Simulating the Repatriation of Canadian Forces Materiel from
Afghanistan
Bohdan Kaluzny and Raman Pall
Centre for Operational Research and Analysis, Defence R&D Canada, Department of National Defence,
Ottawa ON K1A 0K2, Canada
Keywords:
Airlift, Discrete Event Simulation, Lines of Communication, Military Logistics, Modelling, Sealift.
Abstract:
The Canadian Forces ceased combat operations in the Kandahar province of Afghanistan in 2011 and were
instructed by the Government of Canada to complete its redeployment out of Kandahar by the end ofDecember
2011. Materiel and equipment were transported back to Canada over several lines of communications. Nearly
1500 sea containers full of materiel, 800 vehicles, and 200 air pallets of material were returned to Canada
by combinations of air, sea, and ground transport. This paper describes a discrete-event simulation model
developed to analyze the repatriation of Canadian equipment from Afghanistan to Canada via the applicable
lines of communication. The objective was to develop a model that could be used to analyze the repatriation
in order to enable and improve future mission planning. The discrete-event simulation model is shown to be
representative of the actual repatriation effort and is subsequently used to determine the impacts of different
potential courses of action, measured mainly through results on the total cost and duration of the returns.
1 INTRODUCTION
In March 2008, the Government of Canada directed
the Canadian Forces (CF) to cease combat opera-
tions in the Kandahar province of Afghanistan by July
2011, ending Operation ATHENA, Canada’s 10 year
contribution to NATO’s International Security Assis-
tance Force. The CF was instructed to complete its re-
deployment out of Kandahar by the end of December
2011 and a Mission Transition Task Force (MTTF)
was set up to conduct the closure. MTTF planning
commenced in early 2010. Materiel and equipment
were transported back to Canada over several lines of
communications (LOCs). Time, the sensitivity of ma-
terial being shipped, threat and costs were the main
factors used to determine the mode of transportation
and the LOC that would be used for shipments from
Afghanistan back to Canada. Sensitive materiel con-
sisted of most vehicles, communications equipment,
weapons systems, spare parts, munitions and high
value items. Roughly 1500 sea containers, 800 vehi-
cles, and 200 air pallets of material were transported
by LOCs destined for Canada.
The main lines of communication were:
1. Air lines of communication (ALOC) from Kan-
dahar direct to Canada or to Intermediate Staging
Terminals (ISTs), referred to hereafter as ALOC
Direct and ALOC IST, respectively. The IST
was initially located in Cyprus and transitioned to
Kuwait in September 2011. These ALOCs were
used to retrograde material and vehicles that was
deemed time or security sensitive
1
. The 1 Cana-
dian Air Division devoted daily CC-177 Globe-
master III airlift assets to support mission closure.
Contracted Antonov AN-124 airlift was also em-
ployed to move large and heavy vehicles (such as
main battle tanks). Restrictions were imposed on
use of the AN-124 for the move of weapons sys-
tems, ammunition and classified materiel due to
security considerations. Escorts were utilized on
a number of flights to mitigate risk.
2. Ground lines of communication (GLOC) from
Kandahar to Karachi, Pakistan. Contracted trucks
were used to transport containers from Kanda-
har Airfield to the port of Karachi for onward
movement by scheduled maritime liner service to
Canada. Only non-sensitive materiel was trans-
ported through GLOC. During mission planning,
this route represented the least costly LOC, but
was potentially risky due to the possibility of
1
Abiding by the request of the Cypriot Government,
there would be no movement of Main Battles Tanks or am-
munition and explosives through the island. The Kuwait
IST provided no such restrictions.
264
Kaluzny B. and Pall R..
Simulating the Repatriation of Canadian Forces Materiel from Afghanistan.
DOI: 10.5220/0004206900660075
In Proceedings of the 2nd International Conference on Operations Research and Enterprise Systems (ICORES-2013), pages 66-75
ISBN: 978-989-8565-40-2
Copyright
c
2013 SCITEPRESS (Science and Technology Publications, Lda.)
criminal activity along the route.
3. Sea lines of communication (SLOC) from
Karachi or the IST back to Canada. Regularly
scheduled liner service was used from Karachi
to the Port of Montreal. Sea travel from Cyprus
and Kuwait was done by four specially contracted
ships. The International Port of Kuwait was used
for the loading of materiel and vehicles while the
Kuwait Naval Base provided a safe and secure
area for the loading of ammunition.
The MTTF recorded and managed the retrograde
operations with a spreadsheet-based tool that pro-
vided a common operating picture (COP). This was
used as the primary planning tool and was updated
daily—it provided a comprehensivelisting of all vehi-
cles, sea containers, ammunition, and general freight
identified for retrograde, prioritized for movement
along the applicable line of communication. The
MTTF COP, while providing a wealth of detailed data
such as vehicle/container weights, aircraft departure
times and loads, etc., was limited with respect to as-
sisting in options or predictive analyses.
1.1 Objective and Scope
This paper describes a discrete-event simulation
model developed to analyze the repatriation of Cana-
dian equipment from Afghanistan to Canada via the
LOCs. The objective is to enable and improve fu-
ture mission planning and to capture pertinent data
recorded in the MTTF COP. Operation ATHENA
MTTF was a large undertaking encapsulating sev-
eral efforts. More than 2700 shipping containers and
1000 vehicles were processed, more than 250 struc-
tures/buildings were transfered, over 10000 contracts
were reviewed/closed, over 7000 cubic meters of soil
was remediated, 150 terabytes of electronic data was
processed, and 120 thousand pounds of paper was
repatriated.
The scope of discussion in this paper is focused
on the equipment repatriated and the LOCs used for
the repatriation. Naturally, several simplifying as-
sumptions were made in the attempt to model real-
ity. The paper is structured as follows: Section 2 pro-
vides references to tools and models published in the
open literature and describes the discrete-event sim-
ulation model of the repatriation of Canadian equip-
ment along the different LOCs. Section 3 compares
the results of the baseline simulation model to real-
ity (as captured in the MTTF COP), and subsequently
Section 4 reports on results obtained from select sce-
narios deviating from the baseline. Actual cost figures
are purposely omitted. Instead, results are reported as
relative differences.
2 METHODOLOGY
Research into modelling military logistics has re-
ceived much attention, benefiting from both simula-
tion and optimization approaches.
Schank, Mattock, Sumner, Greenberg and
Rothenberg (1991), and more recently Powell,
Whisman and Wu (2009) provide a review of military
logistic modelling efforts. They group the modelling
approaches into deterministic linear programming,
simulation, and stochastic programming categories
(and furthermore propose a method to combine
simulation and optimization). Primary focus has been
on airlift modelling. Dantzig and Ferguson (1995)
developed one of the earliest mathematical models
for optimizing air-based transportation. Research at
the United States Naval Postgraduate School (NPS)
and the RAND Corporation were combined to create
NRMO (NPS/RAND Mobility Optimizer) described
by Baker, Morton, Rosenthal, and Williams (2002).
The heart of NRMO is a linear programming model
that minimizes the amount of cargo delivered late or
not at all. Burke, Love and Macal (2004) developed
the Transportation System Capability discrete-event
simulation model to simulate the deployment of
forces from U.S. Army bases. The U.S Air Mo-
bility Command employs a rules-based simulation
model called the Air Mobility Operations Simulator
(AMOS) for strategic and theater operations to
deploy military and commercial airlift assets.
Open literature publications from Defence Re-
search and Development Canada include an aircraft
load allocation optimization model (Ghanmi et al.,
2009) which uses a hybrid of simulated annealing and
genetic algorithm methods to solve a multi-objective
optimization problem associated with allocating a set
of cargo items across a heterogeneous fleet of avail-
able airlift assets. The model was used to con-
duct analysis of some of the strategic lift options
for the Canadian Forces and to develop a simula-
tion framework to study the effectiveness of a va-
riety of pre-positioning options (Ghanmi and Shaw,
2008). Campbell and Moorhead (2010) developed a
spreadsheet-based movement options analysis simu-
lation tool useful for military move planners to deter-
mine rough time and cost estimates of strategic move-
ment of materiel.
2.1 Discrete-event Simulation Model
A discrete-event simulation model was built in or-
der to recreate the repatriation of materiel from
Afghanistan to Canada via the LOCs. A discrete-
event simulation was used over other techniques due
SimulatingtheRepatriationofCanadianForcesMaterielfromAfghanistan
265
to the inherent flexibility of simulation models, their
ease in performing sensitivity analyses, and their abil-
ity to identify critical constraints in the system.
The model was built in the Arena simulation en-
vironment (Kelton et al., 2010). An Arena model
is a computer program containing components called
modules that represent processes or logic. Connector
lines are used to join these modules together and spec-
ify the flow of entities. While modules have specific
actions relative to entities, flow, and timing, the pre-
cise representation of each module and of each entity
relative to real-life objects is subject to the modeller.
In this case, entities within the model represent the
various items that are to be returned to Canada from
Afghanistan. The lowest level of granularity of the
items in the simulation are containers, vehicles, and
air pallets – the specific contents in the containers or
pallets are not specified. At the start of the simula-
tion, all items (vehicles, containers, and so on) are
created and placed at the Kandahar airfield. Upon cre-
ation, each of the items is assigned a specific method
of shipment (or LOC) to use in its return to Canada,
and are placed in LOC-specific queues.
Resources in the model representing the various
modes of shipment govern the availability of the AN-
124 and CC-177 aircraft used for the ALOCs, the con-
tracted trucks used for the GLOC, and the ships used
for the SLOC. Note that the data concerning these
resources is in line with the historical availability of
these resources.
Average temperatures in Afghanistan were be-
tween 20
o
C and 35
o
C at night in June to September
and 45
o
C or higher during the day. This affected the
maximum payload (MPL) that could be loaded on air-
craft, despite the fact that movement was preferred at
night when temperatures were at their lowest.
Each day, available resources are loaded with
items until they reach (but not exceed) their MPLs.
The MPL of the AN-124s was computed as a func-
tion of the external air temperature, which was deter-
mined stochastically as a function of the simulation
date through the use of historical payload data. How-
ever, data concerning the relationship between the
payloads of the CC-177 and the external air temper-
ature was not available – hence, the MPL of the CC-
177 aircraft assets was determined by bootstrapping
historical datasets of all monthly payloads of CC-177
flights departing Afghanistan.
Of particular note is the distinction for waiting
times for the SLOC – upon arrival from the GLOC
at Karachi, items depart for Canada via liner service;
hence do not incur significant waiting times at port.
Conversely, items at the IST depart via contracted
sealift at specified dates. Items arriving at the IST
for transport to Canada wait until the next scheduled
ship departure date, incurring holding costs each day
until they depart.
Each type of resource has its own specified distri-
bution of the time required to reach its next destina-
tion (Canada, one of the ISTs, Karachi, etc.). As items
travel along each of the LOCs, costs of each segment
are determined stochastically, as are the durations for
which the resources are travelling along the LOCs. At
the end of the simulation, all costs are aggregated into
a final figure for the entire cost of the MTTF.
As the model involves various stochastic ele-
ments, there was a need to perform multiple runs of
the model in order to obtain a representative sample of
the various outputs of the model. The results in this
paper are based on 800 runs of the model (50 runs for
each scenario discussed).
2.2 Data
Data used in this study was extracted from the
MTTF COP spreadsheets (containing entries up to 28
November 2011) and consisted of item weights (con-
tainers, vehicles, and air pallets), flight times, air-
craft loads (weight and number of items), and the
LOC chosen for each item shipped. The MTTF COP
records report on 757 vehicles, 1408 containers, and
170 air pallets shipped out of Afghanistan (the air
pallets were shipped exclusively via ALOC Direct to
Canada). To generate realistic weights for the simula-
tion model, lognormal probability distribution func-
tions (pdf) were fitted to the container and vehicle
weights found in the MTTF COP data set. The mean
container weight was 15, 125 pounds (lbs) and the
pdf’s standard deviation was 7, 604. Vehicles weights
were found to be tri-modal due to the presence of very
heavy vehicles over 100, 000 lbs (e.g., tanks) and very
light vehicles under 20, 000 lbs (e.g., trailers). For use
in the simulation model, three pdfs for light, medium,
and heavy vehicle weights were fitted with means of
7, 443 lbs, 39, 648 lbs, and 126, 531 lbs respectively;
and standard deviations of 7, 078 lbs, 13, 123 lbs, and
5, 460 lbs respectively. The weight of air pallets was
set at 4, 500 lbs.
The MTTF COP reports on completed CC-177
flights out of Afghanistan as follows: 27 routed to
Canada, 82 to Cyprus, 75 to Kuwait. It is use-
ful to note that Cyprus was also used by the Cana-
dian Forces for troop rotation referred to as “relief-
in-place.” CC-177 assets were used to transport over
1000 passengers out of Afghanistan. Nearly a quar-
ter of CC-177 flights from Kandahar to Larnaca car-
ried a significant number of passengers (over 40) and
containers/vehicles. The MTTF COP reports on a
ICORES2013-InternationalConferenceonOperationsResearchandEnterpriseSystems
266
125 completed AN-124 flights from Kandahar to an
IST and 11 routed to Canada. MTTF flights out of
Afghanistan started in the first week of May 2011.
CF Move Planners provided cost information for
contracts (AN-124 flights, GLOC trucks, chartered
or merchant sealift). The Canadian Department of
National Defence Cost Factors Manual was used to
estimate the costs of using CC-177 assets native to
the CF. Costs are not disclosed herein, however Fig-
ure 1(a) provides insight into the relative costs of us-
ing each of the different LOCs:
• ALOC Direct using an AN-124;
• ALOC Direct using a CC-177;
• ALOC IST using an AN-124;
• ALOC IST using a CC-177;
• Sealift from an IST;
• GLOC using a truck followed by sea liner service
from Karachi.
All numbers in the figure are mean costs associated
with each of the LOCs, and are expressed as fractions
of the cost associated with the ALOC Direct method
using a CC-177 aircraft.
Since each method of shipment can carry a dif-
ferent number of items, the costs of transporting an
individual container back to Canada are illustrated in
Figure 1(b), calculated under the assumption of full
loads aboard each of the different vessels. In this fig-
ure the costs of the ALOC IST and the subsequent
sealift portion of the return are combined into one ex-
pression.
In simple terms, the different LOCs have widely
varying costs per container. The costs of travelling
directly to Canada via airlift are slightly more than
double the costs of travelling to an IST via airlift and
using sealift to return to Canada; and the costs of for-
going airlift at all and using the GLOC and subse-
quent sea liner service is approximately fifteen times
cheaper than the IST option. Moreover, the costs of
using contracted AN-124 airlift is slightly higher than
using CC-177 airlift native to the CF.
2.3 Notable Assumptions and
Limitations
In this section, the assumptions inherent in the data
and the methodology are described in detail.
Materiel and Vehicle Processing in Kandahar. CF
units handed over their equipment and vehicles to
the MTTF over several months. Vehicle and ma-
teriel production lines were established to process
and prepare the returns. Vehicles and containers
were inspected, sent for maintenance/repair, and
fumigated. All materiel was tagged and grouped
by shipping priority or destination and placed into
containers. Vehicles and containers then were
moved to the airfield or put aside for pick-up (by
truck). These MTTF efforts within Kandahar
were outside the scope of the analysis.
Modelling of Distribution within Canada. After
arriving in Canada (either in Trenton if arriving
by ALOC Direct, or in Montreal if arriving by
sealift) the items were distributed amongst the
various CF supply depots, facilities, and bases.
However, no efforts were made to model the
distribution of the items in Canada, as it was not
within the scope of the analysis.
Disposal, Transfers, and Sales. A significant por-
tion of non-essential items in Afghanistan used
by the CF were disposed (i.e., destroyed), trans-
ferred, or sold to external organizations (e.g., for-
eign militaries, non-governmental organizations,
etc.). These items were not modelled – the simu-
lation only included those items which were to be
returned to Canada.
Other Aircraft used in the Airlift. The model in-
corporates airlift handled by the CC-177 aircraft
and the contracted AN-124 aircraft. However,
there were other types of aircraft involved in
the airlift to a much lesser extent – these in-
cluded the CC-130 aircraft, which was primarily
used to transport materiel from one location in
Afghanistan to another; and the CC-150 aircraft,
which was primarily used to transport personnel
from Kandahar to the ISTs. As the model was fo-
cused on the return of materiel from Afghanistan
these aircraft were not included in the model.
Equipment used by Op ATTENTION. Operation
ATTENTION is Canada’s participation in the
NATO Training Mission - Afghanistan (NTM-A),
which delivers training and professional devel-
opment support to the national security forces of
Afghanistan. The MTTF was directed to move a
small portion of the materiel used by Operation
ATHENA to another region of Afghanistan for
use by Operation ATTENTION. As this materiel
were not repatriated by the MTTF, they were not
included in the model; nor was the airlift that
was used for their movements (which included
AN-124 and CC-130 aircraft).
Aircraft Payload: average vs. max. When deter-
mining the maximum payload able to be carried
by the aircraft under the different temperature
conditions, the actual payloads of the aircraft
were used instead of their maximum payloads,
SimulatingtheRepatriationofCanadianForcesMaterielfromAfghanistan
267
2.12
1.00
0.94
0.38
2.48
0.01
Nominal Costs
AN-124 Direct to Canada
CC-177 Direct to Canada
AN-124 to IST
CC-177 to IST
Sealift IST to Canada
GLOC and Sea Liner service
(a) The relative costs of the different LOCs.
1.14
1.00
0.53
0.40
0.03
Costs per Container
AN-124 Direct to Canada
CC-177 Direct to Canada
AN-124 to IST
+ Sealift to Canada
CC-177 to IST
+ Sealift to Canada
GLOC and Sea Liner service
(b) The relative costs per container of the different LOCs.
Figure 1: The relative nominal and per-container costs of the different LOCs are illustrated in Subfigures 1(a) and 1(b),
respectively. All numbers in the figures are expressed as fractions of the ALOC Direct cost using a CC-177 aircraft, which is
given a cost of 1.
which was unavailable to the authors. As the
items loaded on the aircraft were discrete in
number, the modelled payloads were necessarily
smaller than they were in reality. The impacts
of this assumption will be discussed further in
Section 3.1.
Warehousing Costs. Items arriving at the IST incur
holding costs each day until they depart on a ship
for transport to Canada. Due to the unavailability
of information on the the warehousing costs at the
IST, the holding costs in the model were specified
to be equal to those of comparable military ware-
houses in Canada. Holding costs at the Kandahar
airfield were not considered.
2.4 Key Outputs
After the simulation has run to completion, several
outputs are collected. These outputs include the route
taken by each item to return to Canada, along with
costs incurred, and timestamps along each section of
the route. However, the main outputs of the model
consist of the following four quantities:
Total Cost of the MTTF. This quantity consists of
all costs incurred by the MTTF, and includes con-
tracted costs of the AN-124 airlift, costs of the
CC-177 airlift (operating costs, crew costs, and
amortization costs of the equipment), costs of the
contracted sealift from the IST, sea liner service
costs from Karachi, holding costs for items at the
IST, and costs of the contracted trucks used for
transport of items for the GLOC.
Number of Flights Required. The total number of
flights (AN-124 and CC-177) required to repatri-
ate the items using the ALOC Direct and ALOC
IST methods.
Completion of 50% of the Returns. The simulated
time at which 50% of all items have been repa-
triated from Afghanistan using any of the various
LOCs.
Completion of 95% of the Returns. The simulated
time at which 95% of all items have been repa-
triated from Afghanistan using any of the various
LOCs. N.B. the return date of the final item was
not used as there were instances when vast major-
tity (in the order of 99%) of all items were repa-
triated by a given date, but a few items were de-
layed in their arrival to the IST and so had to wait
for several months for another ship to depart for
Canada. Considering the end date in such an in-
stance would have unfairly skewed the results.
3 VALIDATION OF THE MODEL
In this section, the outputs of the simulation are com-
pared to corresponding actual quantities in order to
validate the model.
3.1 Aircraft Bulk and Payloads
The first set of quantities that are compared are the
number of items on-board the aircraft flights (also re-
ferred to as the flight’s bulk), and the payloads of these
flights. All figures concerning the actual historical
quantities were taken from the COP.
When comparing these quantities for the CC-
177 aircraft, it was found that the mean simulated
bulk (3.44 items) was comparable to the mean actual
bulk (3.36 items). However, a Kolmogorov-Smirnov
test (Massey, 1951) found that the null hypothesis that
the datasets have the same distribution was rejected
at the 95% confidence level. A similar result was
found when the mean simulated payload (64,700 lb)
was compared to the mean actual payload (71,500 lb).
These distributions are presented graphically in Fig-
ure 2.
The difference in these distributions may be ex-
plained by the simulation using an unrefined method
when selecting items to load on to the aircraft (in the
order in which the items were received at the loading
ICORES2013-InternationalConferenceonOperationsResearchandEnterpriseSystems
268
area), whereas in actuality experienced load masters
carefully select items in order to maximize the load
of the aircraft. Evidence for this conjecture comes
from the fact that the average unused payload of the
CC-177 aircraft was 4,100 lb in the simulation; and
when this quantity was added to the simulated pay-
loads, the null hypothesis was not rejected at the 95%
level (with a p-value of 0.42).
The simulated and actual bulk and payload quanti-
ties were also compared for the AN-124 aircraft. The
mean simulated bulk of the AN-124 (6.38 items) was
comparable to its mean actual bulk (6.25 items), and
in this case the Kolmogorov-Smirnov test did not re-
ject the null hypothesis at the 95% confidence level
(with a p-value of 0.27). Moreover, the Kolmogorov-
Smirnov test did reject the null hypothesis at the 95%
confidence level when comparing the payloads of the
AN-124 aircraft (the mean simulated payload was
149,600 lb and the mean actual payload was 157,600
lb). These distributions are presented graphically in
Figure 2.
The average unused payload of the AN-124 air-
craft was 6,900 lb in the simulation. When this quan-
tity was added to the simulated payloads, the null hy-
pothesis that the simulated plus unused payload of
the AN-124 and the actual payload datasets have the
same distribution was not rejected at the 95% level
(with a p-value of 0.12). Hence the difference in
the payloads may be explained by the aircraft load-
ing method, which can reduce the sizes of the unused
payloads on the aircraft.
3.2 Repatriation as a Function of Time
One of the key measures used by the MTTF to es-
timate the progress made in the repatriation was the
number of items (containers and vehicles) that had
been returned to Canada as a function of time. One
item listed in the MTTF COP is the number of items
that had left the main Canadian base in Afghanistan
(Kandahar Airfield) by week.
The actual number of items repatriated was com-
pared to their corresponding simulated quantities.
Both series can be found in Figure 2(e). The multiple
series shown in the figure for the simulated quantities
are due to the variance in the runs of the simulation.
Note that the number of items returned to Canada
as a function of time in the simulation differs
from the actual quantities by at most 12% un-
til day 180 (approximately November 2011), when
the series diverge substantially. This divergence
corresponds to the date of creation of a hold-
ing yard off-base where items could be stored
for eventual onward movement to Karachi via
the GLOC (Department of National Defence (DND),
2011). Once created, items placed in the holding yard
were considered off-base from a CF perspective, and
were counted as such in the MTTF COP.
3.3 Number of Flights Required
Finally, the last point of validation concerned the
number of flights needed to complete the ALOC por-
tion of the repatriation. The number of flights (CC-
177 and AN-124) needed in the simulation ranged
from 333 to 370, with a median of 351. The actual
number of flights needed was 357, which corresponds
to the 75% quartile in the range found in the simu-
lation. This information is presented graphically in
Figure 2(f).
3.4 Validation Summary
Given that the simulation satisfactorily models reality
to the extent that it can replicate the number of flights
needed to return items from Afghanistan to Canada,
at the correct rate, and subject to similar constraints
on the load placed on each aircraft, the simulation is
representative enough of reality to be an illustrative
model for sensitivity analyses.
Up to this point in the paper the model has been
used as a descriptive model of reality. In the next
sections, the model is used to evaluate the impact
of changes to physical or procedural aspects of the
MTTF repatriation efforts.
4 RESULTS AND DISCUSSION
Several versions of the model were run with changes
made to its various parameters to determine the im-
pact of these changes on the four measures of per-
formance of the MTTF (specified in Section 2.4: to-
tal cost, number of flights, 50% completion time, and
95% completion time). The unaltered version of the
model is referred to below as the baseline model. The
relative increases or decreases of the outputs over the
baseline are specified in this section.
The analyses performed were fivefold - wherein
one (and only one) of the following things were var-
ied, with all else remaining constant:
1. Increases to the number of items returning by
ALOC Direct, with proportional removal from
one of the other two methods.
2. Decreases to the number of items returning from
GLOC, with increases going to the number return-
ing by ALOC IST;
SimulatingtheRepatriationofCanadianForcesMaterielfromAfghanistan
269
2 4 6 8 10
CC-177 Bulk: Simulated and Actual HNumber of ItemsL
Actual
Simulated
(a) The number of items aboard the CC-177 flights.
20 000 40 000 60 000 80 000 100 000
CC-177 Payload: Simulated and Actual HlbL
Actual
Simulated
(b) The payloads of the CC-177 flights.
5 10 15
AN-124 Bulk: Simulated and Actual HNumber of ItemsL
Actual
Simulated
(c) The number of items aboard the AN-124 flights.
50 000 100 000 150 000
AN-124 Payload: Simulated and Actual HlbL
Actual
Simulated
(d) The payloads of the AN-124 flights.
0 100 200 300 400 500
0
500
1000
1500
2000
Simulation Date
Number of Items Returned
Actual
Simulated
(e) The number of items returned to Canada as a function of time (in days).
Note that Day 1 corresponds to May 1, 2011.
330 340 350 360 370
Number of Flights Required: Simulated and Actual
ActualMedianHSimL
(f) The total number of CC-177 and AN-124 flights required.
Figure 2: Comparison of several actual and simulated quantities. Subfigures 2(a) to 2(d) illustrate the bulk and payloads of
the CC-177 and AN-124 flights departing Afghanistan in support of the MTTF. Subfigure 2(e) presents the number of items
returned to Canada as a function of time; and Subfigure 2(f) presents the total number of flights required for the ALOCs.
3. Ordering of the types of aircraft used – selecting
the AN-124s ahead of the CC-177s if both were
available, or random assignment to either type of
aircraft (the baseline assumes loading of the CC-
177s first if they are both available);
4. Delaying the first scheduled departure of all
flights (ALOC Direct, and ALOC IST) by several
months (30, 60, 90, or 120 days) to take advantage
of the lower temperatures in the later months; and
5. Increases to the availability of the different air-
craft (AN-124s and CC-177s), i.e., the number
of aircraft of each type available for use in the
ALOCs each day.
The results of each of these changes is discussed in
sequence in the following subsections.
4.1 Increases to ALOC Direct
It was hypothesized that increasing the number of
items transported via ALOC Direct would decrease
the total time to complete the total return but would
increase the total costs (due to increases in the number
of flights required). The analysis found that this hy-
pothesis was correct: increasing the number of items
ICORES2013-InternationalConferenceonOperationsResearchandEnterpriseSystems
270
returning by ALOC Direct, and correspondingly de-
creasing the number of items returning via one of
the other two methods (ALOC IST or GLOC), would
have had the effect of increasing costs while decreas-
ing the total time to complete the total return by ap-
proximately equal amounts.
More specifically, doubling the number of items
returning by ALOC Direct would have had the effect
of increasing the total cost by 10% while decreasing
the time to complete 95% of the repatriation by 7%. If
instead the number of items returning by ALOC Di-
rect was increased fourfold, the total cost would have
increased by 28% and the time to complete 95% of the
repatriation would have decreased by 22%. The re-
sults for this modification are presented in Figure 3(a).
4.2 Decreases to GLOC
Another possibility for speeding up the repatriation
was a reduction in the number of items requiring
transport by GLOC, which was thought to be the
slowest of all options. It was found that decreasing
the number of items returning by GLOC (and subse-
quent movement by SLOC) would have had the effect
of increasing costs and the number of flights required,
and has a marginal effect on the time to complete half
the returns. However, it would have had a substantial
effect on the time to complete 95% of the return of
the items. This result is due to the fact that the the
airlift portion of the repatriation is done much earlier
than the sealift portion; hence, hastening the sealift
portion of the returns has a large impact on the time
required for the total return.
More specifically, decreasing the number of items
returning by GLOC by 67% of its original value
would have had the effect of increasing the total cost
by 14%, increasing the number of flights by 18%, and
decreasing the time to complete 50% of the returns by
3%. However, the time to complete 95% of the repa-
triation would have decreased by 47%. The results for
this modification are presented in Figure 3(b).
4.3 Ordering of the Aircraft
AN-124 aircraft are contracted airlift vehicles, and
are thus more expensive than the CC-177 fleet which
is native to the CF. It was suspected that if more
items travelled by CC-177 instead of AN-124, costs
incurred by the MTTF could be reduced.
However, it was found that changing the ordering
of the aircraft for loading and departing for the ALOC
(Direct or IST) would have had a negligible effect on
all outputs studied. None of the outputs would have
varied by more than 1% by selecting the AN-124s
ahead of the CC-177s if both were available as op-
posed to the baseline case of selecting the CC-177s
first. A similar result was found for the case of ran-
domly selecting either type of aircraft instead of al-
ways selecting the CC-177s first. The results for this
modification are presented in Figure 3(c).
4.4 Delaying the Aircraft Departures
Recall that the external air temperature has a large ef-
fect on the maximum allowable payload for both the
AN-124 and CC-177 aircraft. It was hypothesized
that by delaying the first scheduled departure of all
flights (for ALOC Direct as well as ALOC IST) by
several months (30, 60, 90, or 120 days), one could
take advantage of the lower temperatures in the later
months, potentially reducing the flights required.
It was found that that these delays would have had
a small effect on total costs (up to a 5% savings), but
the time needed to complete half the returns was in-
creased by substantial amounts (up to 38%). How-
ever, the time to complete 95% of the returns would
have been unchanged due to the final items always
awaiting a ship for sealift via the SLOC. The results
for this modification are presented in Figure 3(d).
4.5 Increasing Aircraft Availability
It was hypothesized that if there were more aircraft
(CC-177s and AN-124s) available to the MTTF, the
repatriation of the items may have progressed at a
faster rate. This analysis tested that hypothesis, by
increasing the number of aircraft available for ALOC
purposes by 33%, 66%, or 100%.
It was found that increasing the availability of
the aircraft would have increased the cost and total
flights by small amounts (up to 4% and 6%, respec-
tively), but the time to complete 50% of the returns
would havebeen dramatically reduced - by up to 35%.
Again, the time to complete 95% of the returns would
have been unchanged due to the dates of the sealift
portion of the returns. The results for this modifica-
tion are presented in Figure 3(e).
5 CONCLUSIONS AND
RECOMMENDATIONS
The objective of this study was to develop a model
that could be used to analyze the repatriation of Cana-
dian equipment from Afghanistan to Canada via the
LOCs, in order to enable and improve future mission
planning.
SimulatingtheRepatriationofCanadianForcesMaterielfromAfghanistan
271
(a) Effects of increasing the number of items returning by ALOC Direct, and correspondingly decreasing the number of items returning via one of the other two
methods (ALOC IST or GLOC).
7% 14% 23%
Total Cost
10% 18% 28%
Num Flights
-1%
-3%
3%
50% Completion
-36%
-47%
-43%
95% Completion
Decrease GLOC by 33%
Decrease GLOC by 67%
Decrease GLOC by 100%
(b) Effects of decreasing the number of items returning by GLOC, and correspondingly increasing the number of items returning via ALOC IST.
1% 0%
Total Cost
0%
Num Flights
1%
0%
50% Completion
-1%
0%
95% Completion
AN-124 before CC-177
Random assignment
(c) Effects of altering the ordering in which the aircraft were selected – selecting the AN-124s ahead of the CC-177s if they were both available, or random
assignment to either type of aircraft (the baseline assumes loading of the CC-177s first if they are both available).
0%
0%
-4%
-5%
Total Cost
0%
-1%
-4%
-6%
Num Flights
3% 7% 22% 38%
50% Completion
0%
0%
0%
0%
95% Completion
Delay first flight by 30 days
Delay first flight by 60 days
Delay first flight by 90 days
Delay first flight by 120 days
(d) Effects of delaying the first scheduled departure of all flights (ALOC Direct, and ALOC IST) by several months (30, 60, 90, or 120 days) to potentially take
advantage of the lower temperatures in the later months.
1% 4% 4%
Total Cost
2% 5% 6%
Num Flights
-16%
-29%
-35%
50% Completion
0%
0%
0%
95% Completion
Increase aircraft avail. by 33%
Increase aircraft avail. by 67%
Increase aircraft avail. by 100%
(e) Effects of increasing the availability of the different aircraft (AN-124s and CC-177s) for use in the ALOC.
Figure 3: The results of specific changes to the model’s parameters are shown in each of the subfigures. The relative changes
of each of the key outputs (total cost of the MTTF, the number of flights required, completion of 50% of the materiel returns,
and completion of 95% of the materiel returns) are illustrated as relative increases or decreases to the baseline value found in
the unchanged version of the model.
A discrete-event simulation model was developed
and shown to be representative of actual the repatria-
tion of items from Afghanistan. The model was then
used to determine the impacts of different potential
courses action, measured mainly through results on
the total cost and duration of the returns.
ICORES2013-InternationalConferenceonOperationsResearchandEnterpriseSystems
272
Results indicate that increases to the number of
items using the ALOC Direct method of shipment
would have both increased total cost and decreased
the time to complete the repatriation. Decreasing the
number of items returning by GLOC would have in-
creased cost to a moderate extent while significantly
decreasing the time to complete the repatriation. De-
laying the departure of the flights to the cooler months
would have decreased the total costs as well as the
number of flights required of the aircraft while hav-
ing no effect on the time to complete the repatria-
tion. Finally, increasing the availability of the air-
craft would have increased the cost and the number
of flights while having no effect on the time to com-
plete the repatriation.
There is a subtle and complex relationship be-
tween the parameters of the model and its main out-
puts (the total cost incurred and the time to complete
the repatriation) that is due to the scheduling of the
returns via airlift and sealift, as well as the overall
per-container cost for each different method of ship-
ment. This type of model can provide insights into
the factors affecting these outputs – insights which
would perhaps otherwise go unnoted. It is recom-
mended that in future operations a model similar to
one detailed here be constructed to investigate how
the repatriation could be optimized prior to the ini-
tial departures and during mission execution. More
specifically, this model can be used in the repatriation
of the materiel from Operation ATTENTION, which
is expected to conclude in 2014.
REFERENCES
Baker, S., Morton, D., Rosenthal, R., and Williams, L.
(2002). Optimizing military airlift. Operations Re-
search, 50(4):582–602.
Burke, J., Love, R., and Macal, C. (2004). Modelling
force deployments from army installations using the
transportation system capability (transcap) model: A
standardized approach. Math. Comput. Model., 39(6-
8):733–744.
Campbell, G. and Moorhead, P. (2010). A movement op-
tions analysis simulation tool for the canadian opera-
tional support command. In Proceedings of the 2010
Winter Simulation Conference, B. Johansson, S. Jain,
J. Montoya-Torres, J. Hugan, and E. Yucesan, eds.
Dantzig, G. and Ferguson, A. (1955). The problem of rout-
ing aircraft: a mathematical solution. Aeronaut. Eng.
Rev., 14:51–55.
Department of National Defence (DND) (2011). Mission
transition task force closure of operation athena -
2011. Technical report.
Ghanmi, A., Kaluzny, B., Kim, B., and Shaw, R. (2009). An
optimization model for airlift load planning: Galahad
and the quest for the holy grail. In Proceedings of the
IEEE Symposium on Computational Intelligence for
Security and Defence Applications, Ottawa, Canada.
Ghanmi, A. and Shaw, R. (2008). Modeling and analysis of
canadian forces strategic lift and pre-positioning op-
tions. Journal of the Operational Research Society,
59:1591–1602.
Kelton, W., Sadowski, R., and Swets, N. (2010). Simulation
with Arena. McGraw-Hill, 5th edition.
Massey, F. (1951). The Kolmogorov-Smirnov Test for
Goodness of Fit. J. Am. Stat. Ass., 46(253):68–78.
Powell, W., Whisman, A., and Wu, T. (2009). The
optimizing-simulator: An illustration using the mil-
itary airlift problem. ACM Trans. Model. Comput.
Simul., 19(3).
Schank, J., Mattock, M., Sumner, G., Greenberg, I., and
Rothenberg, J. (1991). A review of strategic mobility
models and analysis. Technical report, RAND Corpo-
ration.
SimulatingtheRepatriationofCanadianForcesMaterielfromAfghanistan
273