kanafani a - the consistency of traffic forecasts for airport master planning
DESCRIPTION
Airport Planning, Terminal Planning, Forecasting, Assumption Rectangle, Demand, Airport Investment, Kanafani ModelTRANSCRIPT
Institute of Transportation Studies
Working Paper UCB-ITS-WP-81-2
THE CONSISTENCY OF TRAFFIC FORECASTS
FOR AIRPORT MASTER PLANNING
Adib Kanafani
Paper prepared for presentation to the Third World Airports Conference, Singapore, March 1981
University of California
Berkeley, February 1981
ISSN 0192 4141
2
THE CONSISTENCY OF TRAFFIC FORECASTS FOR AIRPORT MASTER PLANNING
Adib Kanafani
University of California, Berkeley
Introduction
The importance of traffic forecasting in the preparation of airport
master plans cannot be overemphasized. Traffic forecasts provide an in-
dispensable input for the preparation of facility designs, economic evalu-
ations, and investment programs for airport development. If it is recog-
nized that the purpose of airport master planning is to provide for an
orderly and rational development of an airport through to its ultimate
potential, then it is possible to argue that the purpose of traffic fore-
casting in this process is not to predict future volume, per se, but to
provide for a consistent set of assumptions regarding the parameters of
design and evaluation. The important inputs into the master planning pro-
cess are not necessarily the predictive statements of what the traffic
levels might, or will, be at some future date, but the planning statements
of what traffic levels the airport is to be planned for. The difference be-
tween these two is vital and often overlooked in airport master planning.
Aggregate traffic forecasts are almost certain to be inaccurate, particu-
larly as the forecast horizon becomes longer. The airport master plan
should be made robust to this possible discrepancy. It should be based on a
consistent set of assumptions regarding the constitution and the charac-
teristics of the traffic volumes for which the airport is being planned. It
is the characteristics of traffic, such as aircraft mix and peaking pat-
terns, that determine the ultimate potential capacity of a given airport
site. When this ultimate level will in fact materialize is a prediction
that cannot be made with certainty. Nor need it be. The development of an
3
airport system can be made to adapt to actual traffic growth and evolution
with only short-term forecasting necessary, if sufficient flexibility is
built into the master plan.
Current State-of-the-Art
In current practice, it is customary for the airport sponsor to provide
to the master planner a set of forecasts of total traffic. These forecasts
are usually provided for passengers and for total aircraft operations, as
these are the sorts of forecast variables that are amenable to econometric
forecasting. The record on this type of forecasting is clear: most long-
term forecasts of passenger traffic and of aircraft operations have tended to
overestimate actual developments. Figures 1 and 2 show the results of
forecasts made in and around 1969 for 1980 traffic levels compared with the
actual volumes of 1978 and 1979. Figure 1 shows the comparisons for annual
passenger forecasts, and Figure 2 for annual aircraft operations. These
comparisons cover a range of airports including high volume ones such as
London Heathrow and medium volume ones such as Tampa, Florida. The trend is
self-explanatory: the forecasts in all cases have overestimated traffic.
Particularly, in the case of operations forecasts, there seems to be a tend-
ency for the overestimation of volume to increase with the level of traffic.
It should be noted that these figures are approximate and indicate that bar-
ring any dramatic traffic growth between 1979 and 1980, an overestimation of
traffic by the 1969 forecasts is demonstrable.
The impacts of such an overestimation should not be as critical as one
might believe at first. It usually results in a master plan in which the
development of the airport system is seen to occur at a faster pace than what
in fact occurs. Larger facilities may be justified on the basis of the high
forecasts, but then their construction is delayed because traffic does not
materialize as anticipated earlier. At worst, some facilities may be built
4
earlier than the optimal year, with the result that resources are misallocated
for a period of time. Eventually, traffic grows and the available facilities
will be put to use.
The need for accuracy is much more critical when dealing with some of the
design and planning parameters that are either derived from the forecasts, or
based directly on independent assumptions. Errors in forecasting some of
these parameters may result in irreparable damage to the airport system plan.
As an example of this, consider the number of passengers per operation. The
forecasts for this parameter made in 1969 for 1980 are shown in Figure 3
compared with actual 1978/79 figures. Here the trend is also clear. There is
not consistent over- or underestimation, but in the majority of cases, the
forecast and the actual numbers are far apart. The type of planning error
that can result from this type of forecast discrepancy can be detrimental to
the efficient functioning of the resulting airport system. If the number of
passengers per operation is overestimated, then the terminal facilities and
the access facilities will be overdesigned in relation to the airfield and the
apron-gate subsystems. Additional traffic growth, whether rapid or slow,
will not correct this inconsistency, resulting in landside facilities that are
underutilized. If, on the other hand, the number of passengers per operation
is underestimated, then the opposite will occur, and landside facilities will
be overcrowded in relation to the airfield and apron/gate facilities. The
crowding of the passenger terminal facilities at Los Angeles (LAX), for
example, should not be surprising given the actual and the forecast passengers
per operations (86 and 55, respectively).
Traffic Assumption Needs in Master Planning
In airport master planning, numerous assumptions need be made regarding
the constitution and characteristics of forecast traffic volumes. Total
annual passenger volumes and aircraft operations are of only limited value
5
as planning inputs by themselves, as most of the facility designs and evalua-
tions are based on considerably more detailed breakdowns of these volumes. It
is usually not possible to forecast these detailed characteristics, and hence
assumptions must be made regarding them. These assumptions should be arrived
at by a consensus of the various parties involved in airport master planning.
Many of them are based on observed historic characteristics which are modified
to account for changes in technology and in operating procedures.
The following is a listing of the traffic assumptions most often used
in planning, and the purpose for which they need to be made:
Table 1
Assumptions Needed in Airport Master Planning
Assumption
Annual Passengers
Annual Operations
Design Hourly Passengers
Design Hourly Operations
Fleet Mix
Use in Planning Economic evaluation.
Economic evaluation,and general sizing of airfield facilities.
For design of terminal facilities, and access system.
For design and layout of airfield system, and sizing of apron and air-craft gate positions.
For airfield design, apron and gate position design, and for integration of airside and landside facilities.
In addition to these basic inputs, assumptions have to be made regarding
parameters that provide for a consistent linkage between them. The following
are important characterizations of traffic about which assumptions must be
made: passengers per operation; load factor; proportion of transfer and
transit traffic; access mode choice; and incidence of passenger companions. In
addition, these assumptions have to be made separately at the annual level, and
at the design hourly level. For example, the annual number of passengers per
operation is different from that during the peak hour. The former may
6
be useful in some economic analysis, but the latter is crucial for the de-
sign of the airport facilities and for integrating the plans of the airside
and the landside.
Consistency in Forecast Assumptions
The concern of this section is not with the assumptions made as a part of
the econometric forecasting process used to arrive at forecasts of total annual
passenger and operation volumes, but at the assumptions made in order to arrive
at the more specific parameters used in planning and design. It is usually
necessary to take a set of figures that have been forecast exogenously and use
them to derive these planning parameters.
In order to illustrate the concept of consistency between forecast assump-
tions, we take the case where the following four traffic figures are given
exogenously: Annual Passengers AP, Annual Operations AO, Design Hourly Pas-
sengers HP, and Design Hourly Operations HO. These last two could signify peak
hour volumes, or some other hourly volume such as the thirtieth busiest hour or
the peak hour of the average day of the peak month. The relationship between
these four traffic figures is illustrated in Figure 4. It can be given by the
following formulae:
HP = AP. x
HO = AO. y
HP = HO. m
AP = AO. n
It is clear from the requirements of consistency that for the rectangle of
Figure 4 to close, the following relationships must hold:
(x/y) = (m/n)
Note that m and n are important planning parameters; m more so than n. The
number of passengers per operation during the design hour, m, is an important
determinant of landside facility sizing, and of the balance between landside
7
and airside capacity. Attempts are often made to forecast this indicator
exogenously by projecting and adapting historic trends. It is noted that the
assumptions regarding m are closely related to those regarding the fleet mix
and the prevailing load factor during the design hour. Let ACH be the
aircraft size during the design hour, and LFH the corresponding load factor.
Then
m = ACH.LFH
which means that assumptions regarding m should be consistent with assump-
tions regarding the load factor and the fleet mix. Such assumptions cannot
be made by the master planner without consultation with the airline, for
example, whose planning is based on similar assumptions of fleet mix and load
factors. It is also important to note that projections of m must be made
separately from n which is usually derived simply by dividing the exogenously
given forecasts of AP and AO (n = AP/AO). The reasons for this is that m is
usually larger than n, indicating that the number of passengers per operation
during the design hour is larger than that averaged over the whole year. The
same applies to the projection of the design hour factors x and y. While both
of these factors will vary over time, basically declining as the traffic
volume increases at the airport, it is common to expect that x will always
exceed y. While airline schedules may offer more flights during periods of
high passenger flows, it is usual for load factors to increase during those
periods and hence for the operation's design hour factors to be lower than
those for passengers.
When constructing such an "Assumptions Rectangle" for an international
airport, it is important to distinguish between domestic and international
traffic, and between origin-destination and transit traffic. Even if sim-
ilar peaking characteristics can be assumed similar for both, it is likely
that seating factors m and n are different due to the different fleet mixes
8
used in domestic and international operations. This is illustrated by showing
an example "Assumptions Rectangle" construction done as a part of the
elaboration of the master plan for the Viracopos International Airport serving
the Sao Paulo region in Brazil. The rectangle shown in the following
illustration is for the year 2000 and is based on the following exogenous
forecasts:
AP = 6.488 million 0-D plus 1.5 million international transit
AO = 119,035
HP = 152 domestic plus 2130 international 0-D and 526 international transit
ACH = 133 seats domestic 248 seats international
LFH = 13% domestic
60% international
In order to complete the assumptions rectangle, it is necessary to com-
pute the hourly operations. This is done on the basis of the information
given regarding average aircraft size ACH, the load factors LFH, and the
hourly passenger flows HP.
Domestic Operations = (Domestic HP). (Domestic ACH.LFH) = (152)/(133 x 0.13) = 9 operations
International Operations = (2130 + 526)/(248 x 0.60) = 18 operations
Total Hourly Operations HP = 9 + 18 = 27.
The given information can be used to construct the rectangle and to
calculate the corresponding values of m and n for total traffic, and m' and n'
for traffic excluding the transit flows.
A check of the consistency can be simply made by comparing (x/y) with
(m/n) or (m'/n') in this case all having a value of 1.59. It is noted that
the separation of transit flows in this case is essential since these flows
9
would count in the calculation of aircraft operations and hence in runway
system planning, but not in landside flows and facility planning. In other
words, while HO = 27 is a planning figure for the airfield system, HP' = 2270
would be the corresponding passenger flow figure for terminal and landside
planning.
Another aspect of consistency in assumptions is the need to adjust some
of the parameters. For example, the rectangle in Figure 5 can be constructed
for different forecast years, or levels of traffic volume. For each, dif-
ferent assumptions regarding m, n, x, and y need be made. Fleet mixes usually
change over time resulting in different aircraft sizes, and so do the peaking
characteristics resulting in different design hour factors x and y. It is
customary to observe that as the overall traffic volume increases at an
airport, the peak hour factor declines. This is expected since for an airport
with a single operation, the peak hour factor will have a maximum value of
unity, and for an airport with constant traffic flow throughout the whole
year, the peak hour factor will have a minimum value of (1/365 x 24 =
0.00011). The phenomenon is illustrated in Figure 6 where typical trends in
design hourly traffic DHT and design hour factor DHF are shown. Such trends
would apply similarly whether traffic is measured by passengers or by opera-
tions. The adjustments of the design hour factor are essential for achieving
consistency in the master plan, and for avoiding imbalance between the various
components of the airport system. We illustrate this with an example from the
Viracopos International Airport master plan. In that master plan, passenger
terminal units are designed for a load of 2000 hourly passengers each. In
preparing a staging plan for the airport toward the ultimate capacity of the
site, it was necessary to seek a balance between the ultimate capacity of the
airfield area, with three parallel runways, and the terminal
10
building area with space available for eight terminal units. To do this, it
was necessary to project the annual capacity of the terminal system with dif-
ferent numbers of terminal units, and then to compare that with the corre-
sponding operations capacity of the runway system as it evolves from the
single runway currently in use to the two and consequently three runway con-
figuration. Figure 7 illustrates how the capacity projection of the terminal
system is done. As the number of terminal units is increased, the peak
hourly flow capacity is correspondingly increased in increments of 2000 pas-
sengers. These hourly flow capacities correspond to increasingly higher an-
nual capacities since the peak hour factor is assumed to decrease. Thus,
while two terminal units giving a capacity of 4000 peak hour passengers will
correspond to an annual capacity of ten million passengers, four terminal
units with an hourly capacity of 8000 passengers correspond to an annual
capacity of twenty-seven million passengers. The following table illustrates
this projection.
Table 2
Traffic Projections for Viracopos Airport
Year Annual Passengers AP'
Design Hour Factor X
Hourly Passengers HP'
Terminal Units
1985 1.1 M 0.04% 440 1
1990 1.4 0.04% 560 1
1995 5.0 0.04% 2000 1
2000 6.4 0.035% 2240 2
2005 8.0 0.035% 2800 2
2010 24.0 0.030% 7200 4
2015 28 0.030% 8400 5
2020 33 0.030% 10000 5
11
Note that if the system capacity were obtained by simply multiplying the
number of terminal units by five million annual passengers, then a totally
different staging plan would be obtained, and one that is likely to lead to an
oversizing of the passenger terminal system. Figure 7 illustrates this by
comparing the forecast capacity with the flows obtained by this simple
multiplication, according to which seven terminal units would be needed by the
time the annual traffic volume reached thirty-three million versus five
according to the project capacity. The implications on the airport layout
plan are significant. In this particular case, the number of runways is
limited by the physical characteristics of the site to three. The approach to
the layout plan is to allocate only as much space to the terminal system as
would be needed to provide a passenger capacity corresponding to what can be
handled by three runways. Even taking into consideration the uncertainties
regarding future aircraft size and runway capacity flow rates, it would still
make a difference whether one allocated space for a maximum of six or of eight
terminal units, since this means a total land area of 9.4 or 6.8 sq. km.,
respectively.
Summary and Conclusions
Achieving consistency in the assumptions used in airport master planning
is more important than attempting forecasting accuracy. Long range forecasts
of airport traffic are not likely to be accurate. A flexible airport master
plan can provide robustness against forecasting errors. However, errors in
some of the planning assumptions cannot be corrected against and may result in
lasting suboptimalities in the development of the airport system. The attempt
to achieve consistency in making planning assumptions is an attempt to avoid
some of these pitfalls in airport master planning.
Finally, it is recognized that airport master planning continues to be
based on assumptions rather than accurate predictions of the future. This,
12
however, should not preclude making these assumptions realistic. This can be
done by, first, seeking consensus on the assumptions among all the parties
involved in the planning process, and second, maintaining consistency among
all the planning assumptions made.
13
Fig. 1 COMPARISON OF ACTUAL AND FORECAST AIRPORT TRAFFIC
14
Fig. 2 COMPARISON OF ACTUAL AND FORECAST AIRPORT OPERATIONS
15
Fig. 3 COMPARISON OF ACTUAL AND FORECAST PASSENGERS PER OPERATION
16
Fig. 4 ASSUMPTIONS RECTANGLE OF AIRPORT TRAFFIC
17
Fig. 5 ASSUMPTIONS RECTANGLE FOR VIRACOPOS AIRPORT
YEAR = 2000
18
Fig. 6 TRENDS IN DESIGN HOUR FACTOR
19
Fig. 7 FORECASTING TERMINAL SYSTEM CAPACITY