ID of the block in which it is located. Parking locations between blocks are allocated to the nearest blocks. The nearest block is the one with the shortest Euclidean distance between its centroid and the parking coordinate. The Euclidean distance is the ‘ordinary’ straight-line distance between two points. 

Setting destinations within blocks In terms of each block, rather than selecting the centroid as the block destination, a series of points were selected to form a group of destinations for each block. This avoids the phenomenon of considering the same point, namely the centroid as a destination after the parker arrives. In each block, four points from the existing LineString polygon array were selected to produce a destination matrix. Thereafter, the distance to all other block centroids were determined for each of the four destination points using the Euclidean equation. The nearest block to each of the four destinations was consequently determined, and this process was repeated for every other block. For each simulation, once all parking bays in the current block are fully occupied. It was no longer necessary to re-calculate the distance between all blocks to define parkers’ next target block. 

3.3 Parking searching model 

The PSM is built in Python based on the formulated parking and block GeoJSON files. The scope of the model is the city of Sydney. (Figure 3.6) The primary purpose of this model is to simulate Cruising and Walking after a vehicle arrives at its destination. The clearway was constructed by removing a certain percentage of on-street parking locations. This enables comparison of Tlocal for different clearway cases with the on-street parking, whilst this comparison process is repeated under different parking demands. The model procedures are shown in Figure 3.7 and explained in detail in the following sections. 

3.3.1 Assumptions 

• Firstly, the simulation model assumes that the number of vehicles in the system does not exceed the total number of parking bays and that all bays are initially unoccupied. 

• In addition, the number of parkers in each block is proportional to the corresponding employee rate. 

• In the theoretical model, it is assumed that the spatial distribution of off-street parking is sufficient to satisfy the city’s parking requirements. Therefore, on-street parking can be eliminated. 

• All parkers know the exact route from the departure point to the destination; thus no additional time is wasted on searching for the destination. 

Figure. 3.8. Rate of parkers entering the modeling region 

are assigned ‘0’ (step 6). In the model, on-street parking is the regular situation, in which all parking locations are accessible to the public. In contrast, the clearway situation is simulated by omitting a portion of random on-street parking locations (step 7). 

The parking supply is the number of available parking bays remaining in the system after the clearway ratio (ρ) is implemented. It is crucial to confirm whether the input Q is smaller than the parking supply as the supply has been reduced for the clearway (step 8). If Q exceeds supply, the simulation will be halted to avoid the infinite loop because parkers that exceed the supply will continue to search for a parking bay in the system. 

If supply exceeds Q, parkers will be assigned to the study region. The model simulates the Q within a three-hour period from 6:00 am to 9:00 am with the highest Q at 8:00 am (step 9). Figure 3.8 shows an example of the distribution of parkers who enter the study region. The Q increases progressively from 6:00 am and reaches the maximum at 8:00 am. Thereafter, Q declines as time approaches the end of the peak hours. 

3.3.4 Parking searching behavior 

The modeling region comprises numerous irregular blocks. Each block contains four predetermined destinations. When a parker enters the study region, it is randomly distributed into one of the blocks according to the blocks’ employee rate (step 11). For example, if block i has an employment rate of 5%, a Parker will have a 5% probability of being distributed into this block. Once a parker reaches the assigned block, the parker randomly selects one of the four predefined points as the final destination (step 12).

Setting a series of destination points in a block has its advantages. The presence of four destinations enables the distribution of vehicles to various locations in a block. The searching strategy is to find the nearest available parking location. The consequence of clustering at one particular destination is that the parking location within the closest proximity is always the first to be occupied. Moreover, establishing more locations as destinations enables more realistic modeling as the selected four points are boundary LineString points that define each block. Parkers tend to start searching for available parking bays from the moment they enter the block. 

Once vehicles reach their final destination, the model classifies parkers into two categories based on the specified part-time employee rate for each block, namely part-time and full-time employees (step 13). In terms of the first block that parkers enter, part-time employees prefer on-street parking (step 14), whereas full-time employees tend to prefer off-street parking (step 15). This is because full-time employees travel daily to their workplaces and those who drive are more likely to park their vehicle for the whole day. The off-street parking fee is cheaper for regular users than the on-street parking zones. 

Conversely, for casual users, such as part-time employees, on-street parking is less expensive than off-street parking. During the peak hour, the highest on-street parking fee is $7.40 per hour. Figure 3.9 The lowest price for casual parking in off-street parking locations starts from $10 per hour. 

When a parker reaches the closest parking location (A) near the final destination (D), the parker verifies the UA(t) of location A (step 16). If it is smaller than 1, the parker parks the vehicle and the model systematically adds 1 to the parking location A’s UA(t) (step 18).

The model simultaneously calculates the cruising distance and walking distance of AD. If location A’s UA(t) is greater than or equal to 1, the parker has to find the next closest parking point B, to location A (step 17). If point B has a UB(t) smaller than 1, the vehicle can be parked. Subsequently, the model calculates the accumulated distance DA + AB and walking distance DB. If the UB(t) of point B is equal to 1, the parker must continuously search until he or she locates an available parking bay (step 21). Parkers only can enter the next closest parking block when all parking bays in the current block are fully occupied (step 20). 

It is worth acknowledging that the searching behavior for parkers in both on-street parking and 23

Figure. 3.9. On-street parking price (City of Sydney (2017)) 

clearway situations are identical. The only difference is that a percentage of on-street parking locations is removed to construct the clearway situation. In addition, the parking location preference from different groups of employees only exists in the first block where they arrive. If parkers cannot successfully locate a parking bay within their destination block, they will park their vehicles in the next closest available parking bay they encounter.

Furthermore, the objective of the present study is to investigate the difference in time spent for Tlocal for both on-street parking and clearway situations. When the parking demand is high, the model’s searching behavior may result in some vehicles cruising an extensive distance as their drivers search for an available parking bay. However, the difference in Tcruising for both on-street parking and clearway situations remains for investigating the effect of clearways. Moreover, if a portion of the on-street parking locations is removed, they will completely disappear from the system. 

3.3.5 Travel time calculation 

In the PSM, the parking locations and potential destinations are being accurately pinpointed on the map according to their longitudes and latitudes. If the Euclidean metric formula is applied to calculate the cumulative vehicle cruising distance, the distance error will be significant. Consequently, the Manhattan distance is used to approximate the actual distance that a vehicle travels. As shown in Figure 3.10, this is the distance between two points in a grid based on a strictly horizontal and vertical path rather than the diagonal distance.(Teodorovic and Janic, 2017) Equation 5 depicts the Manhattan distance formula: Õn 

d1(p, q) = ||p q||1


where (p, q) are the geometric points 

|pi qi| (5) 

Figure. 3.10. Example of cruising and walking distance using the Manhattan distance The black dots are parking locations. The solid black line is the Manhattan distance. The solid red line is the Euclidean metric distance. The dashed blue line is the straight-line air distance. 

Figure 3.10 demonstrates the difference between the cumulative Euclidean metric and Manhattan distances using parking locations in the system. In reality, parkers walk from their parking location back to their workplaces. Pedestrians are not as tightly constrained by road systems as vehicles, and their trajectories are relatively casual. According to Benenson et al. (2008), the walking distance is approximately 1.3-1.4 times larger than the air distance. Hence the straight-line distance is multiplied by a factor of 1.3 to give a reasonable approximation for the walking distance. 

With regards to the cruising speed, Belloche (2015) demonstrates that the average search speed is nearly constant at approximately 11 to 15 kilometers per hour (km/hr). The average human walking speed is approximately 1.4 meters per second (m/s).(Marshall et al., 2008) In order to account for the effect of traffic lights, the pedestrian walking speed is recognised as 1.2 m/s or approximately 4.3 km/hr. The vehicle’s cruising speed is presumed to be 13 km/hour.

3.4 Origin to destination travel time 

TOD for each Q and ρ is approximated based on numerous journal articles, such as Wijayaratna (2015) and Guo et al. (2012). Even though the destinations of parkers are determined in PSM, whereas their origins are unknown, which makes TOD hard to be approximated. In this research, the origin is defined as one of the polygon points of the simulated region that a parker first arrives at. The distribution of the origins is determined through the Traffic Volume views provided by the NSW Transport service.

As it can be seen in Figure 3.11, the region surrounded by eight vehicle counters is similar to the PSM region. These points are then pinpointed to the nearest boundary points in the PSM region. Once a parker is about to enter the simulated region, it is distributed to one of the counters. Therefore, the origins of parkers can be determined as the travel distance for each parker. As a result, the average travel distance for each demand can be consequently calculated. 

Moreover, the parking demand (Q) represents the number of total parking bays being occupied for a simulation, and it is assumed to be proportional to the vehicle flow of the system. Therefore, different Q represents the traffic flow at a different time of the day. For example, the 20% demand is highly likely to occur at 4:00 am, whereas the 80% demand is more likely to occur at 8:00 am. The approximated Q and the corresponding time of the day are shown in the appendix. 

Based on the corresponding time of the day for each Q, the google map allows the travel time at a different time of the day (different Q) to be predicted with input origins and destinations. For each Q, the origins are set at eight different starting points. The destination is set towards the CBD based on the generated travel distance (as the CBD clusters with the most number of employees). This process is repeated for eight origins. Afterwards, the average TOD for the baseline on-street parking of each Q can be predicted. The TOD at each clearway ratio can be approximated through applying the travel time reduction factor from Wijayaratna (2015) and New South Wales Roads and Maritime Services (2015). 

Figure. 3.11. Distribution of traffic volume viewers Transport for NSW (2019) 

It is worth mentioning that the TOD required in this research is at a microscopic level within a network. The primary purpose of this study is to build a PSM model to quantify the negative impact of clearway. Moreover, at the current stage, the clearway benefits that have been investigated at a network level are insufficient. The approximate TOD might be inaccurate in the Sydney region. However, it does reflect the benefits of the clearway. The given reduction factor in the term of different clearway ratios(ρ) is on a single road segment; it assumes the factors also could be applied in Sydney. 

4 Results and Findings 

The parking searching model (PSM) simulates the Tcruising for parkers to find a parking bay with different parking demands (Q) and Clearway ratios (ρ). It also considers the Walking for parkers to walk from the parking location to their final destination. In summary, the three output datasets are as follow: 

• The rate of parkers entering the modeling region for each minute in a three-hour scope. • The average cruising time taken for a parker to locate a parking bay, Tcruising 

• The average walking time between the parking bay and the destination: Twalking 

These three sets of data are collected under 9 different parking demands (Q) and each corresponds to 11 clearway ratios (ρ). 

It is worth reiterating that the on-street parking case is where all parking supply is available. The clearway case is obtained by removing different proportions of on-street parking data points in the model. In the PSM, different parking demands (Q) correspond to different numbers of parkers who enter the simulation region, whilst various on-street parking elimination ratios represent different clearway ratios (ρ) in the system, each with its own parking supply. For example, where ρ is 1, it means that there is no on-street parking in the road network, all roads are clearways. However, where the ρ is 0, it is the baseline on-street parking case, in which on-street parking supply is all available. 

In addition, where Q is 0.6, it indicates that 60% of the current available parking bays will be occupied at the end of the simulation period. Figure 4.1 further confirms whether the ρ in different clearways satisfies the Q

This section is divided into four subsections. Firstly, Section 4.1 compares the trend of Cruising and Walking across different Q and ρ. Secondly, Section 4.2 quantitatively analyzes the impact of the clearway when the Cruising and Walking are combined. Thirdly, Section 4.3 further evaluates the effects of clearway on parkers for different time segments in the three-hour time frame. Lastly, Section 4.4 combines the Tlocal with the TOD to evaluate the TTT

Figure. 4.1. The satisfaction of parking supply(ρ) in different demand (Q) ; Red: Q > ρ, Green:Q < ρ 

4.1 Walking and cruising time tendency 

This subsection focuses on the trends in both Cruising and Walking within the simulation time. In different Q and ρ scenarios, the Cruising and Walking for each minute are plotted over 180 minutes. As stated, the three-hour simulation time frame (6:00 am to 9:00 am) is divided into 180 time steps. Figure 4.2 presents the tendency of cumulative Walking whilst Figure 4.3 to 4.7 depicts the tendency of the average Cruising and Walking per minute over the measured period. 

Figure 4.2a plots the total Twalking for all parkers within different minute steps and Figure 4.2b depicts the cumulative Walking for each the corresponding time point. It is apparent that when Q is reduced, the total Walking is also reduced correspondingly. This indicates that a parker has a greater probability of locating a parking bay near their destination. 

In Figure 4.2a, Walking initially increases, attaining its peak at approximately the 130th time step (8:10 am) before decreasing. It is consistent with the parking demand distribution in Figure 3.8. As the number of vehicles entering the system per minute increases with time, the difficulty of locating a parking bay increases. Parkers tend to park their vehicles further away from their destinations. After the peak at 8:00 am, although the Twalking of each parker continues to increase, the slope begins to decrease due to the decrease of Q

It is worth noting that the peak value exhibited by Figure 4.2a is shifted approximately by 10 minutes 29

(a) The cumulative Walking for each minute (b) The cumulative Walking for the peak period 

Figure. 4.2. The cumulative Walking for parkers of each time step and the peak period 

to the right in comparison to the previous parking demand distribution in Figure 3.8. This is due to the lagging effect. The impact of the peak period will become predominant after it terminates. As the parking occupancy increases dramatically after the peak hour, parkers have to spend more time on searching for a parking bay, which potentially causes them to park further away from their destination. Meanwhile, the parking demand (Q) has not been significantly reduced. Hence, the total Walking continues to increase and reaches its maximum after the 8:00 am peak hour. 

Figure 4.2b demonstrates an S-curve shape, the slope is relatively flat at the beginning and end, whilst the middle is steep. The middle part of the S-curve corresponds to a greater Q. Moreover, in higher Q cases, cumulative Walking increases at a much faster rate than it does in the lower Q cases. The difference in cumulative Walking becomes more significant after the 60th minute. The greater Q has a more pronounced increase. The time difference between Q increases significantly and it finally stabilizes after the 160th minute. 

Figure 4.2 is an example of many different cases, but the trends are expected to be repeated in cases with different clearway ratios (ρ). 

Figure 4.3 is the on-street parking baseline case. The implication is that Figure 4.3 introduces Cruising and Walking of parkers under different parking demands (Q), based on the existing Sydney parking space supply. Then, commencing with the Q presented in Figure 4.4, the model begins to simulate Cruising and Walking in different clearway ratios (ρ). This is compared with the baseline case to ensure that the relationship between the two is more intuitive.

(a) Tcruising Baseline: On-street parking (b) Walking baseline: On-street parking 

Figure. 4.3. Cruising and Walking: On-street parking 

In Figure 4.3, Parker’s Cruising and Walking rise as time proceeds. However, within the first 80 minutes, both Cruising and Walking are relatively flat and exhibit no significant difference. This is because the model assumes all parking spaces are initially unoccupied and parkers who enter the system initially will easily find a parking space. In both Figure 4.3a and 4.3b, slopes of 80%, 75%, and 70% Q start to become steeper after 80 minutes, at which the occupancy rate in the three Qs is approximately 30%. It can be concluded that, under the current Sydney parking supply (ρ = 0), when Q is higher than or equal to 70%, and 30% of the parking locations has been occupied, the difficulty of finding parking spaces begins to increase significantly. 

The 80% Q seen in Figure 4.4 corresponds to the Q during Sydney’s daily peaks. According to Shoup (2006), nearly 20% of the parking spaces are unoccupied even in the peak period. In Figure 4.4a and 4.4b, a higher ρ corresponds to a higher Cruising and Walking value throughout the 180 minutes, in particular after the first 80 minutes. With the existence of a clearway, more and more parkers are being affected. When removing a portion of the on-street parking from the system, it forces drivers to seek parking further away from their destination as the previous on-street parking locations are prohibited. Parkers also have to walk further to reach their final destination. 

When the Q is at 80% and ρ is over 70% , If there is no additional off-street parking is being implemented, then the parking supply will be just enough to meet the Q. Ideally, the occupancy of parking facilities should be sufficiently high to ensure that they are occupied to a level that justifies the supply but does not make it unreasonably difficult to find a space. Generally speaking, parking is considered ‘at capacity’ when available spaces are 85% occupied at times of peak demand. (Shoup, 2006) This indicates that approximately one out of seven parking bays should be available. 

Potentially long Cruising and Walking for a parking space may result in enhanced driver frustration and congestion. Initially, this might make parkers more willing to risk parking illegally, which will cause even greater congestion. In summary, if a full or nearly-full clearway ratio (ρ) is to be implemented, the number of the off-street parking bays cannot fail to satisfy the current current peak hour Q

With regard to the output Walking in the 80% Q. The different ρ did not result in significantly longer Twalking than on-street parking for the first 25% of the measured time (1 minute to 45 minutes). However, from the 80th minute onwards, especially in the last 25% of the time period (135 minutes to 180 minutes), the impact of clearway becomes much more significant. 

Figure 4.5 exhibits a similar tendency as Figure 4.4. The initial phase with a relatively flat slope in Figure 4.5 is extended, in comparison to the 80% Q. This demonstrates that, as Q decreases, the number of vehicles affected by the clearway decreases. As such, when less on-street parking is prohibited, parkers can locate a nearby vacant parking bay without long distance cruising. 

When Q continues to drop down to 60%, parkers spend no more than nine minutes for Walking. (Figure 4.6) The Walking distribution in the appendix reveals approximately 28 time steps with a Walking within four to five minutes for the baseline case (ρ=0). If ρ equals 20%, the total number of Walking located within four to five minutes increases to 75 time steps. If ρ is 100%, there are approximately 15 time steps that allow parkers to walk less than two minutes. The quantity is significantly reduced. At 

(a) Tcruising: 70% parking demand (Q) (b) Twalking: 70% parking demand (Q

Figure. 4.5. Cruising and Walking: 70% parking demand (Q

At the same time, 25% of the time (approx. 45 minutes), parkers have to spend more than six minutes walking. The specific walking time distribution is demonstrated in appendix Figure 7.1. 

(a) Tcruising: 60% parking demand (Q) (b) Twalking: 60% parking demand (Q

Figure. 4.6. Cruising and Walking: 60% parking demand (Q

Figure 4.7 shows the 50% Q. Its Cruising and Walking follow a different trend from previously presented trends in the other four Qs. The Cruising and Walking still increase with time, but the increase rate is not as significant as previously. As such, the time variation in different ρ becomes insignificant. At the beginning, the 50% Q case demonstrates almost the same cruising and walking trends as the other demand cases, as its time slope is initially flat. However, further along in the simulation, the difference between different ρ is still not significant. 

When the Cruising and Walking are compared, Cruising is less than Walking at the beginning as most parkers can find a parking space in less than a minute. However, at the end of the modeling period, it 

(a) Tcruising: 50% parking demand (Q) (b) Twalking: 50% parking demand (Q

Figure. 4.7. Cruising and Walking: 50% parking demand (Q

increases more rapidly than the Twalking. According to the search behavior in the model, parkers tend to search for parking bays closest to their current location. As such, they are liable to cruise around their destination – even going back and forth – until they find an unoccupied bay. Consequently, the cumulative Cruising increases whilst the Walking remains relatively short. 

From the apparent trends seen across six different cases, the following conclusions can be drawn. • The total Cruising and Walking decrease as the parking demand (Q) decreases. • The peak demand has a lagging impact. 

• All clearway ratios (ρ) in all parking demand (Q) cases have a higher Cruising and Walking than the on-street parking case. 

• The later a parker enters the system, the more significant the impact of the clearway. • Cruising is initially shorter than the Walking but ultimately becomes longer. 

• Lastly, the impact of clearway becomes less significant at 50% parking demand (Q). 

4.2 Interaction effects 

In terms of the combination of Q and ρ, the average Tlocal that each parker spends after reaching their final destination is plotted in Figure 4.8. Figure 4.9 demonstrates the extent to which each ρ is greater than the on-street parking baseline case. It is worth acknowledging that the time taken for the parking maneuver is not taken into consideration because the PSM assumes that the time spent on the parking maneuver is the same in both on-street parking and clearway situations. 

Figure. 4.8. Average Tlocal for each parking demand and clearway (minute) 

In Figure 4.8, it is apparent that in low Q of 20% and 30%, the difference between the on-street parking and clearways is not significant. Nevertheless, the clearway cases still tend to have a longer Tlocal. This is because when the Q is low and ρ is high, the unoccupied parking bays usually cluster at several off-street parking locations. It causes parkers to cruise a longer distance to reach one of them. As a result, a longer Tlocal is presented by clearways even with a small Q. For Q higher than 50%, the slope is steeper and exhibits a larger growth tendency. This further verifies the finding that Tlocal of higher Q increases at a faster pace. 

Figure 4.9 plots the Tlocal increment percentage against Q for each ρ. For higher ρ, such as the 100%, the corresponding Tlocal increment is higher. It represents that the clearway impact will be further emphasized at a larger Q, which is the higher traffic flow. 

As a result, the following conclusions can be observed: 

• When the parking demand (Q) rises, the range of Tlocal becomes greater and the extent to which Tlocal rises will also increase. 

• Provided that there is a clearway, the Tlocal will increase regardless of the parking demand (Q) or clearway ratio (ρ) 


Figure. 4.9. Tlocal increment percentage 

An additional observation worth acknowledging indicates that vehicles that enter the system in the earlier stages are less affected by the clearways than those that arrive later. Although the average Tlocal is able to reflect the impact of clearways, the generated result is not comprehensive as it varies in different time segments, as demonstrated in Section 4.1. As such, the discussion will presently focus towards the difference between clearways and on-street parking during different time segments. 

4.3 Tlocal in different time segments 

The PSM simulates a three-hour time frame. The number of parked vehicles varies at different points of time. Figure 4.10 depicts the Tlocal, ρ, and Q in subsequent hour-long intervals. When the three figures are compared, the gap between graphs increases significantly from the first hour to the third hour. This correlates with the trend presented in Section 4.1. Furthermore, the change in Tlocal caused by different ρ within the first hour is not obvious. As stated, the model assumes all parking bays are initially unoccupied. In terms of parkers entering into the simulated region at an earlier stage, they can immediately locate a parking bay near their destination. The graphical upward trend becomes obvious for the second and the third hours. 

A closer perusal of the graphs reveals that in the first hour, Tlocal remains under three minutes regardless of the Q and ρ. This value is much smaller than that of the average case. In the second hour, time begins to rise slowly. The last hour, however, reveals a much larger Tlocal than the average. This verifies the 

(a) Tlocal: The 1st hour (b) Tlocal: The 2nd hour 

(c) Tlocal: The 3r d hour 

Figure. 4.10. Tlocal in different time segments 

results from Section 4.1, which suggests that the time slope in the later stages of the simulation is much steeper than in earlier stages. In terms of the third hour, taking the most extreme Q (80%) as an example, Tlocal has a value of 13 minutes in the on-street parking case (ρ = 0). In the 100% clearway case, its Tlocal closes to 18 minutes. 

The results observed can be summarized as follows: 

• Initially, drivers can easily find parking spaces across various parking demands (Q). 

• Parkers will be affected by clearway ratio (ρ) and parking demand (Q) to varying degrees depending on the time they are parking. 

• Lastly, the Tlocal slope in the later stages of the simulation is much steeper than in earlier stages. 37

4.4 Total travel time 

In the combination of the collected and approximated TOD, the TTT for each Q and ρ is listed in both Table 7.9 (Appendix) and Figure 4.11. In Figure 4.11, an unexpected outcome is observed when the Q is at 20% and 30%, namely TTT increases as the ρ increases. This type of Q usually occurs during the early hours of the night, between 3 a.m. and 4 a.m. Vehicles are more likely to travel at the free flow speed, which is the road’s speed limit. The clearway’s benefit is negligible, as there is initially no congestion on the road. It is because the benefit of the clearway is mainly generated through mitigating road congestion. 

Figure. 4.11. TTT for each demand and clearway case 

Even under such low Q, the Local time of some parkers will still be affected by the clearway. If the parking location nearest to their destination appears to be prohibited, parkers will still need to travel further to search for the next parking bay. This leads to a phenomenon in which, at very low Q, clearways have not only failed to reduce travelers’ TTT, but can, in fact, increase their TTT

When the Q reaches 40% and 50%, the clearway benefits gain significance. The TTT decreases as 38

the ρ increases to a point. This indicates that the positive impact of the clearway is more significant than its negative impact for a certain range. The increase in the road capacity mitigates the congestion in the system. 

As Q grows, the reduction in TTT also increases. For example, each 10% increase in ρ corresponds to the time saving, which increases from a period less than one minute under 40% of Q to a period greater than one minute under 60% of Q. When the traffic flow increases, more vehicles encounter congestion. Therefore, an increase in road capacity can benefit more vehicles. 

However, Figure 4.11 indicates that the relationship between the TTT and the ρ is non-linear. From 40% Q onwards, there is a presence of one optimized point that changes the decrease tendency of TTT into an increased tendency. This is because when more on-street parking is removed, the TOD will continue to decrease due to greater road capacity. When the ρ exceeds the optimized point, the time saved on the TOD is no longer sufficient to compensate for the Tlocal. Consequently, it results in the phenomenon that an increase in the ρ leads to an increase in the TTT

The optimized point occurs earlier as the Q increases. At 40% Q, the turning point occurs at 90% ρ. Whereas at 60% Q, it occurs at 80% ρ. The optimized point resembles a step function. It is primarily because, when the Q is low, the total number of parkers entering in the system is small. A large portion of the parking spaces can be removed and it still does not have a significant impact on parkers because the parking supply is significantly larger than the Q. Parkers do not need to cruise for substantial periods of time to locate a parking bay.

Therefore, it can tolerate a larger ρ to prevent the additional Tlocal from exceeding the saved TOD. However, when the Q is high, even a small ρ present in the system can lead to a significant impact. As the parking supply becomes saturated with the increased Q. The difficulty of finding a parking space occurs at a smaller ρ. As a result, the clearway optimized point becomes smaller as the Q increases. 

The results of this subsection can be summarized as follows: 

• Firstly, when the parking demand (Q) is low, the implementation of clearways will cause the increase of the TTT

• Firstly, as the parking demand (Q) increases, the increment of TTT for each 10% clearway ratio (ρ) also increases. 

• The optimized points for parking demand (Q) at 40% 50%, 60%, 70% and 80% are at 90%, 90%, 80%,70% and 60% clearways ratios (ρ), respectively. 

• Lastly, the optimized point occurs at a smaller clearway ratio (ρ) when the parking demand (Q) increases. 40

5 Discussion 

The PSM model primarily quantifies the negative impact of on-street parking. Section 1 proposes all the questions that drive the research. The results and corresponding analysis are presented comprehensively in detail in Section 4. The discussion section is composed of two sub-sections, namely the first section which discusses the feasibility of the clearway implementation in reality and the second section which explains the complications of the PSM. 

5.1 Feasibility analysis on clearway implementation 

According to the results in Section 4, at different Q, the implementation of various ρ leads to different TTT. Q differs according to the time of day. From the travel time perspective, the clearway should only be implemented at particular time intervals. As concluded in Section 4.4, when Q is low, it is counterproductive to implement the clearway as drivers’ TTT will increase conversely. When there is no congestion, vehicles drive at free flow speed. Under this circumstance, the clearway no longer reduces TTT, but it does increase Tlocal. Therefore, clearway implementation can only reduce TTT if congestion exists in the network. 

With regards to parkers who prefer to park their vehicles on the side of the road, a decrease in the number of available parking bays results in an increase in the time spent searching for a parking bay; as a result, the resulting traffic congestion is greater as well.

However, when all on-street parking is removed in a particular block to create a clearway zone, the congestion caused by cruising essentially decreases. At the start of each clearway zone, a clearway warning sign is placed before the entrance. Therefore, parkers do not slow down to search for a parking bay in this block. Instead, they immediately drive to another block, where available parking bays may potentially exist. In reality, in comparison to the 90% ρ, the full ρ might result in a significant decrease in TTT. In summary, rather than partially implementing clearways within a block, it is best to construct a full clearway zone. 

It is worth acknowledging that TTT is not the only standard that should be used in determining the feasibility of the clearway. Factors such as accessibility to local businesses and parking supply should also be taken into consideration. 


Parking supply This research provides Sydney with transport suggestions for clearway implementation, particularly in the CBD district. When the city government intends to remove on-street parking to implement the clearway, it first needs to evaluate whether the existing parking supply is able to meet the Q. According to this research, when ρ constitutes over 40% of the road segments, the existing parking supply cannot meet the Q

Local business accessibility Some on-street parking locations serve more than just parking purposes. For example, a reduced number of parking bays may potentially lead to reduced business accessibility, especially those without designated off-street parking locations. With regards to local businesses that depend on on-street parking, having available parking bays nearby can potentially create business opportunities. 

When a clearway is implemented on a road segment, it is critical to quantify its negative impact on Cruising and Walking, and compare it with the benefits generated from the increased road capacity. It is also imperative to balance Q whenever parking is removed to introduce a clearway. As such, before removing parking, an alternative nearby parking location that can meet current parking demand should first be identified. 

This research can assist urban planners in two ways. Firstly, it can enable them to re-define the meaning of TTT. Secondly, it can help them obtain an in-depth understanding of the negative impact of clearway implementation, which include increased Cruising and Walking for park workers. Ultimately, these insights should be used as part of a comprehensive consideration of clearway implementation, one that also considers parking supply and demand at different times of day as well as the unique needs of different groups impacted by parking availability. 

5.2 Complications 

The parkers’ cruising behavior is far more complex than a single model can portray. There are many factors, such as the parking price and the value of travel time (VTT), that need to be taken into account. This section will suggest several undiscussed complications and illustrate the shortcomings of this model. 

Additional traffic congestion caused by vehicle cruise and parking maneuvers This cruising effect is not only imposed on the parkers, but also on commuting vehicles. The most direct impact is a reduction in driving speed. However, the relationship between the congestion level and a vehicle’s cruising speed is not studied in the model. The congestion influence is neglected and assumes the cruising speed remains constant at 15 km/hr. 

Furthermore, for each vehicle that is being parked on the street, their parking maneuvers cause traffic blockages. When parkers perform a parking maneuver, they obstruct one of the lanes on the road. For park workers themselves, it takes time to conduct the parking maneuver, which should also be considered as an aspect of the TTT. In the PSM it is assumed that once a vehicle finds an available parking bay, they would immediately park the vehicle. 

The value of travel time VTT refers to the cost of time spent in transport. It is not a constant variable; different people place different values on time saving at diverse times, in different locations, and for different trips. High VTT includes paid travel, urgent personal trips, travel under congested conditions, unexpected delays and long trips. Even within a single trip, the VTT will still vary with regards to time.

The more time Parker spends on cruising, the more likely they will arrive at their destination late. The desirability of a parking bay increases and, as a result, they might risk parking illegally. For these travelers, they are more willing to finish the trip as soon as possible and no longer follow any parking preference. The simulation model did not distinguish between parkers with different VTT, nor assign any priorities to people with potentially high VTT. The model merely divided parkers into part-time and full-time travelers. 

On-street parking and off-street parking are not perfect substitutions for one another. In the clearway example, travelers either park at a further on-street parking location or an off-street parking lot. However, the on-street and off-street parking are not perfect substitutes for one another as they have different characteristics. 

1. Off-street parking is less convenient than the on-street parking because the additional time required to access a higher-level floor might be significant and vehicles need to cruise between floors to find 

a parking bay. However, cruising within a building does not cause congestion on the roads. The cruising time within a parking location is not considered in the model. 

2. From both the perspective of construction costs and charging fees, on-street parking is less expensive than the off-street options. 

3. On- and off-street parking serve different groups of parkers. On-street parking usually serves the casual parkers; but for those who wish to park for an extended time, they normally select off-street parking. 

On-street parking is limited by the length of stay On-street parking is limited by the length of stay. Within the simulated region, the City of Sydney, the two-hour parking control is implemented for on-street parking locations. However, a survey conducted by Shoup (2006) indicated that the time constraints are difficult to enforce. The survey results revealed that more than half of the vehicles either violate the time restraints or park in an illegal location. (Shoup, 2006) Within the model, the time limitation on on-street parking is not considered. The model does not quantify the portion of parkers that only park for a short period of time. 

Illegal on-street parking When the time is limited or travelers are unable to find a parking space, there is an option beyond the simple choice of cruising. Travelers also can park in an illegal on-street parking bay and risk receiving a parking ticket. Vickrey (1994) stated that the illegal on-street parking could generate more search traffic and waiting cars than spaces in off-street parking locations. Illegal parking is random in space and time. It may result in congestion caused by the reduction in road capacity and the reduction in the speed of the traffic flow. 

Model imperfections In addition to the aforementioned factors, the model itself also has imperfections when demonstrating the process of finding a parking bay. Parkers have imperfect information in the system; they do not know in advance how long it will take to find a parking bay, and they lack information regarding the parking availability in each location.

Therefore, the time it will take to find a parking bay is a random variable with unknown distributions and expected values. Whereas, the model assumes parkers do have accurate information about the location of the next nearest parking location, and the searching behavior always compels them to travel there. In reality, it is not the case; parkers do not follow such cruising patterns, rather, vehicles will be cruising casually on the roads. 

For some employees, they park their vehicles on a daily basis. Therefore, they do not need to cruise to know where to park their vehicles. Nonetheless, the model does not distinguish between the purposes of different trips; it only classifies employees as part-time or full time. 

Drivers do not use mathematical models when determining where to cruise if the current parking location is unavailable. A model cannot perfectly predict how individual parkers behave nor comprehend precisely known reality functions. However, all model assumptions are reasonable. The PSM does simulates and mathematically quantifies the negative impact of clearways and on-street parking under numerous parking demand (Q) and clearway ratios (ρ). 

6 Conclusion 

6.1 Summary of the present study 

This research, for the first time, evaluates clearways considering not just travel time savings for motorists, but also time losses for park workers who now must search longer for parking, park farther away, and walk longer distances. This study proposes a parking searching model based on the parking distribution and employment rate within the City of Sydney to investigate the effect of clearway versus on-street parking on travel time. The model simulates the vehicle cruising behavior while searching for a parking bay after reaching the final destination in a three hour time frame. The simulation is repeated for different parking demand and clearway cases. The generated vehicle cruising time and the parkers’ walking time is further evaluated with the collected origin to destination travel time to investigate the significance of the clearway. 

Within the three-hour simulation, all ρ exhibited a longer Cruising and Walking than on-street parking. The total Cruising and Walking decrease as the Q decreases. When the Cruising and Walking are compared, the Cruising is less than the Walking initially. At the end of the modeling period, Cruising increases more rapidly than the Walking

With regards to the interaction impact of Cruising and Walking. As the demand rises, the range of 45

Tlocal increases and the extent to which Tlocal rises will also increase. At the beginning of the simulation, parkers can easily locate a parking bay across various Q. The Tlocal slope becomes steeper in the later stage of simulation. Moreover, parkers will be affected by ρ and Q to varying degrees depending on the time they are parking. 

When the Q is low, the implementation of clearway will cause the increase in TTT. The optimized TTT occurs at a smaller ρ when the Q increases. Ultimately, it will be concluded that the presence of a clearway does not always benefit the traffic network and mitigate the TTT

6.2 Directions for future research 

There are other factors, such as the value of time and the price of parking, which should be taken into consideration in order to provide a more realistic model. The methodology and techniques employed within this research to assess the distribution of parking demand can ensure future and indeed ongoing analysis of parking preferences based on trip purpose, as different trip purposes would lead to different requirements in terms of parking bays. The impact of the clearway should be tested in other cities to obtain a more comprehensive view of the negative impact regarding clearways. It is also worth comprehensively investigating the different types of clearway distributions, rather than solely examining the percentage of clearways in the system, which could maximize the road capacity whilst minimizing the cruising and walking time. 

Moreover, due to the insufficient parking supply for the peak hour demand when the network is entirely clearway, more effort should be attributed to investigate the precise level of clearways that require application at different time periods. 


AASHTO (2011). A Policy on Geometric Design of Highways and Streets (6th ed. ed.). Washington DC: American Association of State Highway and Transportation Officials. 

Arnott, R., A. d. Palma, and R. Lindsey (1999). A temporal and spatial equilibrium analysis of commuter parking. Journal of Public Economics 45(3), 301 – 335.

Axhausen, K. and J. Polak (1991). Choice of parking: Stated preference approach. Transportation 18(1), 59,81. 

Belloche, S. (2015). On-street parking search time modeling and validation with survey-based data. Transportation Research Procedia 6, 313–324. 

Benenson, I., K. Martens, and B. Slave (2008). Parkagent: An agent-based model of parking in the city. Computers,Environment and Urban Systems 32(6), 431–439. 

Bonsall, P. and I. Palmer (2004). Modeling drivers’ car parking behavior using data from a travel choice simulator. Transportation Research Part C: Emerging Technologies 12(5), 321 – 347. 

Box, C. (2004). Curb-parking problems: Overview. Journal of Transportation Engineering 130(1), 1,5. 

Byrd, J. E. and V. P. Sisiopiku (2006). Comparison of level-of-service methodologies for pedestrian sidewalks. In the Transportation Research Board 85th Annual Meeting, Washington DC, United States. 

Cao, J., M. Menendez, and V. Nikias (2015, Sep). The effects of on-street parking on the service rate of nearby intersections. Journal of Advanced Transportation 50(4), 406–420. 

Desjardins, R. (1977). Effective low cost traffic engineering. In the Compendium of Technical Papers of the 47th Annual Meeting of the Institute of Transportation Engineers at the Fourth World Transportation Engineers Conference, Mexico City. 

Geroliminis, N. (2015, Sep). Cruising-for-parking in congested cities with an mfd representation. Economics of Transportation 4(3), 156–165. 

Goolsby, M. E. (1971). Influence of incidents on freeway quality of service. In the 50th Annual Meeting of the Highway Research Board, Washington District of Columbia, United States, pp. 41–46. 

Gordon, R., R. Reiss, H. Haenel, E. R. Case, R. L. French, A. Mohaddes, and R. Wolcott (2005). Traffic control systems handbook. Washington, DC: Federal Highway Administration. 

Guo, H., Z. Gao, X. Yang, X. Zhao, and W. Wang (2012). Modeling travel time under the influence of on-street parking. Journal of Transportation Engineering 138(2), 229–235. 

Guo, H., W. Wang, and W. Guo (2012, Sep.). Micro-simulation study on the effect of on-street parking on vehicular flow. In 2012 15th International IEEE Conference on Intelligent Transportation Systems, pp. 1840–1845. 

HRB (1971). Special report: Parking principles. Technical report, Washington, D.C. 

Humphreys, J. Bishop, and S. T. D. Wheeler, D. J (1978). Safety aspects of curb parking. Federal Highway Administration. 

KPMG (2017). Solving congestion by rewarding people for changing their driving habits. Wolfson Economics Prize

Marshall, W. E., N. W. Garrick, and G. Hansen (2008). Reassessing on-street parking. Transportation Research Record 2046(1), 45–52. 

ArcGIS (2019). Geospatial data formats and file types. https://doc.arcgis.com/en/ arcgis-online/reference/geojson.htm. 

City of Sydney (2017). 2017 floor space and employment survey block data. https: //www.cityofsydney.nsw.gov.au/learn/research-and-statistics/ surveying-our-community/floor-space-and-employment-survey/ local-area-overview-summary. 

City of Sydney Data hub (2019). Parking meter. https://data.cityofsydney.nsw.gov.au/ datasets/parking-meters. 

New South Wales Roads and Maritime Services (2015). Clearways proposal – mona vale road from pacific highway, pymble to kitchener street, st ives : Sydney clearways strategy, community consultation report. Roads and Maritime Services

Transport for NSW (2019). Sydney clearways strategy : keeping sydney moving. [Chippendale, New South Wales] Transport for NSW

Ommeren, J. N. V., D. Wentink, and P. Rietveld (2012, Jan). Empirical evidence on cruising for parking. Transportation Research Part A 46(1), 123–130. 

Portilla, A. I., B. A. Oreña, J. L. Berodia, and F. J. Díaz (2009). Using m/ m/∞ queueing model in on-street parking maneuvers. Journal of Transportation Engineering 135(8), 527–535. 

Segal, M. D. (1972). Highway safety research study. Maine State High-way Commission. Shoup, D. (2005). The High Cost of Free Parking. American Planning Association, Routledge. Shoup, D. C. (2006, Nov). Cruising for parking. Transport Policy 13(6), 479–486. 

Smith, B. L., L. Qin, and R. Venkatanarayana (2003). Characterization of freeway capacity reduction resulting from traffic accidents. Journal of Transportation Engineering 129(4), 362–368. 

Teodorovic, D. and M. Janic (2017). Freight transportation and logistics. Transportation Engineering, 569 – 634. 

Thompson, R. G. and R. A. J. (1998). A parking search model. Transportation Research Part A: Policy and Practice 32(3), 159 – 170. 

Vickrey, W. (1994). The economizing of curb parking space. Journal of Urban Economics 36, 42–65. Weant, R. and H. Levinson (1990). Parking. Technical report, CT United States. 

Wijayaratna, S. (2015, Sep). Impacts of on-street parking on road capacity. Australasian Transport Research Forum, 1–15. 

Wilbur, S. and Associates (1965). Parking in the city center. New Haven, Conn. 

Young, W., D. Beaton, and S. Satgunarajah (2010). An analysis of the spatial distribution of parking supply policy and demand. Australasian Transport Research Forum

Yousif, S. and Purnawan (1999, 01). On-street parking: Effects on traffic congestion. Traffic Engineering Control 40

Zhang, K. and A. Excell (2011, 09). A micro-simulation approach to quantifying clearway benefits. Australasian Transport Research Forum 2011 34(0014).