Analytical Models of Rail - Transportation

Transportation Research Part E 93 (2016) 294315
Contents lists available at ScienceDirect
Transportation Research Part E

j o u r n a l h o m e p a ge : w w w.e l s e v i e r.c o m / l o c a t e / t r e
Analytical models of rail transportation service in the grain

supply chain: Deconstructing the operational and economic
advantages of shuttle train service
Michael F. Hyland
a,c
, Hani S. Mahmassani
b,c,
, Lama Bou Mjahed
a,c
a
Northwestern University Civil & Environmental Engineering, Transportation Center, 600 Foster Street, Evanston, IL 60208, United States
Northwestern University, 215 Chambers Hall, 600 Foster Street, Evanston, IL 60208, United States
c
Northwestern University, Transportation Center, 600 Foster St., Evanston, IL 60208, United States
b
a r t i c l e
i n f o
Article history:
Received 4 February 2016
Received in revised form 17 June 2016
Accepted 20 June 2016
Available online 4 July 2016
Keywords:
Rail transportation
Rail cost model
Shuttle trains
Grain export supply chain
Terminal elevators
a b s t r a c t
This paper introduces conceptual and mathematical models of the domestic grain supply
chain incorporating trucking, elevator storage, and rail transportation. We compare conventional rail service supported by country elevators with shuttle service supported by terminal elevators across three critical transportation service dimensions: travel time, cost,
and capacity. Even after taking into account trucking and elevator storage, the time and
cost model results indicate that shuttle service transports grain faster and reduces logistical supply chain costs, respectively, relative to conventional service. The rail capacity
model results demonstrate that shifting grain from conventional to shuttle service significantly increases rail capacity.
2016 Elsevier Ltd. All rights reserved.
1. Introduction
Grain production costs in the United States are significantly higher than in South American countries that compete on
the global market (e.g., Brazil and Argentina); however, the United States domestic (i.e., inland) transportation system is
significantly more efficient in terms of cost, speed, and reliability (Frittelli, 2005). These efficiencies are the result of
significant restructuring within the grain industry and its associated transportation services over the past twenty-thirty
years. Main- taining a competitive advantage in terms of transportation costs and efficiency is crucial in order for the
United States to continue selling grain to export markets. The analysis presented in this paper examines and
quantifies the advantages and disadvantages of the restructured grain logistical supply chain. For context, the analysis is
patterned on the Upper Mid- west region (i.e., North Dakota, South Dakota, and western Minnesota) of the United States,
though the models consider a highly idealized, and general, context.
The restructuring of the grain supply chain included the introduction of (1) shuttle train service by the railroads, and
(2) larger, more efficient, storage elevators (henceforth referred to as terminal elevators) by grain shippers. The restructuring
of the grain supply chain coincided with increased yield volumes, due to technological advances and weather changes, on
the grain production side. Additionally, the railroads promoted and incentivized the construction of terminal elevators
throughout the Upper Midwest.
Corresponding author at: Northwestern University, 215 Chambers Hall, 600 Foster Street, Evanston, IL 60208, United States.
E-mail addresses: michaelhyland2013@u.northwestern.edu (M.F. Hyland), masmah@northwestern.edu (H.S. Mahmassani), lamabm@u.nortwestern.edu

(L. Bou Mjahed).
http://dx.doi.org/10.1016/j.tre.2016.06.008
1366-5545/ 2016 Elsevier Ltd. All rights reserved.
M.F. Hyland et al. / Transportation Research Part E 93 (2016) 294315
295
The motivation for providing shuttle and unit train service lies in a simple realization that freight rail transportation is
most cost efficient when a large number of railcars move together on a single train. As the number of railcars on a train
increases, the costs of rail transportation are spread over more railcars, resulting in a lower cost per railcar; hence, railroads
aim to move as many railcars as possible on a single train. In order to obtain the cost efficiencies associated with large trains,
railroads have historically consolidated railcars at classification yards. Under conventional rail service, single or multiple railcars
are typically moved from their origin to a nearby classification yard where they are grouped with other railcars traveling to a
similar destination. Railcars often travel through multiple classification yards before reaching their destination. The disassembly
and reassembly of railcars at classification yards is a time and resource consuming process, subject to considerable variability,
and hence is a source of unreliability in service times. Historically, railroads and shippers have accepted the inefficiencies of
classification yards as the price of railcar consolidation (Keaton, 1991). However, the introduction of shuttle train service for
bulk commodities, such as coal and grain, allowed railroads to move a large number of railcars directly from origin to destination, thus bypassing classification yards.
In order for the railroads to move 100 or more railcars directly from origin to destination, it is necessary for the grain
storage elevators at the origin and destination to be able to load and unload 100 or more railcars in a reasonable time. Prior
to the restructuring of the grain supply chain, most grain elevators either did not have enough storage capacity to fill 100
railcars with grain or it would take the elevators multiple days to load an entire shuttle train with grain. Hence, the
railroads decided to incentivize existing elevator owners and other grain industry stakeholders to either retrofit their
existing elevators or build new, terminal, elevators in order to efficiently load 100 or more railcars in a short period.
Because of the incen- tives, larger, more efficient terminal elevators began popping up throughout the grain producing
regions of the United States. These larger, more efficient terminal elevators handle the same volume of grain as
multiple country elevators (U.S. Department of Agriculture, 2014). This shift from single and multi-railcar shipments
originating at country elevators to
100 or more railcar shipments originating at terminal elevators increased the efficiency of grain transportation in many
ways; however, it did result in two negative impacts: longer trucking distances between farms and storage elevators and
potentially longer storage times at elevators. The longer trucking distances stem from the fact that each terminal elevator
handles the same volume of grain as multiple country elevators. Hence, the draw areas of terminal elevators in the restructured grain supply chain are much larger than the draw areas of country elevators prior to restructuring. Moreover, many
grain elevator cooperatives now truck grain from multiple country elevators, which act as feeders, to a single terminal elevator. The longer storage times at terminal elevators are due to the fact that shippers need to consolidate more grain to fill
100 railcar shipments than one-two railcar or even six-26 railcar shipments. Therefore, while it is clear that transporting
grain over the rail network via shuttle service is more efficient than moving grain on mixed-manifest trains, previous
studies have not explicitly examined whether or not the negative impacts that may occur upstream of the rail network
outweigh the benefits on the rail network. The analysis presented in this paper aims to fill this gap, while providing a
simple framework that helps convey the main trade-offs involved. The approach adopted is to present a stylized
representation of the under- lying system, capturing essential features without the clutter of non-essential detail.
This paper presents three analytical models to determine the advantages and disadvantages, from the perspective of
global competitiveness, of the restructured grain collection and transportation system (shuttle service and terminal
elevators) relative to the old system (conventional service and country elevators). The three mathematical models address
three critical performance dimensions of grain transportation, namely time, cost, and throughput. The analysis explicitly
takes into con- sideration the components of the grain supply chain upstream of the rail transportation network.
Understanding the tradeoffs associated with these two types of grain collection/transportation systems across the entire
supply chain, not just on the rail network, is crucial to providing low cost, fast, and reliable transportation to the grain
industry that allow it to compete in export markets.
The remainder of this paper is structured as follows: Literature related to grain logistics and transportation is reviewed
in Section 2. Section 3 presents a conceptual model of a grain transportation network and lists the necessary model inputs
and model outputs. Sections 46 present the time, cost, and throughput models, respectively. Each of these three sections
presents mathematical models, base parameter values, numerical results, and sensitivity analyses. Lastly, Section 7
presents conclusions along with potential extensions to the capacity model.
2. Literature review
This section reviews the literature related to the grain supply chain. The integrated models of the grain supply chain
presented in this paper are rooted heavily in the existing literature. In fact, many of the individual component models,
model assumptions, and parameter values used in this paper to compare shuttle service/terminal elevators with
conventional service/country elevators come from the existing literature.
Frittelli (2005) finds that between 1980 and 1998: considerable consolidation occurred in grain production, rail transportation, and grain storage. The number of farms decreased by 15% but farm size increased by 11%; the number of terminal
1
Shuttle trains and unit trains are defined differently throughout the literature and in the rail industry. Consistent with most recent studies, the term shuttle
train in this paper refers to 90120 railcar shipments wherein the railcars and the locomotives that power shuttle trains are not detached from one another
during normal service (Sparger and Prater, 2012); whereas, the term unit train refers to 5055 railcar shipments that are often combined with other single
railcar, multi-car, and unit train shipments (Frittelli, 2005).
elevators increased but the total number of grain elevators decreased dramatically due to country elevators becoming
obso- lete; farms now own more trucks and those trucks are more likely to transport grain directly from farm to terminal
elevator bypassing country elevators and short line railroads; and railroads abandoned a large portion of their short lines.
Griffin et al. (1984), Tolliver et al. (2005) and Vachal (2015) surveyed farmers in the Upper Midwest about their on-farm
storage capacity, truck fleets, truck transportation habits, and marketing patterns in the years 1984, 2000 and 2013, respectively. The 2000 survey focuses on wheat and barley movements whereas the 2013 survey includes wheat, soybeans, and
corn. The surveys yielded a number of findings relevant to the models, model assumptions, and parameter values used in
this paper. The average trucking distance from farm to local market for wheat increased from approximately 12 miles to
32 miles between 1984 and the early 2000s. The average trucking distance for wheat, corn, and soybeans remained around
30 miles between the early 2000s and 2013. The number of terminal elevators nearly doubled over this same period. Even
back in 2000, nearly all farmers owned their own trucks (90%) or used leased trucks (4%) to transport their own grain to
market. Farmers have a large amount of storage capacity on their farms, which allows them to sell their grain
throughout the entire year, spreading out the demand for rail transportation service. In 2013, 36% of wheat, 32% of corn, and
66% of soybeans were sent directly to market after harvest. Farmers increased the hauling capacity of the trucks in their
fleets considerably between 1984 and 2013. In 1984, 80% of grain trucks were single-axle compared with 18% (52%) in
2013 (2000).
Between 2008 and 2012, the American Transportation Research Institute surveyed for-hire trucking carriers to analyze
the operational costs of trucking (Fender and Pierce, 2013). They separate operating costs into two categories, vehiclebased and driver-based. Vehicle-based costs include fuel, purchase and lease payments, repair and maintenance, insurance,
tires, permits, and tolls. The driver-based costs include wages and benefits. Fuel costs and labor costs are the two largest
components according to their results. Berwick and Dooley (1997) develop truck cost models for owners/operators of trucks
that are later updated by Berwick and Farooq (2003). Their literature review finds that variable trucking cost models
generally include the following parameters, maintenance and repair, fuel, labor, and tires. The study finds that the
utilization rate is the most important factor in the cost of truck transportation.
Bangsund and Leistritz (2005) and Bangsund et al. (2011) use cost models of each component of the grain supply chain
to examine the economic contributions, direct and secondary, of the wheat industry and the soybean industry to the state
of North Dakota (ND), respectively. According to Bangsund et al. (2011), the soybean industrys direct annual impact to
the state of North Dakota via soybean production, handling, truck transportation, and rail transportation were $1.1
billion,
$27.5 million, $10.9 million, and $38.9 million, respectively. Bangsund et al. (2011) use the same variable and fixed
trucking cost model as Berwick and Dooley (1997). To estimate rail transportation costs they employ Tolliver et al. (1987)
cost model, which includes the following variable costs: train crew (44% of the total variable cost), locomotive (23%),
railroad car (21%), and a miscellaneous transportation charge (11%). The country elevator cost model includes the following
costs: labor, taxes and licenses, insurance, utilities, services, interest, equipment depreciation and repairs, and general
expenses.
Bahizi, in a feature article in a USDA weekly Grain Transportation Report (U.S. Department of Agriculture, 2014), explains
the role of grain elevators in the marketing, storage, and cleaning of grain. Bahizi explains how elevators make money by
not only charging for storage but also buying grain from farmers at a given price and selling it to the next entity in the
grain supply chain at a higher price. Bahizi highlights how the shift towards larger shipments and shuttle-loading
facilities has negatively affected country elevators located far from Class I railroad mainlines. According to Vachal et al.
(1997) elevators with higher storage capacity, higher annual throughput volumes, high diversity of grains handled, and
more services have longer survival times. Elevators owned by larger companies and elevators with a high proportion of
grain transported via truck had lower survival times.
Kenkel et al. (2004) develop a variable cost model of terminal elevators as part of a broader cost-benefit analysis of
upgrading a country elevator to load 100110 railcar shuttle trains. The study breaks down the costs of constructing and
operating shuttle-loading facilities into initial infrastructure investment costs, annual operating costs, and variable operating
costs. Kenkel et al. (2004) list overtime (wages), grain inspection, grain inspection overtime, electricity, and grain cleaning
(labor, electricity, and shrinkage) as the variable costs of shuttle loading elevators. The report found that the results of the
cost-benefit analysis were heavily dependent on the elevators annual volume of grain handled. The break-even grain volume
was around 8 million bushels of wheat per year.
Kenkel (2008) examines the handling and storage costs of country grain elevators. Kenkel reports that fixed costs historically have accounted for 2/3rds of an elevators total cost. The high proportion of total cost being fixed incentivized
elevators to maximize throughput at small margins. However, Kenkel (2008) suggests that with the increasing price of
grain and cost of electricity, the variable cost accounted for 52% of the total cost in 2008. A higher market price of grain
increases inventory costs as well as shrinkage and loss costs. The variable storage and handling costs in their model include
grain shrinkage and moisture loss, conveying, turning and aeration, and fumigation expenses.
An Association of American Railroads (AAR) report provides a concise overview of the current state of grain rail transportation at a high-level and highlights the complexities associated with understanding grain production and
transportation due to the interaction between various stakeholders (Association of American Railroads, 2015).
According to Moore (2013), there are two types of railroad costing models: the more common statistically estimated
models of reported freight rates (see Ndembe, 2015) and engineering cost models. The Surface Transportation Boards
(STB) Uniform Railroad Costing System (URCS) is the best-known rail cost model. The URCS model is a complex engineering
cost model that uses reported railroad expenses and activity data to determine the costs of specific railroad movements.
Despite its complexity, URCS does a relatively poor job of estimating the cost of individual rail shipments. The methods used
by STB to assign variable costs to shipments by allocating portions of a railroads total expenses are economically invalid and
produce unreliable results because most railroad costs are shared by traffic and cannot be unambiguously divided and allocated to individual units of traffic. (Transportation Research Board, 2015, p. 4). The railroad cost model in this paper is
much simpler and is tailored toward the movement of grain via shuttle trains and conventional mixed-manifest trains. It
captures the key variable cost differences between shuttle and conventional railroad service.
A recent empirical study by Ndembe (2015) examines the impact of shuttle trains on railroad pricing for red spring
wheat in North Dakota via regression models. The paper notes the huge increase in transport via shuttle trains for grain
across the United States and especially wheat movements in North Dakota. Ndembe (2015) finds that price rates for
shuttle trains are significantly lower than other shipment size categories including single car, multi-car, and unit trains.
Other significant model parameters include distance from the nearest barge facility, total shipment distance, load factor
(tonnage per railcar), origination region, season, and year.
Notwithstanding (1) the large volume of literature examining components of the grain supply chain, (2) the importance
of having an efficient grain collection and transportation system (especially to compete in export grain markets), and (3)
the significant restructuring of the grain supply chain that has occurred over the past two-three decades, no study in the
literature appears to have attempted to model the entire grain supply chain in order to quantify the advantages and
disadvantages of the restructured grain supply chain. The models and analysis presented in this paper aim to fill this
gap. The objective of the paper is to provide robust insights into the advantages and disadvantages of the restructured grain
collection and transportation system (terminal elevators/shuttle rail service) relative to the old system (country
elevators/conventional rail service). In order to provide these insights, we present simple mathematical models that aim to
capture the key differences between the two grain collection/transportation systems. A preliminary analysis is performed
using the mathematical models and reasonable parameter value estimates obtained from primary and secondary sources.
The analysis uncovers a very important result: even after taking into account longer trucking distances and increased
storage time, the restructured grain transportation and collection system moves grain faster, at a lower cost, and has
higher maximum throughput on the rail network than the old system. Taking into account the upstream impacts of the
restructured grain supply chain is crucial in performing a fair comparison of the two grain collection/transportation
systems. Moreover, we show that the results hold across a range of parameter values.
3. Modeling the grain supply chain
3.1. Conceptual model
Grain moving from the Upper Midwest to the Pacific Northwest travels a long distance (e.g. 1500 miles from Fargo, ND
to Portland, OR). The three most common modes for grain transportation include barges, railroads, and trucks. There is no
fea- sible waterway between North Dakota and the Pacific Northwest for barge transportation; moreover, trucking lowvalue commodities such long distances is prohibitively expensive (Vachal and Bitzan, 2002; Frittelli, 2005). Hence, rail
transportation is the ideal mode of transportation for grain moving from the Upper Midwest to the Pacific Northwest
ports.
Fig. 1 presents a conceptual model of the grain logistical supply chain prior to and after restructuring occurred. Trucks
are used for the initial leg of grain transport from the farm to either a country elevator, in the case of the old grain supply
chain, or a terminal elevator, in the case of the restructured grain supply chain. In the restructured grain supply chain,
terminal elevators are (almost) exclusively located on the mainline of the rail network; whereas, in the old grain supply
chain, country elevators were located on short lines throughout the region. As displayed in Fig. 1, the average trucking
distance from the farm to the nearest terminal elevator is, on average, considerably longer than was the distance from the
farm to the nearest country elevator.
After the grain is transported from the farm to a local elevator via truck, the grain remains in storage until the elevator
consolidates enough grain to fill an order. A shuttle service order is typically between 90 and 120 carloads of grain,
whereas, a conventional service order ranges from a single carload to 55 carloads. After the elevator consolidates enough
grain to fill an order, railcars arrive and move the grain from the local elevator to the export elevator over the rail network.
Grain railcars traveling via conventional service are processed at classification yards on the rail network, whereas, shuttle
trains bypass classification yards and travel directly on the rail network from terminal elevator to export elevator. Bahizi, in
a USDA report (U.S. Department of Agriculture, 2014), conceptualizes the grain supply chain before and after the
introduction of shuttle trains and terminal elevators in a manner that is consistent with Fig. 1.
3.2. Model inputs and outputs
The principal input to the time and engineering cost models is the grain rail transportation demand rate, which has
units of tons per day. The grain rail transportation demand rate is simply the rate at which grain from a given area of
land surrounding a local elevator arrives at the local elevator en route to its final destination. The demand rate depends on
a number of factors including the number of farms served by an elevator, the production rate of the farms during the
current season, the percentage of farms selling their yield to an elevator (as opposed to selling it to other entities, e.g. flour
mill or ethanol plant, or storing it and waiting for a better price). Capturing all of these factors is beyond the scope of this
analysis; therefore,
Fig. 1. Conceptual depiction of the restructured and old grain supply chains.
rather than assuming a fixed demand rate, we vary the demand rate between 50 tons per day and 5000 tons per day, an
extremely large range. Tolliver et al. (2006) states that during the peak harvest period a large terminal elevator may
generate
225 trucks per day. Assuming 20 (30) tons per truck, that is equivalent to 4500 (6750) tons per day. The Alton Grain terminal
elevator in North Dakota handles 25 million bushels of grain per year, which is approximately 19002700 tons per day on
average. Eq. (3.1a) (Eq. (3.1b)) calculates the average tons per day assuming elevators operate 365 (260) days per year. Section 3.3 provides the conversion rate between tons and bushels used in the equation below and throughout the report.
25; 000; 000 bush

year
1 year
365 days
0:0275 tons
25; 000; 000 bush

year
1 year
260 days
0:0275 tons
1 bush
1 bush
1883
2664
tons
day
tons
day
3:1a
3:1b
The capacity model includes a second input in addition to the demand rate: the percentage of rail transport demand
served by shuttle trains. The analysis varies the percentage of demand served by shuttle service between 0% and 75% in
increments of 25%. Naturally, the percentage of demand served by conventional service is equal to 100% minus the percent
served by shuttle service. The capacity model also includes background traffic (which remains fixed) representing other
commodities moving through the network.
The time model outputs the total time it takes to transport grain from the farm to export elevator. The cost model
returns the cost per ton to transport the grain from the farm to the export elevator. The throughput model returns the total
number of railcars per day that the rail network can handle.
3.3. Modeling assumptions and unit conversions
The following assumptions hold for all three models. First, the following unit conversions are assumed:
1
1
1
1
bushel of grain = 55 pounds = 0.0275 tons.

jumbo hopper = 286 k pounds (220 k pounds of grain + 66 k tare) = 110 tons of grain.
shuttle train = 110 railcars = 12,100 tons of grain.
conventional train = 24 railcars = 2640 short tons of grain.
Because country elevators are sometimes located off the mainline track, bridges on some branch lines have weight
restrictions that prevent railcars from being loaded with 110 tons of grain. However, we make the conservative
assumption2 that conventional train railcars, as well as those on shuttle trains, hold 110 tons of grain. In this paper, twentyfour railcars are associated with conventional service because, according to the USDA (Prater et al., 2013), shuttle service is
disproportionately replacing 626 railcar blocks.
In order to perform the intended comparison between the restructured grain supply chain (shuttle train service and
terminal elevators) and the old grain supply chain (conventional rail service and country elevators) the following
additional simplification is made. We assume that the old system did not include any terminal elevators. We also
assume that in the new system all the elevators are terminal elevators. Moreover, we assume that one terminal elevator
handles the same volume of grain as three country elevators. Discussion of these assumptions, and their implications on
the model results are presented in Section 5.1.
4. Time model
The time model presented in this section measures the amount of time it takes for a unit of grain to move from a production node (i.e. a farm) to the export terminal. The time model is subdivided into three components: the time it takes to
truck grain from a farm to the nearest country or terminal elevator for consolidation and temporary storage; the time that
the grain remains in storage before a train arrives to pick up the grain; and the time that the grain takes to travel from
the local country or terminal elevator to the export elevator over the rail network. The ability of shippers to transport their
goods to market quickly is essential (Swan and Tyworth, 2001), especially in a competitive global market wherein
customers are located all over the world.
4.1. Time model parameter values and mathematical formulation
Each of the three components of the time model are presented in this section. For the variables that have different base
values for shuttle and conventional service, the base value for shuttle service is presented first, followed, in parenthesis, by
the base value for conventional service.
4.1.1. Trucking and unloading
Presented below are the variables including their meanings and base values used to determine trucking and
unloading time.
v truck
Truck speed
Farm to elevator distance
Truck unloading time
dist truck
timeunload
45 MPH
30 (15) miles
10 min
The trucking and unloading times make up a miniscule portion of the total time to move grain from farm to export elevator. However, we include this portion of the model for completeness. Berwick and Dooley (1997) use an average trucking
speed of 45 miles per hour in their trucking cost model. Berruto (2004) found that the average unloading time at a large
grain elevator ranged from 4 min to 7.4 min over 3 days in 1993. Since 1993, grain elevators have become significantly
more efficient at unloading, weighing, and testing grain. Vachal (2015) reports that the average trucking distance from farm
to market for wheat in 1984 was 12 miles. The average distance increased to around 32 miles in the early 2000s and
remained around
30 miles between 2000 and 2013.
Mathematical formulation of truck in-transit and unloading time:
Eq. (4.1) states the average time needed to truck grain from the farm to a country or terminal elevator and the time to
unload the grain at the elevator. The value is less than one hour for both conventional and shuttle service.
Timetruck
dist truck
v truck
timeunload < 1 h
4:1
4.1.2. Elevator storage and handling time

The amount of time that a ton of grain spends at an elevator depends on a number of factors including the demand rate
(a model input), the minimum amount of grain needed to fill a railcar order, and the handling rate at which grain is loaded
onto a train. Each of these factors are considered in the elevator time model component presented in this section. Longterm
2
Throughout the paper, we refer to a number of assumptions as conservative. The results presented in this paper suggest that shuttle train service is
superior to conventional service across all three performance dimensions. A conservative assumption implies that the assumption is making shuttle train
service look worse or conventional service look better than what a deeper analysis would likely suggest.
storage is implicitly captured in this model through the demand rate. A farmers decision to either sell grain to an elevator
or put it in long-term storage directly affects the demand rate for grain transportation.
Presented below are the variables including their meanings and base values used to determine elevator storage and
handling time.
Lmin
drate
timehandling
Elevator loading minimum

Demand rate of grain into elevator
Elevator handling time
12,100 (2640) tons

Input [tons/day]
13 (24) h
The loading minimum is the amount of grain that must be consolidated at an elevator before a train can pick-up and
transport the grain. Railroads typically incentivize terminal elevators to load shuttle trains as quickly as possible. For example, BNSF provides rate discounts if elevators load railcars in under 15 h (Burlington Northern Santa Fe, 2015). A value of 13
h is used because, according to conversations with 34 terminal elevator operators in North Dakota, grain shuttle loading
elevators typically load grain shuttle trains in significantly less than 15 h. In contrast, multi-car (630 railcars) trains are
typically loaded in one day (Tolliver and Bitzan, 2002).
As mentioned previously, the demand rate is the amount of grain produced in the geographical area surrounding an elevator per unit of time that is to be served via rail transport. This analysis assumes that the demand rate of a terminal
elevator is three times higher than the demand rate of a country elevator; i.e. assuming a region wherein only terminal or
country elevators are allowed to operate, not both, a terminal elevator would serve the same demand as three country
elevators. The 3:1 ratio used in this analysis is a conservative assumption. Since the introduction of shuttle trains and the
restructuring of the grain supply chain, the number of country elevators has decreased significantly due to the ability of
farmers to truck directly to terminal elevators as well as nearby end-users, including ethanol plants and livestock feed lots
(U.S. Department of Agriculture, 2014; Vachal, 2015).
Mathematical formulation of elevator storage and handling time:
Eq. (4.2) determines the amount of time a ton of grain spends at an elevator. In order for a fair comparison between
shuttle and conventional service, in terms of storage time, the analysis assumes that as soon as a terminal (country)
elevator consolidates enough grain to fill a shuttle (conventional) train, the train arrives and picks up the grain. This is
assumption is made in order to present a fair comparison between the two grain collection/transportation systems.
Timeelev ator
Lmin
drate
timehandling
4:2
The first term on the right-hand side of Eq. (4.2) represents the average time a ton of grain spends in storage; Lmin =drate is
the total time needed to accumulate enough grain at an elevator to fill a train. Assuming a uniform arrival rate of grain
into an elevator, dividing the total time needed to accumulate enough grain to fill an entire train by two gives the average
time a ton of grain spends in storage. The authors interviewed a number of grain industry stakeholders in the Upper
Midwest. One farmer stated that he typically waits until he knows a shuttle train is coming before selling his grain to, and
dropping it off at, an elevator. We asked three terminal elevator operators whether this was common practice. Each of the
three elevators operators suggested that farmers are not usually notified when shuttle trains are coming to pick-up
grain.
4.1.3. Rail transportation
The two models presented in this section determine the time to transport grain from a terminal (country) elevator to an
export elevator. This section presents two distinct rail transport time models because the factors influencing rail transportation time for shuttle and conventional service are different.
Shuttle Rail Transportation:
Presented below are the variables including their meanings and base values used to determine the time to transport
grain over the rail network via shuttle train service.
dist rail
v shuttle
f crew
delaycrew
Distance over the rail network

Shuttle line-haul train speed
Frequency of crew changes
Crew change delay
1400 miles
22 miles per hour (MPH)
1/7 changes/h
1 h/change
The 1400 miles value is approximately the distance from Fargo, North Dakota to the Ports in the Pacific Northwest
(PNW). A 2008 Bureau of Transportation Statistics report (U.S. Department of Transportation, 2008) shows that the quarterly
average line-haul speed between 1999 and 2008 ranged from 21 to 25 miles per hour. The STB now requires railroads to
report on
performance metrics every week (Railroad Performance Metrics, 2015). According to the self-reported performance metrics
of Canadian Northern (CN), Union Pacific (UP) and Burlington Northern Santa Fe (BNSF) the average line-haul speed for unit
grain trains over the past 52 weeks was approximately 22 miles per hour (Railroad Performance Metrics, 2015). Line-haul
speed is a measure of the over-the-rail train speed and does not include the time trains and/or railcars spend at
classification yards or at origination or termination points. According to a GAO report (Mead, 1992), railroad train crews
typically switch every 7 h. This paper assumes that the delay per crew change is 1 h this is an upper bound (i.e. a
conservative value) on the time required for a rail crew change.
Mathematical formulation of shuttle train service in-transit time:
Eq. (4.3a) determines the amount of time that a ton of grain spends on the railroad network when traveling via shuttle
train service. The first term on the right-hand side of the equation represents the in-transit time wherein the train is moving
over the rail network. The second term represents the lost time due to mandatory crew changes. Using the parameter values
listed above, Eq. (4.3a) returns a value 73 h. The results are independent of the demand rate.
Time shuttle
rail
dist rail
rail
dist
shuttle
v shuttle
f crew
delaycrew
4:3a
Conventional Rail Transportation:

Presented below are the variables including their meanings and base values used to determine the time to transport
grain over the rail network via conventional service.
dist rail

Conventional line-haul train speed
Distance between classification yards
Delay per classification yard
v conv en
dist yards
delayyard
1400 miles
21 MPH
250 miles/yard
30 h/yard
The distance between classification yards depends on a number of factors. Railcars traveling between two major rail hubs
usually do not have to travel through many classification yards; in contrast, railcars traveling between two low-density
points usually travel through many classification yards. The other key determinant is the number of railcars traveling
together (i.e. the shipment size); as the number of railcars traveling together increases the number of inter-train and
intra-train switches typically decreases (Tolliver and Bitzan, 2002). Tolliver and Bitzan (2002) state that the average distance
between classification yards for single car shipments is 200 miles. The value is typically larger for multi-car ( 24 railcar)
shipments; hence, a reasonable value of 250 miles is used in this analysis. Section 4.3 includes a sensitivity analysis on
the model results with respect to the average distance between classification yards.
According to the self-reported performance metrics of CN, UP, and BNSF, the average speed for conventional trains over
the past 52 weeks was approximately 21 miles per hour (Railroad Performance Metrics, 2015). Additionally, BNSF selfreported that the average time a railcar spent in the Pasco, WA classification yard over the past 52 weeks was 31.2 h. The
average time in the Northtown, MN classification yard was 34.7 h (Railroad Performance Metrics, 2015). These two
classification yards are located on BNSFs northern corridor. We use a value of 30 h, representing a lower bound (i.e. a
conservative parameter value), for the average dwell time of a conventional train at a classification yard.
Mathematical formulation of conventional train service in-transit time:
Eq. (4.3b) determines the amount of time that a ton of grain spends on the rail network when traveling via conventional
train service. The second term on the right-hand side represents the time that conventional train railcars spend in classification yards. Using the parameter values listed above, Eq. (4.3b) returns 235 h.
Timeconv en
rail
dist rail
dist rail dist
yards
vconv en
delayyard
4:3b
4.1.4. Total travel time Farm to export elevator

This section presents the abbreviated mathematical formulation used to calculate the total logistical supply chain time
between farm and export elevator. The parameter values listed above were used to calculate the time for each component
of the time model; the only remaining variable in these two equations is the demand rate.
Shuttle Service
Total Timeshuttle Timetruck Timeelev ator Timeshuttle

Total Timeshuttle 0:83 h
6050
drate
4:4a
rail
13 h 72:7 h
Conventional Service
Total Timeconv entional Timetruck Timeelev ator Timeconv en

Total Timeconv entional 0:50 h
1320
drate =3
rail
4:4b
24 h 234:7 h
The denominator in the elevator time component of the conventional service model is drate =3 because the demand
arriving at a single country elevator is one-third the demand arriving at a terminal elevator (i.e. one-third the total
demand).
4.2. Time model results
Figs. 2 and 4 display the results of the time model described in Section 4.1. Fig. 2 shows that shuttle service is faster
than conventional service when the rail transport demand rate is greater than 350 tons per day a relatively low rate.
As the demand rate increases to values greater than 2500 tons per day, shuttle train service transports grain nearly 50%
(12 days vs. 6 days) faster than conventional service. Fig. 3 displays the total time for a demand rate of 1500 tons per day
broken down
Fig. 2. The results of the time model using the parameter values presented in Section 4.1.
Fig. 3. Breakdown of time model results by segment of the logistical supply chain (demand rate = 1500 tons/day).
by trucking, elevator storage, and rail transport time. Shuttle trains are significantly faster at moderate-high demand rates
because they are able to bypass classification yards. The reason why conventional service is comparable/slightly faster at
low demand levels is that it takes a long time to consolidate enough grain at a terminal elevator to fill an entire 110-railcar
shuttle train; therefore, the average grain spends a long time in storage at a terminal elevator. The main takeaway from
the time model presented in Section 4 is that at moderate and high demand levels, the total time it takes for grain to travel
from the farm to the export elevator is significantly lower when it travels via shuttle train service.
4.2.1. Validation of time model results
In this sub-section, we attempt to validate some of the time model results based on actual data. The trucking and truck
unloading time results are inconsequential and likely accurate given that all the components come directly from
secondary sources, therefore no attempt is made to validate these results. The terminal elevator handling time for shuttle
trains is relatively fixed (the maximum time is 15 h); hence, further validation is not required. The country elevator
handling time value came from a secondary source (Tolliver and Bitzan, 2002). The average short-term storage at
elevators is subject to many factors, most importantly the overall demand for rail service and the availability of grain
railcars. This uncertainty is captured in the model via the demand rate that we vary between 50 and 3000 tons per day
in Fig. 2.
Unfortunately, origin to destination shipment travel time data for rail shipments is not something that railroads are
required to report (Transportation Research Board, 2015); therefore validating the model results was difficult for conventional and shuttle rail service. We were unable to obtain reliable data for conventional service rail travel times between
the Upper Midwest and the Pacific Northwest. We validated the shuttle service rail time model results via conversations
with representatives of large grain companies in the Upper Midwest including Archer Daniels Midland (ADM) and Cenex
Harvest States (CHS). Representatives from both companies stated that under normal operating conditions, the average shuttle train makes three round trips between the Upper Midwest and the Pacific Northwest in a month (i.e. approximately
5 days per one-way trip). The results in Fig. 3 suggest that the one-way rail transportation time is approximately 3.0 days;
however, this does not include the 13 h required to load the shuttle train at the terminal elevator and the time spent
unloading the grain from the shuttle train at the export terminal. Adding these two values to the 3.0 days obtained in
the shuttle train rail time model returns a value between 4 and 5 days. This result suggests that the shuttle train time
model results are reasonable.
4.3. Sensitivity analyses on time model results
The purpose of this section is to analyze how the results displayed in Fig. 2 would change if the parameter values listed
in Section 4.1 were different. The sensitivity analyses presented here reinforce the key result of Section 4.2: shuttle train
service is faster than conventional service at moderate and higher demand levels (i.e. the time model results are
relatively stable across a wide range of model parameter assumptions). Fig. 4 shows that the average distance between
classification yards significantly impacts the travel time of conventional service; however, shuttle service still outperforms
conventional service even if the average distance between classification yards parameter assumption is quite large. Fig. 5
shows a similar result for the classification yard dwell time parameter assumption.
5. Engineering cost model
The engineering cost model presented in this section aims to capture the variable logistical supply chain costs of moving
grain between the farm and the export elevator. Moreover, the models aim to capture the key variable engineering cost differences between the restructured grain collection and transportation system and the old system. The model excludes the
capital costs of trucks, country and terminal elevators, classification yards, and railroad infrastructure. Similar to the time
Fig. 4. Sensitivity analysis on the time model results with respect to the average distance between classification yards. We vary the average distance
between 200 miles and 400 miles. Conv is an abbreviation for conventional rail service.
Fig. 5. Sensitivity analysis on the time model results with respect to the average dwell time of railcars in a classification yard. We vary the dwell time
between 24 h and 36 h.
model, there are three cost model components including a trucking cost model, an elevator cost model, and a rail
transportation cost model. Note that engineering costs are different from accounting costs, intended for tax and financial
reporting, in that they refer to direct costs actually incurred in providing the service in question.
5.1. Cost model parameter values and mathematical formulation
The mathematical models for each of the three cost model components are presented below. The costs obtained in each
of the three components are additive. The engineering cost model parameter values in this section come mainly from
secondary sources.
5.1.1. Trucking costs
Presented below are the variables including their meanings and base values used to determine trucking costs.
Trucking cost per mile
Truck carrying capacity
Distance between farm and elevator
Length of analysis period
ratetruck
captruck
dist truck
drate
time
$1.4/mile
25 tons
30 (15) miles
Input [tons]
1 days
There is one major trucking cost difference between the restructured grain collection and transportation system and the
old system. The restructured system with terminal elevators and shuttle trains has longer trucking distances than the old
system (Vachal, 2015). This key difference is captured in the truck cost model presented in this section.
According to Berwick and Farooq (2003), the literature typically lists fuel costs, maintenance and repairs, tires, and labor
as the variable costs of trucking for farmers that haul their own grain to market. Table 5-1 provides estimates of the
trucking cost per mile from two sources Bangsund et al. (2011) and Fender and Pierce (2013). Fender and Pierce (2013)
estimate trucking costs for every year between 2008 and 2012 and for the five major geographical regions in the United
States in 2012. Fender and Pierces trucking cost estimates are for all truck transportation not just grain hauling.
Conversely, Bangsund et al. (2011) estimate the variable trucking cost for grain shipments in 2011. Table 5-1 shows that
the two highest cost components are fuel costs and labor costs. The $1.4 per mile parameter value is a conservative
estimate of the variable cost of transporting grain. To clarify, $1.4 per mile is not the cost of purchasing truck service from a
third party; it is the variable engineering cost of moving grain from farm to elevator. We assume that farmers truck their
own grain from farm to elevator.
The average trucking cost per mile obviously depends on the volume of grain being transported. However, as Barnes and
Langworthy (2003) lament, despite there being a wide variety of truck sizes and uses, it is most common for researchers to
Table 5.1
Breakdown of trucking cost per mile.
Variable costs
Fuel costs
Repair/maintenance
Tires
Labor wages
Labor benefits
Total
Bangsund et al. (2011)
Fender and Pierce (2013)
2011
2008
2009
2010
2011
2012
2012 Midwest
$0.51
$0.17
$0.17
$0.37
$0.63
$0.10
$0.03
$0.44
$0.14
$1.35
$0.41
$0.12
$0.03
$0.40
$0.13
$1.09
$0.49
$0.12
$0.04
$0.45
$0.16
$1.25
$0.59
$0.15
$0.04
$0.46
$0.15
$1.40
$0.64
$0.14
$0.02
$0.42
$0.12
$1.33
$0.64
$0.14
$0.04
$0.42
$0.11
$1.35
$1.23
estimate a single composite trucking cost value for all of them. However, from Vachal (2015) we do know that, over the past
few years in the Upper Midwest, most grain is hauled to market with 5-axle semi-trucks that have an average loaded
weight of approximately 25 tons. Hence, we use a truck carrying capacity of 25 tons in this paper.
As mentioned previously, Vachal (2015) reports that the average trucking distance from farm to market for wheat in 1984
was 12 miles. The average distance increased to around 32 miles in the early 2000s and remained around 30 miles between
2000 and 2013. Vachal (2015) also states that the average trucking distances for corn and soybeans in 2013 were around
30 miles.
The model results are independent of the length of the analysis period. This term is only necessary to have consistent
units when adding the trucking, elevator, and rail transportation costs.
Mathematical formulation of trucking costs:
Eq. (5.1) presents the mathematical formulation used to determine the cost of trucking grain from farm to export
elevator. The term on the right-hand side of Eq. (5.1) includes a two in order to account for the round-trip distance between
farm and elevator.
Costtruck
rate
ratetruck
time
captruck
disttruck
5:1
Using the assumed parameter values for trucking cost displayed above, Eq. (5.1) reduces to a function of the demand
rate for both conventional and shuttle service.
d
Conventional: Costtruck
drate
Shuttle: Cost truck
1 day
rate
25 tons
1 day
25 tons
$1:
4
mile
$1:
4
mile
$ 1: 68 day
15 miles
ton
drate .
$3: 36 day
30 miles
ton
drate .
5.1.2. Elevator storage and handling costs

Presented below are the variables including their meanings and base values used to determine terminal (country)
elevator storage and handling costs.
chand :
P grain :
Irate :
Lmin
drate
time
Handling cost
Price of grain
Annual interest rate
$0.76/ton ($0.66/ton)
$170.0/ton
6.0%/yeara
12,100 (2640) tons
Input [tons/day]
1 days
6% is the short-term annual interest rate. Source: Edwards, William. Cost of Storing Grain|Ag Decision Maker. Cost of Storing Grain|Ag Decision Maker.
Iowa State University: Extension and Outreach, 2013. Web. January 2015. <http://www.extension.iastate.edu/agdm/crops/html/a2-33.html>. The link to the
excel file on the right hand side of this page is the actual source.
Short-term storage and handling costs at country and terminal elevators are quite similar. The only differences include
higher inventory costs and overtime wage costs at terminal elevators. These two cost differences are captured in the
elevator cost model presented in this section. It would be reasonable to assume that the newer, more efficient terminal
elevators have lower per unit variable costs but unfortunately, we were unable to find any data or secondary sources that
directly support this assertion.
The handling cost value of $0.76 per ton for terminal elevators was obtained from a study by Kenkel et al. (2004). The
study reports an aggregate terminal elevator variable cost of $0.60 per ton in 2004 dollars ($0.76 per ton in 2015 dollars
CPI indexed). The $0.60 per ton variable cost is comprised of labor overtime ($0.041 per ton), grain inspection ($0.073
per ton), grain inspection overtime ($0.061 per ton), electricity ($0.33 per ton) and grain cleaning ($1.833 per ton).
However, Kenkel et al. (2004) assume that only 5% of the grain is cleaned which equates to an average grain cleaning cost of
$0.091 per ton. A later study by Kenkel (2008) analyzes the handling and storage costs of grain in country elevators but the
breakdown of costs is not consistent with Kenkel et al. (2004).
The only significant difference in the variable handling cost at country elevators and terminal elevators stems from the
fact that terminal elevators need to fill an entire 110-railcar train in under 15 h. In order to complete this task it takes a
team of 35 employees working non-stop 1015 h straight, depending on how efficient they are, to fill the entire shuttle
train. This task requires the elevator operator to either pay employees overtime or higher part-time employees just to
load a shuttle train. It is also necessary to account for the overtime pay given to the grain inspector(s) that work during the
loading of shuttle trains. According to Kenkel et al. (2004), the elevator employee and grain inspection overtime wages
amount to $450 per
train and $675 per train. This equates to an increase in handling cost between terminal and country elevators of $0.10 per
ton. Hence, the handling cost rate of country elevators in this analysis is set at $0.66 per ton.
As mentioned previously, the model assumes that as soon as enough grain is consolidated to fill either an entire shuttle
train or set of conventional train railcars, the grain exits the elevator. Hence, the cost of long-term storage is not accounted
for in the engineering cost model. However, the model includes the inventory cost associated with storing grain in the
short term. The market price of grain varies considerably based on the type of grain and general market fluctuations.
However, the market price of grain does not have a significant impact on the overall variable supply chain cost of grain
collection and transportation.
Mathematical formulation of elevator storage and handling costs:
Eq. (5.2) determines the storage and handling costs of grain at an elevator. The first term in parenthesis determines the
handling costs and the second term determines the inventory cost of the grain sitting in storage. The model assumes that
a terminal elevator needs to have the financing to store an entire shuttle train full of grain independent of the
production (arrival) rate of grain. The country elevator, which serves 24 railcar trains, has lower inventory costs because it
only needs to store 24 carloads full of grain. However, because we assume that one terminal elevator handles the same
volume of grain as three country elevators, the aggregate inventory costs are relatively close.
Costelev ator time
drate
chandle P grain
Lmin
Irate
5:2
Using the assumed parameter values for elevator costs displayed above, Eq. (5.2) reduces to a function of the demand
rate for both conventional and shuttle service.
Conventional: Cost elev ator
Shuttle: Cost elev ator
$0:66 day
ton
$0:76 day
ton
drate $221.
drate $338.
5.1.3. Rail transportation costs

The final component of the engineering cost model, relating to rail transport costs, is presented in this section. Presented
below are the variables including their meanings and base values used to determine shuttle (conventional) rail transportation costs.
timeyard :
wyard :
wtrain :
rfuel :
cfuel :
ccrew :
v rail
dist rail
dist yards
Irate :
Pgrain :
Timerail
time
Lmin
cloco :
crailcar :
nloco :
nrailcar :
tturn :
a
Delay per rail yard

Yard workers per train
Workers per train
Fuel Consumption Rate
Fuel Cost per Gallon
Labor Cost
Line-haul speed
Distance between classification yards
Annual interest rate
Price of grain
Rail transportation time
Annualized locomotive cost
Annualize hopper car cost
Locomotives per train
Railcars per train
Round-trip time for train
NA (30) h
NA (0.5) workers/train
2 (1.5) workers/train
470 ton-miles/gallon
$4.0/gallon
$40/h
22 (21) MPH
1400 miles
250 miles
6.0%/year
$170.0/ton
72.7 (234.7)a h
1 days
12,100 (2640) tons
$125,000/year
$3500/year
3 (1.5) locomotives/train
110 (24) railcar/train
10 (20) days
Values come directly from the sub-section entitled Rail Transportation from Section 4 in the time model.
There are four significant differences in the cost of rail transportation between shuttle and conventional service; the first
three cost differences stem from the fact that shuttle trains bypass classification yards. The first cost difference is that
shuttle trains do not incur switching costs at classification yards. The second difference is that the utilization of railcars
and locomotives on shuttle trains are significantly higher than on conventional trains because shuttle trains do not incur
delays at classification yards. Third, the inventory cost of moving grain via shuttle trains is lower over the rail network
because shuttle trains are faster than conventional trains. Last, because shuttle trains always have the optimal number of
locomotives and
railcars this allows them to spread out labor costs, locomotive costs, and railcar costs more efficiently. All four of these differences are captured in the model presented in this section. We also include fuel costs. We make the conservative assumption that fuel costs are the same for shuttle and conventional service even though they are probably lower for shuttle trains
because they do not stop at and require switching at classification yards.
Shuttle trains do not enter classification yards, by definition; therefore, the delay per classification yard for shuttle
trains and the number of classification yard workers per shuttle train are not applicable (NA). The time that a conventional
service railcar spends in a classification yard is highly variable and dependent on a number of factors. BNSF self-reported
that the average time a railcar spent in two classification yards along the northern corridor were 31.2 h and 34.7 h
(Railroad Performance Metrics, 2015). We use a value of 30 h, representing a lower bound (i.e. a conservative parameter
value), for the average dwell time of a conventional train at a classification yard.
In recent years, the Federal Railroad Administration (FRA) tried to implement regulations requiring all trains to have at
least two persons on each train. Two crewmembers per train is currently the unofficial standard for nearly all railroads (U.S.
Department of Transportation, 2014); hence, we assume that each 110-railcar shuttle train employs two crewmembers. We
also assume two crewmembers are on each conventional train. However, conventional service typically involves the multicar grain shipments (e.g. 24 railcars) being combined with other railcars, therefore we assume, on average, each set of 24
railcars requires 1.5 crewmembers. The fuel consumption rate of 470 ton-miles per gallon was obtained from CSXs website
(CSX, 2015). The $4.0 per gallon fuel cost rate comes from a 2015 U.S. Energy Information Administration report (U.S. Energy
Information Administration, 2015).
The Bureau of Labor Statistics reports that the mean hourly wage for locomotive engineers and railroad yardmasters are
$28.54 and S27.29 respectively (U.S. Department of Labor, 2016). However, the wage received by an employee is significantly
less than the cost incurred by an employer. The employer also pays social security/FICA taxes (6.2% of income), unemployment (6.2%), Medicare (1.45%), workmens compensation (a non-miniscule amount for railroad workers), and benefits (25%)
(Pirraglia, n.d.). Taking taxes and benefits into account increases the labor cost from $28.4 ($27.3) per hour to $39.6 ($37.9)
per hour. In the analysis, the labor cost rate is $40 per hour. The line-haul speed and the distance between rail yards are discussed in Section 4.1.3.
Tita and Hagerty (2014) report that the average heavy-haul locomotive is approximately $2.5 million. Baumgartner
(2001) reports that the average cost of a grain hopper is $50,000 in the year 2000, adjusting for inflation we assume a value
of $70,000 per grain hopper. Baumgartner (2001) reports that the economic lives of diesel locomotives and grain hopper cars
are 20 years. Using these purchase price and lifetime values, we assume an annualized cost of $125,000 for a locomotive and
$3500 for a railcar.
The turnaround time parameter is the time it takes a train to travel from the Upper Midwest to export destination in the
Pacific Northwest and then return to the Upper Midwest. According to ADM and CHS, grain shuttle trains typically make
three round trips between the Pacific Northwest and the Upper Midwest every month, or approximately 10 days per round
trip. A conservative estimate of the round trip time for conventional service railcars is 20 days. The time model presented in
Section 4 estimates a rail transport time of between nine and 10 days and the handling time at a country elevator is one
day (Tolliver and Bitzan, 2002).
It is typical for a grain shuttle train to include three locomotives. Similar to determining the average number of workers
per conventional train, it is difficult to determine exactly how many locomotives are used to transport 24 railcars on average.
We assume 1.5 locomotives are used to transport the 24 conventional service railcars on average. A sensitivity analysis on
the cost model results with respect to the average number of locomotives for 24 railcars is presented in Section 5.3.
Mathematical formulation of rail transportation costs:
Eq. (5.3a) determines the cost of railcars traveling through a classification yard, which only applies to conventional service railcars. The simplified model focuses on the labor costs incurred at classification yards. The first term in Eq. (5.3a) determines the number of trains that pass through classification yards based on the demand rate. The last term in Eq. (5.3a)
determines the number of classification yards a train passes through en route from the Upper Midwest to the Pacific Northwest. Using the parameter values listed above, the classification yard labor cost reduces to a function of the demand rate.
clabor yard
time
rate
Lmin
wyard
timeyard
ccrew
dist rail $ 1: 27
drate
day
distyards
ton
5:3a
Eq. (5.3b) determines the inventory cost incurred during the rail transportation portion of the supply chain. The total
travel time of shuttle service is significantly lower than conventional service; therefore, the inventory cost of shuttle
service is lower than conventional service. Inventory cost is relatively insignificant relative to other costs in the model
though.
$0:27 day
cinv entory P grain
drate
time
Irate
Timerail
ton
drate
5:3b
Eq. (5.3c) determines the fuel cost of rail transportation and is the same for conventional and shuttle service.
cfuel distrail
drate
time
1
r fuel
cfuel
$11:91 day
ton
drate
5:3c
Eq. (5.3d) determines the labor costs over the rail network excluding the labor costs at classification yards. The last term
in Eq. (5.3d) determines the number of conventional or shuttle trains based on the demand rate. The labor costs for shuttle
service are lower than the labor costs for conventional service because we assume that the ratio of laborers to railcars is
2:110 (1:55) for shuttle service and 1.5:24 (1:16) for conventional service.
claborrail dist rail
ccrew
1
wtrain
vrail
drate
time
Lmin
$ 1: 52 day
drate
ton
5:3d
Eq. (5.3e) determines the capital cost per ton for locomotives and railcars. Eq. (5.3e) is independent of the demand rate
because the utilization of railcars and locomotives does not depend on the demand of single terminal elevator or a few
country elevators; it is determined by total demand across the entire country. The capital cost (or rental cost) of
locomotives and railcars per ton of grain transported is significantly lower for shuttle service, relative to conventional
service, because shuttle trains travel significantly faster from origin to destination and back and therefore have higher
utilization rates.
cutil c loco nloco crailcar nrailcar
yr
365 days
tturn
$5:64
1
Lmin
ton
5:3e
Eq. (5.3f) is the summation of Eqs. (5.3a)(5.3e). It includes all of the costs in the rail transportation model.
Costrail claboryard cinv entory cfuel claborrail cutil
5:3f
5.1.4. Total logistical supply chain costs

Eq. (5.4) displays the abbreviated total supply chain engineering cost of transporting grain between farm and export
elevator.
Costdrate Costtruck Costelev ator Costrail
5:4
5.2. Cost model results

Fig. 6 illustrates that the aggregate variable costs per unit of grain transported from farm to export elevator via shuttle
service/terminal elevators are significantly lower than conventional service/country elevators. At demand levels greater than
800 tons per day, the aggregate variable cost per unit of grain transported via shuttle service is approximately $19 per ton
and under conventional service the cost is approximately $23 per ton. This 17% decrease in cost is quite significant in the
grain transportation and grain production industries.
In only two components of the cost model presented in Section 5.1 does conventional service have a cost advantage over
shuttle service; these two advantages are trucking costs and elevator inventory costs. In the model trucks have to travel
twice as far to access a terminal elevator (30 miles) compared with country elevators (15 miles); therefore, trucking costs
are twice as high under shuttle service. However, Fig. 7 below shows that trucking makes up a relatively small portion of
the total logistical supply chain costs compared with rail transport costs. Conventional services elevator inventory cost
advantage over shuttle service is also relatively small. Whereas one terminal elevator needs to finance 110 railcars worth
of grain, we assume it takes three country elevators to handle the same volume of grain as one terminal elevator; therefore,
three country elevators need to finance 72 railcars (24 3) worth of grain.
In contrast to conventional services two minor cost advantages, shuttle train service has a number of major cost advantages, including lower rail transport labor costs, no classification yard costs, and lower in-transit inventory costs. The summation of all of these cost advantages is displayed in Fig. 6; moreover, Fig. 7 shows that the rail transport portion of the
grain supply chain is where shuttle service has a significant advantage over conventional service in terms of costs. The
values in Fig. 7 represent the costs to transport 1500 tons of grain from a farm in North Dakota to an export elevator in
the Pacific
Fig. 6. Engineering cost model results using the base values presented in Section 5.1.
Fig. 7. Comparison of the logistical costs of trucking grain, storing grain, and transporting grain via shuttle train service and conventional service.
Northwest. Of the three grain supply chain components modeled in this paper, railroad transportation is the largest
contrib- utor to total costs and shuttle service has numerous per unit rail transportation cost advantages over conventional
service.
5.2.1. Validation of cost model results
In this section, we attempt to validate some of the cost model results. The trucking cost model parameter values come
directly from secondary sources in the literature; therefore, no further validation is required. The elevator parameter values
also come directly from the literature. In contrast, the railroad cost model includes five different components and the
parameter values in the rail cost model come from a number of different sources.
Fig. 7 displays the shuttle and conventional rail transport costs to ship 1500 tons of grain 1400 miles. Dividing the
shuttle and conventional cost by the shipment weight and shipment distance, the model estimates that the cost per ton-mile
(CPTM) for shuttle and conventional rail are 1.01 cents per ton-mile and 1.47 cents per ton-mile, respectively. In order to
validate the rail cost model results we used the Surface Transportation Boards (STB) 2013 Carload Waybill Sample (CWS),
which is a 1% stratified sample of all major carrier rail waybills originating or terminating in the United States. The CWS
waybill data is revenue data not cost data; therefore, it is not perfect for validation purposes. However, it can be used to
ensure that the cost estimates have the correct order of magnitude and are not too high. In order to ensure that the rail
cost model results are reasonable, the CPTM should be lower than the average revenue per ton-mile (RPTM); it is unlikely
that railroads would ship railcars at a loss. Moreover, the rail cost model presented in Section 5.1.3 does not include the
fixed costs nor does it capture all of the variable costs. Hence, we expect the cost models estimated rail CPTM to be
significantly lower than the CWSs RPTM.
We examined the 909 shuttle train (90 or more carloads) bulk grain shipments originating in the Upper Midwest in 2013
and found that the 0.25 quantile, mean, and 0.75 quantile RPTM waybills were 2.80 cents, 3.68 cents, and 3.41 cents,
respectively. The 0.25 quantile RPTM is significantly higher than the estimated CPTM value of 1.01 cents but the order of
magnitude of the two values is the same suggesting that the shuttle train cost model results are not completely wrong.
For the 411 multi-railcar (640 carloads) bulk grain shipments originating in the Upper Midwest in 2013 the 0.25 quantile, mean, and 0.75 quantile RPTM waybills were 3.35 cents, 5.61 cents, and 6.10 cents, respectively. Once again, the 0.25
quantile RPTM is significantly higher than the estimated CPTM of 1.47 but the order of magnitude of the two values are
the same suggesting that the conventional rail cost model results are not completely wrong.
Also of interest is the percent difference between shuttle service and conventional service in both the STBs average RPTM
and this papers estimated CPTM. The mean RPTM for multi-railcar shipments is 52% higher than the mean RPTM of shuttle
trains. Similarly, the estimated CPTM for conventional rail service is 46% higher than the estimated CPTM for shuttle trains.
5.3. Sensitivity analyses on cost model results
In order to capture how the key parameter values, listed in Section 5.1, affect the aggregate variable costs per unit of
grain transported via shuttle and conventional service, we performed sensitivity analyses. The sensitivity analyses
presented in this section illustrate that the cost model results are relatively stable across a wide range of parameter value
assumptions. Fig. 8 shows that only if trucking costs are assumed to be $5 per mile, a nearly 250% increase over the base
$1.4 per mile parameter assumption, does shuttle service become as costly as conventional service. Fig. 9 illustrates that
the parameter
Fig. 8. Sensitivity of the cost model results with respect to trucking costs. We vary the trucking cost between $2 per mile and $5 per mile.
Fig. 9. Sensitivity of the cost model results with respect to the average number of locomotives used to transport 24 railcars via conventional rail service.
We vary the number of locomotives between 1 and 2.
assumption on the average number of locomotives used to transport 24 railcars of grain via conventional service does
impact the results, just not significantly enough to make conventional service competitive with shuttle service.
6. Capacity model
Measuring the capacity of a rail network is complicated. Three broad dimensions can be considered to measure level of
service or capacity of a rail network: throughput, asset utilization and reliability. The first, throughput, measures the quantity of goods that the transportation network serves per unit of time. Asset utilization measures the percentage of time
assets are being utilized in the manner in which they were intended to be used. Reliability is a broad term that, in
transportation, generally refers to on-time performance. The authors understanding of the dynamics of the grain and
grain transportation industries in the Upper Midwest leads us to believe that throughput is the most relevant level of
service/capacity measure to examine. Two important facts lead us to believe that reliability is not as important as
throughput in terms of level of service: (1) the regions grain storage capacity is very high and (2) grain does not lose its
value quickly. Also, because the utilization of assets is highly correlated with throughput (e.g. given x number of assets, a
higher throughput necessitates higher utilization rates of the x assets), the results of our throughput model provide a
proxy for asset utilization.
Two main factors restrict capacity in rail networks: mainline rail capacity and rail classification yard capacity. Some
mainlines are single track, meaning that two trains moving in different directions both have to travel over the same rail
track to get to their respective destinations. In single-track operation, the width of the train speed distribution increases
with traffic density due to the increased likelihood of crossing paths with another train. A double-tracked rail line
allows trains to travel uninhibited by trains moving in the opposite direction. The speed distribution for double track
lines should remain narrow, as trains do not need to stop as frequently. The rail network capacity model presented below
assumes double track configurations.
Train movements on sections of the railroad network are controlled by rail classification yards. As discussed previously,
classification yards allow railroads to consolidate railcars, traveling in similar directions, in order to transport more railcars
per train and reduce costs. Classification yards also perform a number of other functions such as re-fueling and inspections.
It is possible to model network capacity using either the mainline or the classification yard as the bottleneck. Because
we assume a double track configuration, our model treats the classification yard as the network bottleneck. Moreover,
previous research (Assad, 1980) suggests that classification yards are the major source of delays in rail networks.
The classification yards are modeled using a simple queue-based simulation, where the rate of sorting, fueling, and
inspections is fixed. The classification yard hump acts as the server in the queuing model presented in this section. The
input (i.e. arrival rate of trains) to the classification yard depends on the total demand for grain rail transportation, the percentage of that demand served by shuttle trains and conventional service, and non-grain railcar demand.
Unlike the time model and engineering cost model presented in Sections 4 and 5, respectively, the capacity model
excludes the trucking and elevator storage components of the grain supply chain. Trucking capacity is nearly unlimited
because grain trucks travel on the road network, which is relatively uncongested in most parts of the Upper Midwest. While
trucking companies do face driver shortages, this issue can be overcome without capital investments. Moreover, most grain
producers have their own fleet of trucks that they use to haul their grain to market. Elevator storage capacity is excluded
because land and storage capacity is cheap in the Upper Midwest. There is currently a significant amount of long and
short-term storage capacity (Vachal, 2015) and if the need for more storage presented itself, the cost to increase capacity
would be relatively small compared with increasing capacity via capital investments in the rail network. Bunge (2014)
implies that rail network congestion is disproportionately affecting grain throughput in many parts of the United States.
6.1. Conceptual model and input parameters
Fig. 10 displays a pictorial representation of the classification yard queuing model. Conventional grain and non-grain
trains enter the classification yards. From the conventional trains, each individual railcar is processed and passes over the
hump before being attached to a departing train. As long as the queue does not exceed the capacity of the receiving tracks
at the classification yard, grain and non-grain shuttle trains simply bypass the classification yard without stopping.
The following assumptions and parameter values are used in the queueing model:
Service Rate (l) = 3.0 railcars per minute = 4320 railcars per day.
The capacity of the receiving tracks upstream of the hump in the queuing model is 5000 railcars.
Existing classification yards range widely in terms of receiving track capacity.
The model only includes one classification yard. This classification yard theoretically represents the most congested classification yard between the Upper Midwest and the Pacific Northwest.
The arrival rate of trains, k, is based on a Poisson process.
k f demand; %shuttle.
The simulation is run for approximately 3 days. The length of the simulation only effects the potential for queue
spillback.
Additionally, whereas, the time and cost models examine the geographical area around one terminal elevator (500
5000 tons per day), the capacity model examines a much broader geographical area (e.g. Western North Dakota). Once
again, because demand for grain rail transportation varies considerably depending on a number of exogenous factors, we
vary the demand rate in the analysis between 1 and 30 shuttle train equivalents (110 railcars) per day. We also vary the
percentage of grain demand served by shuttle between 0% and 75% in increments of 25%.
Because railcars carrying grain make up only a portion of the traffic on the rail transportation network, we assume a
fixed level of background traffic. The model assumes 45 shuttle train equivalents per day of non-grain rail transport. We
initially assume that 25% of the non-grain demand is served by shuttle service. We also present results assuming that 35% of
the non- grain demand is served by shuttle service. All of the conventional background railcars and grain railcars travel
through a single classification yard.
6.2. Rail network capacity model results
The queueing model presented in the previous section is simple, yet realistic, for the purpose of this analysis. A more
elab- orate bulk (batch) queueing model of classification yard operations with multiple commodities and destinations
(e.g.
Fig. 10. Conceptual model of a rail classification yard.
Mahmassani et al., 2007) may be used. However, the simplified model presented in this section provides the necessary
level of detail to illustrate the effects of shifting demand to shuttle train service. Both the input and the output
(displayed in Figs. 11a and 11b) of the model are easy to interpret. The input represents the total number of railcars
(conventional and shuttle, grain and background) attempting to pass through, or by, the classification yard. The demand
for grain transportation varies (along the x-axis); the demand for non-grain transportation remains constant. The
output represents the total number of railcars that are actually able to pass through, and by, the classification yard.
Figs. 11a and 11b display the advantages of shuttle service over conventional service in terms of network capacity and
throughput. Fig. 11a assumes that 25% of background traffic is served by shuttle service. Fig. 11b assumes 35% of background
traffic is served by shuttle service. As demand for grain rail transportation increases, the throughput of the network is constrained by the amount of demand served by conventional service; the more grain that is served by shuttle service, the
higher the potential throughput. Fig. 11a shows that after a certain demand breakpoint ( 5600 railcars per day) conventional service (red line 0% shuttle) is no longer able to increase throughput. In contrast, if grain demand is served by
50% conventional service and 50% shuttle service (green line), throughput is only partially constrained at approximately
6100 railcars per day. At demand levels higher than 6100 railcars per day, the 50% of grain demand served by conventional
service, along with conventional background railcars, exceed the capacity of the classification yard; however, the 50% of
grain served by shuttle service is still able to bypass the classification yard. Hence, throughput still increases with demand
for rail transport.
Looking at Fig. 11a, the 0% grain shuttle and 25% grain shuttle lines have two distinct breakpoints; whereas, the 50% and
75% grain shuttle lines have one distinct breakpoint. The first breakpoint for each of the four lines simply represents the
point at which the classification yard throughput is maximized (i.e. even as the rate of railcars entering the classification
yard increases, the rate of railcars exiting the classification yard cannot increase). For the 0% case, all of the grain railcars
need to pass through the classification yard; therefore, as demand increases past 5600 railcars per day, throughput is capped
at 5600 railcars per day. The second breakpoint, for the 0% and 25% cases, represents the demand rate at which the
receiving tracks overflow onto the mainline. Receiving track overflow onto the mainline restricts the flow of the mainline,
including shuttle trains. In the 0% case, the spillback prevents the grain and non-grain shuttle trains from bypassing the
classification yard. This is why the throughput decreases as demand increases beyond 6600 tons per day.
The results in Fig. 11b are very similar to the results seen in Fig. 11a. The main difference between the results in the two
figures is that the first and second breakpoints for every percent shuttle level occur at a higher throughput level in Fig. 11b.
Fig. 11a. Rail network capacity model results. 25% of the background traffic is served via shuttle train service.
Fig. 11b. Rail Network Capacity Model Results. 35% of the background traffic is served via shuttle train service.
The reason for this outcome is straightforward; in Fig. 11a 25% of the background traffic is served by shuttle service,
whereas, in Fig. 11b, 35% of the background traffic is served by shuttle service. Hence, the extra 10% of the background traffic
served by conventional service in Fig. 11a needs to travel through the classification yard. As described previously, the
classification yard causes throughput reductions after input demand exceeds the classification yards service rate.
Therefore, adding
10% of the background traffic to the classification yard naturally results in more congestion at lower levels of total demand
for rail service.
The model currently assumes unlimited mainline capacity downstream of the classification yard. A more realistic network model would include mainline capacity restrictions limiting the throughput of shuttle train service. However, this
limitation does not take away from the main result illustrated by the queueing model: shuttle train service provides significant capacity advantages compared with conventional service when demand is high. Including a mainline capacity restriction would only cap the potential advantage of switching railcars from conventional service to shuttle service.
Increased network capacity and throughput can potentially benefit railroads, elevator operators, and farmers. Higher
throughput, by definition, means more railcars traveling over the rail network. The more railcars moving through the network the more revenue opportunities for railroads. Elevator operators typically buy and sell grain at small margins;
therefore, increased throughput implies a potential for increased revenues and profits (U.S. Department of Agriculture,
2014). Moreover, increased throughput on the rail network allows farmers to sell more of their grain when market prices are
attrac- tive. Lastly, not only do large-volume shuttle train shippers benefit from the network capacity improvements
associated with switching railcars to shuttle service, small-volume conventional service shippers of grain and other
commodities also benefit from the reduced classification yard and mainline congestion.
7. Conclusions and future work

The main grain producing regions in the United States have undergone significant operational restructuring which has
implications for the nations economy as well as its competitiveness on the global grain market. In grain producing regions,
in terms of freight transportation, this restructuring is marked by the noticeable shift from conventional rail service/country
elevators to shuttle train service/terminal elevators. The stark differences between the two systems, and the limited effort
invested in the literature to quantify the operational (time, cost and throughput) differences and their broader implications
across sectors of the economy, motivated the work done in this paper.
Based on transport theory and the main pillars of freight and logistics operations, this paper presents three models of
grain transportation to compare conventional service/country elevators against shuttle service/terminal elevators. The first
model measures the time it takes to move grain from farm to export elevator. The second model determines the aggregate
variable cost of moving grain from the farm to an export elevator. The third model determines rail network throughput as a
function of the demand for rail service and the percentage of railcars that are traveling via shuttle service and conventional
service. Together, the three models allow for a comprehensive comparison of the grain supply chain before and after
the introduction of shuttle trains and terminal elevators in the Upper Midwest.
The results of each model presented in this paper show that the restructured grain supply chain offers meaningful advantages relative to the old grain supply chain under most realistic demand scenarios. These results take into account the time
and costs of storing and transporting grain upstream of the rail transportation network. This implies that restructuring the
grain supply chain in major grain production regions is beneficial for grain logistics operations and the grain sector overall.
Reliability, which is closely related to both time and capacity, and speed (i.e. travel time) are important factors in contracts between grain exporters and their international customers. The results in this paper show that the restructuring of
the grain supply chain significantly improved the speed, capacity, and reliability of the rail network. Going-forward, the ability of grain exporters from the United States to offer faster, more reliable transportation service than competitors,
especially in the context of a highly competitive global grain market, is extremely valuable. In addition to reliability and
speed advantages, the restructured grain supply chain brought grain exporters from the United States significant
transportation and logistics costs advantages. Logistics costs are a major component of the price of grain along with
production costs. Because production costs in the United States are higher than those in South American countries the
United States global competitiveness strongly depends on its ability to keep logistics costs low.
Although the United States currently has grain transportation speed, reliability, and cost advantages, investments in
infrastructure in competing countries could weaken the competitive advantage of the United States if the United States
does not continue to invest, smartly, in infrastructure and supply chain improvements. Identifying these smart
investments, such as terminal elevators and shuttle elevators, is critically important. Moving forward, grain industry
stakeholders need to take a systems-level and forward-looking approach in their evaluation of transportation investments
and their impact on the marketing of grain. Although large upfront investments in grain storage and transportation
infrastructure, such as terminal elevators, are not easily palatable, they can pay off in the long-run as they drive down
operational costs and increase access to a larger quantity and more diverse set of markets.
Additionally, grain stakeholders have to monitor how various factors will play out, all of which have implications for the
grain production and logistics sectors. For example, as demand for food across the world continues to grow and
technological advances increase crop yield volumes, more grain will need to be transported to more places. Another
important consider- ation for future planning is the increase (decrease) of rail network traffic from competing
commodities such as oil (coal).
Other potential impacts on the cost, time, and capacity of the grain supply chain include autonomous trucks, positive train
control, and other technological advances. Future research into the impact of these technologies on the entire grain supply
chain is necessary to properly plan for their eventual arrival.
7.1. Model extensions
The three models in this paper are built on flexible frameworks that may be scaled up in terms of scope and/or complexity. Including additional variable costs and time components to the first two models is relatively straightforward. Similarly,
in the capacity model, the classification yard service and/or arrival rates may be readily adjusted. However, while the capacity model effectively illustrates the benefits of shuttle service over conventional service, added complexity would allow endusers to compare different operational scenarios more effectively and completely. Potential future extensions of the capacity
model include adding capacity restrictions on the mainline and introducing a non-uniform demand rate. The non-uniform
demand rate would allow the model to analyze the resiliency of the network as a function of the amount of demand served
by conventional and shuttle service.
Acknowledgements
The work presented here is based on work conducted at the Northwestern University Transportation Center, with partial
funding through the Centers unrestricted gift account. The authors are grateful to the contribution of Dr. Andreas Frei in the
early part of the study, and to Breton Johnson, NUTC Associate Director, for facilitating grain and rail industry contacts. The
authors remain solely responsible for the content of this paper.
References
Association of American Railroads, 2015. Railroads and Grain. <https://www.aar.org/BackgroundPapers/Railroads%20and%20Grain.pd f> (Accessed on April
5, 2016).
Assad, A.A., 1980. Models for rail transportation. Transport. Res. Part A: Gen. 14 (3), 205220.
Bangsund, D.A., Leistritz, F.L., 2005. Economic Contribution of the Wheat Industry to North Dakota. Department of Agribusiness and Applied Economics,
Agricultural Experiment Station, North Dakota State University.
Bangsund, D.A., Olson, F.E., Leistritz, F.L., 2011. Economic Contribution of the Soybean Industry to the North Dakota Economy. Department of Agricultural
Economics, Agricultural Experiment Station, North Dakota State University.
Barnes, G., Langworthy, P., 2003. The Per-Mile Costs of Operating Automobiles and Trucks. Retrieved from the University of Minnesota Digital Conservancy.
<http://purl.umn.edu/909>.
Baumgartner, J.P., 2001. Prices and Costs in the Railway Sector. Institut des transports et de planification.
Berruto, R., 2004. Logistics and Strategies Evaluation of Grain Receiving Operations Using the System Simulation Approach (Doctoral Dissertation, Purdue
University). <http://docs.lib.purdue.edu/dissertations/AAI3154589 /> (Accessed on April 5, 2016).
Berwick, M.D., Dooley, F., 1997. Truck Costs for Owner/Operators (No. MPC Report No. 97-81). Upper Great Plains Transportation Institute, North Dakota
State University.
Berwick, M., Farooq, M., 2003. Truck Costing Model for Transportation Managers (No. MPC Report No. 03-152). Mountain-Plains Consortium. Upper Great
Plains Transportation Institute.
Bunge, J., 2014. Pain in Trains Falls Mainly on Grain. Wall Street J. <http://wwwwsj.com/articles/pain-in-trains-falls-mainly-on-grain-1414972402>
(Accessed on January 28, 2016).
Burlington Northern Santa Fe, 2015. Agricultural Products Unit Train Facility Design Guidelines. <http://www.bnsf.com/customers/pdf/Guidelines-Shuttle.
pdf> (Accessed on January 28, 2016).
CSX, 2015. Fuel Efficiency CSX.com. <https://www.csx.com/index.cfm/about-us/the-csx-advantage/fuel-efficiency /> (Accessed on January 28, 2016).
Fender, K.J., Pierce, D.A., 2013. An Analysis of the Operational Costs of Trucking: A 2013 Update. Arlington. <http://truckexec.typepad.com/files/atrioperational-costs-of-trucking-2013-final.pdf > (Accessed on January 28, 2016).
Frittelli, J., 2005. Grain Transport: Modal Trends and Infrastructure Implications. Congressional Information Service, Library of Congress.
Griffin, G.C., Wilson, W.W., Casavant, K., 1984. Characteristics and Costs of Operation of North Dakotas Farm Trucks, vol. 51. Upper Great Plains
Transportation Institute, North Dakota State University.
Keaton, M.H., 1991. Service-cost tradeoffs for carload freight traffic in the US rail industry. Transport. Res. Part A: Gen. 25 (6), 363374.
Kenkel, P., 2008. Grain Handling and Storage Costs in Country Elevators. The Cooperative Accountant, Winter.
Kenkel, P., Henneberry, S., Agustini, H., 2004. An Economic Analysis of Unit-Train Facility Investment. In: Southern Agricultural Economics Association
Annual Meeting, Tulsa, Oklahoma.
Mahmassani, H., Zhang, K., Dong, J., Lu, C.C., Arcot, V., Miller-Hooks, E., 2007. Dynamic network simulation-assignment platform for multiproduct
intermodal freight transportation analysis. Transport. Res. Rec.: J. Transport. Res. Board 2032, 916.
Mead, K.M., 1992. Railroad Safety: Engineer Work Shift Length and Schedule Variability. U.S. General Accounting Office, Retrieved from <http://
www.gao.gov/products/RCED-92-133> (Accessed on January 28, 2016).
Moore, A.M., 2013. Estimating Freight Costs Over A Multi-Modal Network: An Auto Industry Supply Chain Example (Doctoral dissertation, Georgia Institute
of Technology), pp. 514.
Ndembe, E., 2015. Hard red spring wheat marketing: effects of increased shuttle train movements on railroad pricing in the northern plains. J. Transport.
Res. Forum 54 (2), 101.
Pirraglia, W. (n.d.). What Costs Do Employers Have to Pay When Paying Employee Salaries? Small Business. <http://smallbusiness.chron.com/costsemployers-pay-paying-employee-salaries-12569.html > (Accessed on January 28, 2016).
Prater, M., Sparger, A., ONeil Jr., D., 2013. The Effects of Increased Shuttle-Train Movements of Grain and Oilseeds (No. 155451). U.S. Department of
Agriculture. http://dx.doi.org/10.9752/TS088.08-2013.
Railroad Performance Metrics, 2015. Railroad Performance Metrics. <http://www.railroadpm.org/> (Accessed on January 28, 2016).
Sparger, A., Prater, M., 2012. A comprehensive rail rate index for grain. J. Transport. Res. 51 (2).
Swan, P., Tyworth, J., 2001. Shipper sensitivity to unreliable service in carload markets. Transport. J. 40 (3), 1625.
Tita, B., Hagerty, J., 2014. Caterpillar falls behind ge in locomotives race. Wall Street J. <http://wwwwsj.com/articles/caterpillar-falls-behind-ge-inlocomotives-race-1405291739> (Accessed on April 5, 2016).
Tolliver, D., Berwick, M., Vachal, K., 2005. Farm-to-Market Transportation Patterns and Truck Use in the Northern Plains, vol. 167. Upper Great Plains
Transportation Institute, Fargo.
Tolliver, D., Bitzan, J., 2002. Analysis of Revenues and Costs for Wheat Shipments Originated in North Dakota on the BNSF Railroad (No. 231806). North
Dakota State University, Upper Great Plains Transportation Institute.
Tolliver, D., Dooley, F., Zink, D., 1987. Short line operation of light density rail networks: economics and public policy. J. Transport. Res. Forum 28 (1).
Tolliver, D., Dybing, A., Mitra, S., 2006. Trip generation rates for large elevators: a North Dakota case study. Transport. Res. Rec.: J. Transport. Res. Board
1966,
8895.
Transportation Research Board, 2015. Special Report 318: Modernizing Freight Rail Regulation. <http://onlinepubs.trb.org/onlinepubs/sr/sr318.pd f>
(Accessed on April 5, 2015).
U.S. Department of Agriculture, 2014. Agricultural Marketing Service. Grain Transportation Report. <http://dx.doi.org/10.9752/TS056.12-25-2014>
(Accessed on January 28, 2016).
U.S. Department of Labor, 2016. Bureau of Labor Statistics. Locomotive Engineers. <http://www.bls.gov/oes/current/oes534011.htm> (Accessed on January
28, 2016).
U.S. Department of Transportation, 2008. Bureau of Transportation Statistics. Transportation Statistics Annual Report. Washington D.C. <http://www.
rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/tr ansportation_statistics_annual_report/2008/pdf/entire.pd f> (Accessed on January 28, 2016).
U.S. Department of Transportation, 2014. Federal Railroad Administration. FRA to Issue Proposed Rule on Minimum Train Crew Size. <https://www.
fra.dot.gov/eLib/Details/L04999 > (Accessed on January 28, 2016).
U.S. Energy Information Administration, 2015. Gasoline and Diesel Fuel Update. <http://www.eia.gov/petroleum/gasdiesel /> (Accessed on January 28,
2016).
Vachal, K., 2015. Northern Plains Grain Farm Truck Fleet & Marketing Patterns, vol. 284. Upper Great Plains Transportation Institute, North Dakota State
University.
Vachal, K.J., Bitzan, J.D., Baldwin, B., 1997. Implications of a North American Grain Marketing System for Prairie Transportation & Elevators (No. MPC Report
No 97-84). Upper Great Plains Transportation Institute, North Dakota State University.
Vachal, K., Bitzan, J., 2002. Long-term availability of railroad services for US agriculture. Transport. Res. Rec.: J. Transport. Res. Board 1790, 6672.

Analytical Models of Rail - Transportation

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Analytical Models of Rail - Transportation

Uploaded by

Copyright:

Available Formats

Transportation Research Part E 93 (2016) 294315

Contents lists available at ScienceDirect

Transportation Research Part E

Analytical models of rail transportation service in the grain

, Lama Bou Mjahed

E-mail addresses: michaelhyland2013@u.northwestern.edu (M.F. Hyland), masmah@northwestern.edu (H.S. Mahmassani), lamabm@u.nortwestern.edu

M.F. Hyland et al. / Transportation Research Part E 93 (2016) 294315

25; 000; 000 bush

25; 000; 000 bush

bushel of grain = 55 pounds = 0.0275 tons.

4.1.2. Elevator storage and handling time

Elevator loading minimum

12,100 (2640) tons

Distance over the rail network

Conventional Rail Transportation:

Distance over the rail network

4.1.4. Total travel time Farm to export elevator

Total Timeshuttle Timetruck Timeelev ator Timeshuttle

Total Timeconv entional Timetruck Timeelev ator Timeconv en

Bangsund et al. (2011)

Fender and Pierce (2013)

5.1.2. Elevator storage and handling costs

Costelev ator time

5.1.3. Rail transportation costs

Delay per rail yard

claborrail dist rail

cutil c loco nloco crailcar nrailcar

Costrail claboryard cinv entory cfuel claborrail cutil

5.1.4. Total logistical supply chain costs

Costdrate Costtruck Costelev ator Costrail

5.2. Cost model results

Fig. 10. Conceptual model of a rail classification yard.

7. Conclusions and future work

You might also like