Analysis of Passenger-Ferry Routes Using Connectivity Measures

Analysis of Passenger-Ferry Routes

Using Connectivity Measures
Avishai (Avi) Ceder and Jenson Varghese
University of Auckland

Abstract

This study examines ferry routes that arrive at a Central Business District (CBD)
during peak periods. Ferries are investigated because in certain locations they pro-
vide an alternative to buses and private vehicles, with potentially faster and more
reliable journey times. The objectives of the study were to (1) conduct a connectivity
analysis of existing commuter ferry services and (2) investigate potential demand
for ferry services and develop potential new routes. The case study is of Auckland,
New Zealand. The first stage of the study analyzed the connectivity of existing ferries
routes to the CBD with bus services within the CBD utilizing measures of connectiv-
ity with attributes of walking, waiting, and travel times, and scheduled headways.
The second stage involved developing new commuter routes from within the greater
Auckland region to the CBD. The origins of these new routes were developed based
on the potential demand of area units derived from journey-to-work data from the
2006 New Zealand Census. These new routes were then compared with existing bus
routes from similar locations to the CBD to provide an additional assessment of the
feasibility of the new routes. Finally, recommendations are made on the establish-
ment of the new ferry routes.

Introduction
Ferries are an alternative to land-based modes of transportation such as buses and
private vehicles, with potentially faster and more reliable journey times, as they do

29
Journal of Public Transportation, Vol. 14, No. 1, 2011

not compete for road space on congested road networks. An effective ferry service
has the potential to reduce traffic congestion on roads by taking people off the
road through modal shift to ferry services.

Objectives
This research study builds on previous research carried out by Ceder et al. (2009).
The objectives of this study are to:
• Conduct a connectivity analysis of existing commuter ferry services.
• Investigate demand for existing ferry services.
• Develop new potential routes based on geographic feasibility and potential
demand.
• Apply the methodology developed to the Auckland area and Central Busi-
ness District (CBD) during the morning peak period.
Background
The Auckland CBD is New Zealand’s largest employment center, with over 70,000
employees and 9,000 businesses (Auckland City Council 2006). In the Auckland
region, the Auckland Regional Transport Authority (ARTA) is responsible for the
coordination, planning, and funding of passenger transport. ARTA’s Passenger
Transport Network Plan 2006-2016 (2006) forecasts that ferry passenger trips will
increase from 4.3 million to 6.6 million per year by 2016.
The ARTA Draft Ferry Development Plan (2008) identifies key issues that will need
to be addressed to ensure that the forecasted increase in patronage is served. The
issues identified include the need for improved coordination of ferry route mainte-
nance and upgrades to the Downtown Ferry Terminal, which is the destination of
all commuter passenger services to the CBD.
The ARTA Sustainable Transport Plan (2007) sets out a 10-year program of scoped
and costed projects and practical actions to help people make safer and more
sustainable travel choices in Auckland. The plan aims to integrate sustainable
transport activities with each other and with planned improvement to infrastruc-
ture and services. The sustainable transport plan identifies a significant increase in
demand management activities, from the current level of around $10 million per
year to an average of $42 million per year for the next 10 years. This investment
is expected to divert 20,000 car trips each morning peak to walking, cycling, and
passenger transport.

30
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

The completion of the Central Connector busway in 2010 will result in more effi-
cient travel between the CBD and Newmarket via key locations such as Auckland
Hospital and The University of Auckland. (Auckland City Council 2009) As part
of the project, there are several planned improvements to bus services, includ-
ing increased frequencies, dedicated bus lanes, new bus stops and shelters, and
improved traffic signals.

Literature Review
This section contains a review of papers that propose methods for optimizing
the configuration of public-transport (PT) routes systems. A passenger ferry rout-
ing system has basically same characteristics as any other PT system in terms of
objectives, constraints, and integration consideration. Baaj and Mahmassani (1991,
1992, 1995) developed PT network design methods based on artificial intelligence
(AI). The discussed methods are based on a typical formulation of the network
design problem as a programming problem with minimal frequency and load fac-
tor and fleet size constraints. The first paper (Baaj and Mahmassani 1991) gives
a quantitative description, using flow charts, of a three-stage design process of
a route network. In the first stage, a large set of routes is generated; the second
stage involves network analysis and determination of frequencies; the third stage
is network improvement. The second paper (Baaj and Mahmassani 1992) focuses
on the method of representing the transportation network, using lists and arrays,
in order to make the solution procedure efficient. The third paper (Baaj and Mah-
massani 1995) concentrates on the stage of creating the initial set of routes, which
are supposed to be modified and improved later. To generate this initial route set,
a set of basic skeletons is created along the shortest paths between nodes with a
high passenger demand; the skeletons are expanded using a set of node insertion
manipulations.
Ramirez and Seneviratne (1996) propose two methods for route network design,
under multiple objectives, using GIS. Both methods involve ascribing an imped-
ance factor to each possible route and then choosing those that have the minimum
impedance. In the first method, the impedance factor depends on passenger flow
and the traveled road length. This method requires use of an assignment model. In
the second method, the impedance factor depends on the number of employees
who have a reasonable walking distance from the route.

31
Journal of Public Transportation, Vol. 14, No. 1, 2011

Pattanik et al. (1998) present a methodology for determining route configuration

and associated frequencies using a genetic algorithm. In genetic algorithms, solu-
tions are chosen out of a large set of possibilities in an iterative process, where the
chances of a solution to survive through the iterations are higher if it yields a high
value to a given fitness function. The method presented here adopts a typical pro-
gramming formulation of the route network design problem, with the objective
of minimizing a weighted combination of passenger time costs and operator time
costs; the objective function is the basis for the calculation of the fitness function
values. A methodology also is presented for the coding of variables as strings with
fixed or variable length.
Soehodo and Koshi (1999) formulated a programming problem for designing PT
routes and frequencies. Similar to other models, the problem is solved by first cre-
ating all feasible routes and then choosing an optimal subset. In addition to some
traditional components, such as minimal frequency and fleet size constraints, the
problem has some unique elements, such as the inclusion of private car user costs,
transit passenger crowding costs, and transfer costs to the minimized objective
function. A sub-model is developed for each of these cost types. Equilibrium of
network flows is another constraint. The model assumes that demand is elastic
and, therefore, the shift of passengers between different modes of transport has a
major role. Both PT and non-PT demand assignment models are used.
Bielli et al. (2002) describe another method for designing a bus network using a
genetic algorithm. As in other genetic algorithms, each population of solutions
goes through reproduction, crossover, and mutation manipulations whose output
is a new generation of solutions. In the proposed model, each iteration involves
demand assignment on each network of the current set of solutions, and a calcula-
tion of performance indicators based on the assignment results. These indicators
take part in a multi criteria analysis of each network, which leads to the calculation
of its fitness function value.
Wan and Lo (2002) developed a network design model with an explicit consider-
ation of intermodal and inter-route transfers. The model has two separate phases.
First, the points that are to be connected with a direct service are determined in a
heuristic algorithm. This algorithm uses a network representation approach named
State Augmented Multi-Model (SAM), which involves inserting imaginary links to
the actual road network where a direct service is provided. Afterwards, an actual
bus route system is built in a mixed integer linear programming problem.

32
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

Yan and Chen (2002) present a method for designing routes and timetables
that aims to optimize the correlation between bus service supply and passenger
demand. The method is based on the construction of two time-space networks: a
fleet flow network and a passenger flow network. Both networks are bi-dimensional
diagrams where the horizontal dimension represents bus stops and the vertical
dimension represents time. While the fleet flow network shows potential activities
of the bus fleet, the passenger flow network illustrates trip demand. The objective
of the model is to flow buses and passengers simultaneously in both networks with
a minimum cost. A mixed integer multiple commodity network flow problem and
a solution algorithm, based on Lagrangean relaxation, are presented.
Tom and Mohan (2003) continued the development of genetic methods for
route network design. In the current model, frequency is the variable, and thus
it differs from earlier models in terms of the adopted coding scheme. While fixed
string length and variable string length codings were used in previous models, the
simultaneous route and frequency coding model is proposed here. The literature
review presented in this section sheds light on what methodologies and quantita-
tive methods recently were used to overcome the planning issues of PT network
design. What follows is a different concept coordination-based with the idea to
bridge between theory and practice.
The literature reviewed provides a spectrum of modeling approaches for public-
transport network design, including the construction of ferry routes. However,
none of the approaches furnishes a sound methodology for the inclusion of con-
nectivity measures within the optimization framework. It is the purpose of this
work to shed light on both the connectivity measures and consideration of realign-
ment of ferry routes.

Public-Transport Connectivity Analysis

Connectivity analysis of PT services is made up of the following quantitative attri-
butes (Ceder at al., 2009; Ceder, 2007):
e1 = Average walking time (for a connection)
e2 = Variance of walking time
e3 = Average waiting time (for a connection)
e4 = Variance of waiting time

33
Journal of Public Transportation, Vol. 14, No. 1, 2011

e5 = Average travel time (on a given transit mode and path)

e6 = Variance of travel time
e7 = Average scheduled headway
e8 = Variance of scheduled headway
In addition to the quantitative attributes, there are also qualitative attributes,
which are not as easily measured and quantified. There are:
e9 = Smoothness (ease)-of-transfer (on a given discrete scale)
e10 = Availability of easy-to-observe and easy-to-use information channels (on
a given discrete scale)
e11 = Overall intra- and inter-agency connectivity satisfaction (on a given dis-
crete scale)
These attributes contribute to an individual’s preference for passenger transport
of alternative modes. It is noted that each individual will have different preferences
and importance (weightings) assigned to the above attributes. To determine the
relative importance and weighting attributed to each of the 11 attributes, surveys
need to be conducted on the preferences of passenger transport users. Such sur-
veys were conducted by Ceder et al. (2009).
As noted above, measuring PT connectivity involves various parameters and com-
ponents. Therefore, the following notations are introduced to ease the explicit
construction of connectivity measures.
For a given time window (e.g., peak-hour, average week-day):
O = {Oi} = set of origins Oi
D = {Du} = set of destinations Du
PDk = {P} = set of inter-route and inter-modal paths to Dk
POk = {Pi} = set of inter-route and inter-modal paths from Ok
Mp = {m} = set of transit routes and modes included in path p
Et = {et} = set of quantitative attributes suitable for connectivity measures
Eℓ = {eℓ} = set of qualitative attributes suitable for connectivity measures
e jmp = the value of attribute ej, j = t, ℓ, related to mode m on path p

34
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

αe = weight/coefficient for each attribute ej, j = t, ℓ

c jp = quantitative and qualitative (j= t, ℓ) connectivity measure of path p
Fp = average number of passengers using path p
cp(i,j) = capacity (flow of passengers) of arc (i,j) between route and mode i,
and between route and mode j; each i can also be an origin Oi or des-
tination Di; (i,j) is contained in path p and is part of a network-flow
model.
Based on the notations, the following equation-based notations are established:
, j = t,ℓ (1)

, j = t,ℓ (2)

, j = t,ℓ (3)

, j = t,ℓ (4)

, j = t,ℓ (5)

, j = t,ℓ (6)

, j = t,ℓ (7)

, j = t,ℓ (8)

, j = t,ℓ (9)

, j = t,ℓ (10)

35
Journal of Public Transportation, Vol. 14, No. 1, 2011

Equation (1) has the purpose of comparing paths (chains of trips) that each have
an origin and destination and may include transfers. This comparison is usually
carried out when there is a change in one or more paths; otherwise, it categorizes
the different paths by their access/egress connectivity quality using the evaluation
tool proposed.
Equation (2) is used to compare destinations. Equations (6) and (7) have the same
purposes as Equations (1) and (2), respectively, but include the consideration of
passenger flow by determining of the average number of passengers exposed to the
calculated level of connectivity. Equation (9) compares groups of destinations with
regard to overall existing connectivity quality.
All the connectivity measures that consider passenger flows should be updated
in the event of changes or improvements to schedules, routes, or services. When
referring to a group of destinations (zonal-based, purpose-based), paths can have a
stop at one destination and continue to others.

Methodology
Connectivity Assessment
To develop an assessment of the existing operation of commuter ferry networks to
the Auckland CBD during the AM peak period of 7 AM to 9 AM, an analysis was
carried out on the connectivity of existing ferry routes to the CBD with outward-
bound bus services from within the CBD. This was the first stage of the study and
is outlined below.
The connectivity analysis was carried out using the following assumptions:
• The analysis incorporated the quantitative measures, while qualitative
measures were excluded. These qualitative measures were ease-of-transfer,
availability of information, and overall intra- and interconnectivity.
• The travel times and service headway information were based on published
timetables from MAXX (see below).
• The weighting attributes used in the analysis were based on the results of
Ceder et al (2009).
• The connectivity measures developed did not incorporate passenger flow
and were normalized by each quantitative measure to allow for cross-
comparison.

36
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

The travel time and headway information for the specific routes was obtained from
published timetable information from MAXX, the regional transport brand for
Auckland managed by the Auckland Regional Transport Authority (ARTA).
The connectivity for each arc was calculated using the formula:

j = t,ℓ (11)

The connectivity attributes and their calculations are outlined below.

Average Walking Time (e1): The walking distances between the CBD ferry
terminals and the location of connecting bus stops were measured using map-
ping tools on Google Maps. The walking times were then calculated on an average
walking speed in Auckland of 1.3 meters per second, based on the findings of the
study by Opus International Consultants, “Factors Influencing Walking Speed.” In
the cases, where the average walking distances were unknown (such as the walking
distances to ferry terminals), an estimated average walking time of 13.24 minutes
(Walton 2008) was used.
Variance of Walking Time (e2): The variance in walking time was based
on the findings of Walton (2008) for variance in walking times in Auckland and
resulted in 6.5 minutes2 for walking times to ferry terminals.
Average Waiting Time (e3): The average waiting time (e3) was determined
from the scheduled headway and is shown in Equation (12):

j = t,ℓ (12)

where:
H is the average scheduled headway (minutes)
Var(H) is the variance of the scheduled headway (minutes2)
Variance of Waiting Time (e4): The variance of waiting time was determined
from the scheduled headway and assumes that delays (which would increase the
average waiting times) are negligible. The formula for the calculation of the vari-
ance of waiting times (e4) is:

37
Journal of Public Transportation, Vol. 14, No. 1, 2011

(13)

Average Travel Time (e5): The travel times were calculated from the differ-
ence between the scheduled departure and arrival times; the average travel time
was determined from all scheduled trips during the AM peak period. It is noted
that the actual travel times are likely to differ from the scheduled travel times.
Variance of Travel Time (e6): The variance of travel time was calculated
from variation (data-based) in scheduled travel times.
Average Scheduled Headway (e7): The average headway was determined
for each arc (data-based) of all services where the origin and destinations were
consistent with their respective arcs.
Variance of Scheduled Headway (e8): The variance of scheduled headway
was determined by the average difference between scheduled and actual (data-
based) departures during the AM peak.

Demand Assessment
This section outlines the methodology used for the second stage of this study,
which was to determine the locations within Auckland where new routes might be
feasible. The first criterion investigated was areas of potential demand based on the
premise that it would not be economically viable to develop routes in areas where
there is no demand.
Census data from Statistics New Zealand were used to determine areas where there
was a potential demand for ferry routes into the Auckland CBD. Journey-to-work
data were generated where the destination for all trips was the Auckland CBD.
The CBD area was defined by the combination of four census area units (Auckland
Harbourside, Auckland Central West, Auckland Central East and Grafton West), as
shown by the shaded regions of Figure 1.

38
Analysis of Passenger-Ferry Routes Using Connectivity Measures

Figure 1. Auckland CBD Area Units

39
Journal of Public Transportation, Vol. 14, No. 1, 2011

It is assumed that the majority of these journey-to-work trips were carried out in
the AM peak period. In the use of the census data, the following limitations were
noted.

Age of Census Data

The census was carried out on March 7, 2006, which meant that the journey-to-
work data were over three years old at the time this analysis was undertaken. An
implication of the time lag is that economic events (such as fuel price rises) and
land use and transport infrastructure changes over the past three years may have
altered the region’s travel patterns. This was noted and taken into account when
estimating the potential demand. Investigations confirmed that all existing ferry
routes were operational by 2006, and the impacts of improvements to bus and
ferry infrastructure were assumed to be negligible for the purposes of this investi-
gation.

Modes of Travel
In the 2006 census, respondents were asked to enter their main mode of travel to
work on March 7, 2006. The modes of travel indicated were:
• Drove a Private Car, Truck, or Van
• Drove a Company Car, Truck, or Van
• Passenger in a Car, Truck, Van, or Company Bus
• Public Bus
• Train
• Motorcycle or Power Cycle
• Bicycle
• Walked or Jogged
• Other
• Worked at Home
• Did Not Go To Work Today
• Not Elsewhere Included
A limitation of the census data is that journey-to-work by ferry was not included as
an option for respondents to select. Discussions with Statistics New Zealand1 indi-
cated that travel by ferry was expected to be covered under the category “Other,”
which also might have included travel by taxi or airplane. Statistics New Zealand

40
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

also acknowledged the possibility that certain respondents who traveled by ferry
may have selected “Not Elsewhere Included” category instead of selecting “Other.”
This limitation does not affect the assessment of future ferry demand, although
in determining existing demand, the assumption was made that the value in the
“Other” field would reflect travel by ferry for areas close to existing ferry departure
points, given that all existing ferry terminals are a significant enough distance away
from the CBD to make travel by taxi unlikely from a cost perspective (and travel by
airplane to the CBD is not possible).
Another constraint with the census data was that any field that had a value of less
than six (for example, five people traveling to the CBD by train from Wellington)
was suppressed to ensure confidentiality of individuals. These suppressed values
were treated as zero values, and it has been identified that this would result in
lower demand estimates for all modes; however, given that this analysis was looking
at area units (catchment areas) of high demand, it was considered acceptable that
areas with potential demand less than six were ignored.
Discussions were held with Fullers Ferries Ltd.,2 the main operator of ferries in the
Auckland region, to determine the patronage on the existing services. Patronage
information on individual routes was not available due to commercial sensitivity,
although archived ARTA business reports indicate a total monthly ferry patronage
of approximately 440,000 in March 2006 and 445,000 in March 2009. This captured
all weekday and weekend trips, and it was not possible to accurately break down
this information to obtain estimates of patronage or demand on existing ferry
routes. The CBD Access Strategy 2006 identified 2,700 passengers arriving in the
CBD during the AM peak (Auckland City Council 2006).
An estimation of existing demand from the census data was made based on the
assumption that all trips from area units near existing ferry infrastructure under
the category “Other” were ferry trips.
The estimation of the potential demand for new ferry routes was developed using
census data on bus patronage. This was based on the premise that bus users would
be more likely to shift modes to ferry transport if the new services provided more
attractive travel times and connectivity. It is also likely that the introduction of new
ferry services might result in commuters currently traveling by private vehicle to
switch to ferries; however, to ensure a conservative estimation of demand only bus
patronage was investigated.

41
Journal of Public Transportation, Vol. 14, No. 1, 2011

Area units with more than 10 commuters traveling by bus were identified. These area
units were then filtered out based on their proximity to the ocean. As the location
of any potential ferry terminal or pier would be by the sea, all area units that were
more than 1,000m away from the ocean were eliminated, based on the assumption
that passenger transport users were willing to walk an average of 1,000 meters to a
transport hub, based on the findings of Walton (2008). This eliminated any potential
demand areas where passengers might drive to ferry terminals to ensure conserva-
tive demand estimation, as this is not consistent with regional policy. Once these
constraints were incorporated in the evaluation, the remaining area units with high
bus patronage were used to develop the potential ferry routes.

Route Feasibility Assessment

Once the areas of high demand were identified, Fullers Ferries was consulted on
the feasibility of new routes to service these areas. Catchment areas that required
routes traveling under bridges with unacceptably low clearances or tidal con-
straints were eliminated. It also was noted that some of the potential piers and
wharfs investigated might require significant infrastructural investment before
they would be feasible ferry terminals. As this study focuses on demand, the
potential cost implications of the engineering requirements were not taken into
consideration.
Once the new potential catchment areas were identified, options for new routes
were developed based on estimated journey times. The journey times of these
new routes were then compared with existing bus routes to provide an additional
assessment of the feasibility of the new routes. The use of intermediate nodes on
ferry routes also was investigated, and this was determined from estimated journey
times.

Auckland Case Study

The Auckland region has the largest population in New Zealand at approximately
1.3 million. It is also one of the fastest growing regions in New Zealand, with a
forecasted increase of 440,000 people over the next 15 years (ARTA 2008). This
increase in population will put additional pressure on an already-congested road
transport network.
This study investigated ferry routes to the Auckland CBD and their connectivity
with bus services to develop an increased understanding of the effectiveness of the
42
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

Auckland CBD as a passenger transport hub. This study also looked at new ferry
routes to potentially improve the efficiency of road-based passenger transport
through the creation of a mode shift from private vehicles to ferries. The initial
analysis conducted benefited from the study of ferry-route design in Hong Kong
by Ceder and Sarvi (2007).

Ferry Operations in the Auckland Region

There are six ferry operators in the Auckland region, with Fullers Ferries being the
largest operator of commuter ferry trips to the Auckland CBD (Downtown). Table
1 presents a summary of ferry services within the Auckland region.

Table 1. Ferry Services in the Auckland Region

Weekday
frequency
(trips Bus
Service Terminals Type Operator each way) Feeder
Devonport Downtown-Devonport Passenger Fullers 31 All trips
Stanley Bay Downtown-Stanley Bay Passenger Fullers 8
Bayswater Downtown-Bayswater Passenger Fullers 21 All trips
Birkenhead Downtown- Passenger Fullers 24 Most
Birkenhead-Northcote trips
West Harbour Downtown-Westpark Passenger Belaire 8
Marina Ferries
Gulf Harbour Downtown-Gulf Passenger 360 2-3 1 trip
Harbour Discovery
Half Moon Bay Downtown-Half Moon Passenger Fullers 11 Most
Bay trips
Waiheke Passenger Downtown-Matiatia Passenger Fullers 20 All trips
Waiheke Vehicular Half Moon Bay- Vehicular Sealink 12
Kennedy Pt
Waiheke Vehicular Half Moon Bay- Vehicular Waiheke 4-7
Kennedy Pt Shipping
Pine Harbour Downtown-Pine Passenger Pine Harbour 11
Harbour Ferries
Great Barrier Island Wynyard-Tryphena Vehicular Sealink 1
Rangitoto Downtown-Rangitoto Passenger Fullers 4
Rakino Pine Harbour-Rakino Passenger Pine Harbour 0 (week-
Ferries end only)
Source: ARTA Ferry Development Plan

43
Journal of Public Transportation, Vol. 14, No. 1, 2011

ARTA also identifies plans for new routes from Beach Haven and Hobsonville, with
an increase in trips for the Half Moon Bay, Bayswater, Gulf Harbour, Birkenhead/
Northcote Pt, Stanley Bay and Waiheke Island routes.

Data Analysis
Existing Ferry Routes. The existing ferry routes that travel to the Auckland CBD
during the AM peak are presented in Figure 2.

Figure 2. Map of Existing Ferry Routes in the Auckland Region

(Wikipedia, 2009)
One observed characteristic of the ferry routes during the AM peak period to the
Auckland CBD is that they all travel directly to the CBD without any interim stops.
The only exception to this is the route from Birkenhead to the CBD, which has a
stop at Northcote and is executed only “on request” and does not normally occur.
Consultations with Fullers Ferries confirmed that the routes were selected in this
way largely to minimize travel time, as the docking procedures can increase total
journey time considerably.

44
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

From the analysis carried out, the origins of ferry trips to the CBD during the
defined AM peak period were:
• West Harbour
• Birkenhead
• Bayswater
• Stanley Bay
• Devonport
• Half Moon Bay
• Matiatia (Waiheke Island)
• Gulf Harbour
• Northcote
The Pine Harbour, Rangitoto, Kennedy Point, and Great Barrier Island routes were
not included, as they did include any trips during the AM peak period.
Connecting Bus Routes. The paths that were analyzed in terms of their con-
nectivity were the ferry arcs defined above added to the arcs of bus services within
the CBD.
Analysis of Census Data. The demand assessment used census data to
identify regions of high numbers of residents traveling by bus. The census mesh
blocks (approximately 9,800 mesh blocks for the Auckland region) were grouped
into area units. This resulted in 253 area units with 6 or more residents traveling
by bus, which were, in turn, grouped into regions based on Work and Income New
Zealand’s regional agglomerations. Table 2 shows a summary of the analysis carried
out on the census journey-to-work data.
Feasibility of Catchment Areas. By analyzing Table 2, several regions and
their respective area units could be eliminated as unfeasible locations for future
ferry terminals. The summary of these findings is presented in Table 3, with five
regions excluded from further analysis due to their geographic location or low
number of commuter bus trips (less than 10). This reduced the number of feasible
area units to 195.

45
Journal of Public Transportation, Vol. 14, No. 1, 2011

Table 2. Summary of Census Journey-to-Work Data

Residents Proportion of Total Bus
Total Traveling Bus Patrons of Commuters in
Region Employed by Bus Total Population Auckland
Auckland City 20,493 3,390 17% 29.4%
Takapuna 12,414 2,937 24% 25.5%
Royal Oak 7,335 2,082 28% 18.1%
New Lynn 5,337 1,044 20% 9.1%
West Auckland 5,106 666 13% 5.8%
Panmure 3,621 546 15% 4.7%
Manukau/ Otara 3,954 450 11% 3.9%
Orewa 1,215 255 21% 2.2%
Mangere/ Otahuhu 1,485 153 10% 1.3%
Papakura 540 6 1% 0.1%
Manurewa 567 0 0% 0.0%
Pukekohe 249 0 0% 0.0%
Total 62,316 11,529 160%

Table 3. Summary of Preliminary Initial Feasibility Assessment

Possible Loca-
tion for Future
Region Ferry Terminals? Reason for Exclusion (If Not Possible)
Auckland City Yes
Takapuna Yes
Royal Oak Geographic location (ferry route to Auckland CBD must
No travel around North Island, a distance of 700+ km)
New Lynn Yes
West Auckland Yes
Panmure Yes
Manukau/ Geographic location (ferry route to Auckland CBD must
No
Otara travel around North Island, a distance of 700+ km)
Orewa Yes
Mangere/ Yes
Otahuhu
Papakura No Low bus patronage
Manurewa No Low bus patronage
Pukekohe No Low bus patronage

46
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

Analysis of Existing Ferry Demand. Analysis was carried out on areas near
the existing ferry terminals to assess the existing ferry catchment areas and identify
proportions of ferry users relative to bus users. This is summarized in Table 4.
Table 4. Estimated Bus and Ferry Patronage of
Existing Ferry Catchment Areas
"Public Bus" Other "Other"
Trips as % (Assumed Trips as % Not Total
Public of Total to be Ferry of Total Elsewhere Employed
Route Origin Bus Employed Trips) Employed Included (in CBD)
Waiheke 0 0% 237 67% 69 354
Island
Stanley Bay 0 0% 105 51% 39 207
Devonport 0 0% 288 52% 93 558
Bayswater 153 13% 303 26% 120 1,149
Birkenhead 456 23% 195 10% 66 1,995
Northcote 219 27% 51 6% 27 798
Half Moon 147 13% 147 13% 27 1,110
Bay
Gulf Harbour 168 9% 69 4% 42 1,806
West Harbour 54 10% 30 5% 9 549
Total 1,197 14% 1,425 17% 492 8,526

The high proportion of journey-to-work trips to the CBD from Waiheke Island
is expected because there is no alternative mode of transport to the CBD from
the island. It is noted that approximately 20 percent of commuters from Waiheke
Island selected the option “Not Elsewhere Included,” which indicates a limitation in
the use of the census data for the purposes of demand estimation.
Stanley Bay and Devonport operators also have a high proportion of CBD commut-
ers traveling by ferry, with no commuters traveling by bus from these identified
catchment areas. This attributed to the difference in journey times where Stanley
Bay and Devonport have scheduled ferry travel times to the CBD of 10 and 15 min-
utes, respectively, while a similar journey by bus from either location would take at
least 45 minutes with one transfer, according to bus schedules. West Harbour has
the lowest number of commuter trips by Ferry at 30, according to the results of the
census data analysis.

47
Journal of Public Transportation, Vol. 14, No. 1, 2011

If Waiheke Island is excluded from the above analysis (because there is no alterna-
tive to traveling by ferry), then there is an even proportion of commuters traveling
to the CBD by bus and ferry.

Analysis of Potential Future Routes

After eliminating these regions, the number of feasible area units was reduced to
195 area units. After grouping area units by their geographic locations, the following
catchment areas were developed, and the estimated modal patronage of each catch-
ment area was obtained, based of the findings of an equal mode share between buses
and ferries in areas of existing ferry patronage, as shown in Table 5.
Table 5. Estimated Ferry Patronage
Existing Situation After the Addition of
(2006 Census) Proposed Ferry Routes
Bus Estimated Ferry Estimated Bus
Future Catchment Areas Patronage Patronage Patronage
City West 540 270 270
Mission Bay 282 141 141
Orewa North 132 66 66
Panmure 150 75 75
Takapuna 462 231 231
West Auckland 27 14 13
West Auckland (Te Atatu) 111 56 55
City West 540 270 270
Mission Bay 282 141 141
Orewa North 132 66 66

It is noted that from the above analysis, the number of potential ferry users from
West Auckland is lower than what was calculated for existing ferry catchments;
however, this has been identified by ARTA for a future ferry route.
Results

Connectivity
The results of the connectivity analysis are presented in Figure 3 in normalized con-
nectivity values of paths from ferry terminals to major locations in the CBD. The
arcs that originate from Matiatia (Waiheke Island) present the highest normalized
connectivity compared with arcs from other ferry routes to the same locations.

48
Analysis of Passenger-Ferry Routes Using Connectivity Measures

Figure 3. Path Connectivity Analysis

49
Journal of Public Transportation, Vol. 14, No. 1, 2011

This is a result of the higher variance in scheduled headways than other ferry
routes.
The destination with the highest normalized connectivity value is Victoria Park, as
this has the highest variance of scheduled headway and the highest average travel
time of all analyzed bus routes within the CBD.

New Route Recommendations

From the geographic feasibility and demand estimations assessments, the follow-
ing routes were identified as possible ferry routes (for ferry trips into the Auckland
CBD), as shown in Figure 4 by the dotted line.

Figure 4. Potential Ferry Routes

50
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

Summary, Future Study and Recommendations

Connectivity Analysis
This research investigated the connectivity of ferries with bus routes in the Auck-
land CBD and can serve as a tool and analytical framework for any ferry-related
connectivity and routing study. The next step to progress the analysis of passenger
transport connectivity within the Auckland would be the inclusion of rail services
and the extension of all bus and rail services beyond the Auckland CBD region.
Auckland is currently undergoing significant passenger transport infrastructure
upgrades, particularly with bus and rail trips to and from the CBD. On completion
of these upgrades, current scheduled services will change significantly (new jour-
ney times, frequencies, and routes), and it is recommended that the connectivity
assessments then be updated to capture these changes.

Demand Estimation
The approach of using existing bus commuters provided an initial estimate of the
potential patronage if new ferry routes were to be established. However, to develop
a more detailed estimate of the potential demand for new ferry routes, several
other factors may need to be taken into consideration. These factors include ferry
fares, bus fares, journey times by car, and socio-economic statistics of the catch-
ment area, such as median incomes and private vehicle ownership, among others.
One potential area for future research is the development of a probability logit
model to incorporate the impact these factors have on potential ferry patronage
and to provide estimates on mode shifts from other modes of transport (bus, car
and train).
The census data used in this analysis were over three years old, and this research has
already identified potential flaws with the use of census data (such as there was no
field for ferry commuters), which contributes to limitations in the demand estima-
tion process. Further investigation into additional sources of patronage informa-
tion such as ferry passenger and occupancy surveys would provide independent
data that could be cross-checked against the census information.
The development of new passenger transport routes is sometimes the result of
political decisions or in preparation for expected increases in land use develop-
ment or population growth. These factors have not been considered in the demand
analysis but may need to be considered in future demand estimation work.

51
Journal of Public Transportation, Vol. 14, No. 1, 2011

Geospatial Analysis
The analysis of ferry catchment areas and ferry routes into the CBD involved a
level of manual analysis of area units, geographic locations, and distances. This has
a higher degree of error than potentially linking this directly with New Zealand
statistics GIS database or carrying out more sophisticated geospatial analysis such
as isochrones.

Financial Feasibility
Because the ferry services in Auckland are commercially operated, a detailed finan-
cial feasibility analysis should be implemented before any new routes would be
recommended. This would involve detailed demand estimation, determination of
appropriate fares, an evaluation of the expected operational costs, and determin-
ing the level of capital investment required (in terms of wharf infrastructure and
procurement of new vessels). Once decisions are made on the number and types
of vessels to be purchased for the new routes, it would be possible to develop an
expected timetable for these new routes, which could then be incorporated back
into the connectivity analysis.

Recommendations
This study conducted an analysis of existing ferry services to the Auckland CBD
and their connectivity with bus services in the CBD during the AM peak period of
7AM to 9AM. This investigation can serve as an analytical framework for any ferry-
related connectivity and routing study. The study examined demand for existing
ferry services using 2006 census journey-to-work data from Statistics New Zealand
and identified new potential routes based on areas of high potential demand and
feasible geographic locations.
Following are recommendations for further analysis:
1. The connectivity analysis be updated on completion of the Central Con-
nector busway, which will significantly change the connectivity in the
Auckland CBD.
2. A potential mode shift should be investigated and financial feasibility of
the proposed new routes determined to gain a better understanding of the
benefits of the development of new routes.

52
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

Endnotes
1
Personal communication, May 21, 2009, Statistics New Zealand Client Servicing
Team Leader.
2
Personal communication, May 25, 2009, Fullers Ferries Operations Manager.

References
Auckland Regional Transport Authority (ARTA). 2006. Auckland Passenger Trans-
port Network Plan 2006-2016, ARTA Report.
Auckland City Council. 2006. Central Access Strategy, Auckland City Council
Report.
Auckland Regional Transport Authority (ARTA). 2007. Sustainable Transport Plan
2006-2016, ARTA Report.
Auckland Regional Transport Authority (ARTA). 2008 . Draft Ferry Development
Plan, ARTA Report.
Auckland City Council. 2009. Central Connector, http://www.aucklandcity.govt.nz/
auckland/Transport/ctc/default.asp, accessed June 17, 2009.
Baaj, M. H., and H. S. Mahmassani. 1991. An AI-based approach for transit route
system planning and design. Journal of Advanced Transportation 25: 187-210.
Baaj, M. H., and H. S. Mahmassani. 1992. Artificial intelligence-based system repre-
sentation and search procedures for transit route network design. Transporta-
tion Research Record 1358: 67-70.
Baaj, M. H., and H. S . Mahmassani. 1995. Hybrid route generation heuristic algo-
rithm for the design of transit network. Transportation Research 3C: 31-50.
Bielli, M., M. Caramia, and P. Carotenuto. 2002. Genetic algorithms in bus network
optimization. Transportation Research 10C: 19-34.
Ceder, A., and M. Sarvi. 2007. Design and evaluation of passenger ferry routes. Jour-
nal of Public Transportation 10(1): 59-79.
Ceder, A. 2007. Public Transit Planning and Operation: Theory, Modeling and Prac-
tice. Elsevier, Butterworth-Heinemann, Oxford, UK.

53
Journal of Public Transportation, Vol. 14, No. 1, 2011

Ceder, A., Y. Le, Y. Net, and C. Coriat. 2009. Measuring public transport connectiv-
ity performance applied in Auckland, New Zealand. Transportation Research
Record 2111: 138-147, 2009.
Pattnaik, S. B., S. Mohan, S. and V.M. Tom. 1998. Urban bus transit route network
design using genetic algorithm. Journal of Transportation Engineering 124: 368-
375.
Ramirez, A. I., and P. N. Seneviratne. .1996. Transit route design applications using
geographic information systems. Transportation Research Record 1557: 10-14.
Soehodo, S., and M. Koshi. 1999. Design of public transit network in urban area
with elastic demand. Journal of Advanced Transportation 33: 335-369.
Tom, V. M., and S. Mohan. 2003. Transit route network design using frequency
coded genetic algorithm. Journal of Transportation Engineering 129: 186-195.
Walton, D. 2008. Impediments to walking as a mode choice. Land Transport New
Zealand Research Report 329.
Wan, Q. K., and H. K. Lo. 2002. Two-phase transit network design in an integrated
transit system. Advanced Modeling for Transit Operations and Service Planning
– Workshop Proceedings, Crusher Advanced Study Institute.
Wikipedia. 2007. Auckland Ferry Routes, http://en.wikipedia.org/wiki/
File:AucklandFerryRoutes.png, accessed April 10, 2009.
Yan, S., and H. L. Chen. 2002 . A scheduling model and a solution algorithm for
inter-city bus carriers. Transportation Research 36A: 805-825.

About the Author

Avishai (Avi) Ceder ([email protected]) is Professor and Chair in Trans-
portation in the Department of Civil and Environmental Engineering at the Uni-
versity of Auckland and Director of the newly-established Transportation Research
Centre (TRC). He was Head of the Transportation Engineering and Geo-Information
Department at the Technion and a Visiting Professor at the Massachusetts Institute
of Technology, the University of California at Berkeley, and others. He was the Chief
Scientist at the Israel Ministry of Transport from 1994 to 1997 and the Israel del-
egate to the Transport Program of the European Community. In 2007, he published
the book Public Transit Planning and Operation: Theory, Modelling and Practice

54
 Analysis of Passenger-Ferry Routes Using Connectivity Measures

(Elsevier, Oxford, UK), which was translated into Chinese by Tsinghua Publishing
Press (Beijing) in June 2010.
Jenson Varghese is a transportation engineer with a broad range of international
and local experience in transport planning, project management, traffic modeling,
procurement, and financial and economic modeling. He currently works for McCor-
mick Rankin Cagney in Auckland, where he provides advice to public and private
sector clients in areas such as transportation policy, sustainability, and economic
development. He has Bachelor of Engineering (Engineering Science) and Master of
Engineering Studies (Transportation) degrees from the University of Auckland and
a Diploma in Economics from the University of the South Pacific.

55