Content uploaded by Pan Lu
Author content
All content in this area was uploaded by Pan Lu on Jul 10, 2020
Content may be subject to copyright.
MPC Report No. 13-250
Analysis of Railroad Energy Efficiency in the United States
Denver Tolliver
Pan Lu
Douglas Benson
Upper Great Plains Transportation Institute
North Dakota State University
Fargo, North Dakota
May 2013
Disclaimer
The contents of this report reflect the work of the authors, who are responsible for the facts and the accuracy of
the information presented. This document is disseminated under the sponsorship of the Mountain-Plains
Consortium in the interest of information exchange. The U.S. Government assumes no liability for the contents or
use thereof.
North Dakota State University does not discriminate on the basis of age, color, disability, gender expression/identity, genetic information, marital status,
national origin, public assistance status, sex, sexual orientation, status as a U.S. veteran., race or religion. Direct inquiries to the Vice President for Equity,
Diversity and Global Outreach, 205 Old Main, (701) 231-7708.
CONTENTS
1. INTRODUCTION ............................................................................................................................................... 1
2. BRIEF REVIEW OF RAILWAY ENERGY STUDIES .................................................................................. 3
3. ANALYTICAL MODELS OF RAILROAD FUEL CONSUMPTION .......................................................... 5
3.1 Resistance Forces and Equations .............................................................................................................. 5
3.2 Estimating Procedures ............................................................................................................................... 6
3.2.1 Car and Locomotive Resistance Factors ......................................................................................... 6
3.2.2 Locomotive Requirements ............................................................................................................... 7
3.2.3 Total Train Resistance ..................................................................................................................... 8
3.2.4 Horsepower-Hours and Gallons of Fuel Consumed ........................................................................ 9
3.3 Uniform Train Simulations ....................................................................................................................... 9
3.3.1 Average Train Speeds .................................................................................................................... 10
3.3.2 Train Speed Profile ........................................................................................................................ 10
3.3.3 Fuel Consumed During Acceleration ............................................................................................ 11
3.3.4 Origin-Destination Switching ........................................................................................................ 12
3.3.5 Average Trip Distance and Cycle Length ..................................................................................... 13
3.3.6 Estimated Speeds and Idle Time ................................................................................................... 14
3.3.7 Fuel Consumed During Drayage ................................................................................................... 14
3.3.8 Grade Profile ................................................................................................................................. 15
3.3.9 Predicted Results ........................................................................................................................... 16
3.4 Summary and Limitations of Analytical Method .................................................................................... 17
4. STATISTICAL MODEL OF RAILROAD FUEL CONSUMPTION .......................................................... 19
4.1 Model Formulation ................................................................................................................................. 19
4.1.1 Main Explanatory Variables .......................................................................................................... 19
4.1.2 Model Statement ............................................................................................................................ 20
4.1.3 Time Variable ................................................................................................................................
21
4.1.4 Regional Variables ........................................................................................................................ 21
4.2 Regression Results .................................................................................................................................. 22
4.2.1 Key Model Properties .................................................................................................................... 22
4.2.2 Parameter Estimates and Standard Errors...................................................................................... 22
4.2.3 Probability Values and Inferences ................................................................................................. 23
4.2.4 Test for Serial Correlation ............................................................................................................. 24
4.3 Model Predictions ................................................................................................................................... 24
4.3.1 Level of Precision .......................................................................................................................... 24
4.3.2 Marginal Estimates Derived from Coefficients ............................................................................. 25
4.3.3 Average Fuel Efficiency Factors ................................................................................................... 25
4.4 Comparisons of Predictions .................................................................................................................... 27
4.4.1 Grain Shipments ............................................................................................................................ 27
4.4.2 Auto Shipments ............................................................................................................................. 27
4.4.3 Coal Shipments .............................................................................................................................. 28
4.4.4 General Inferences ......................................................................................................................... 28
5. NATIONAL ANALYSIS OF RAILROAD FUEL EFFICIENCY ................................................................. 29
5.1 Overview of Waybill Sample .................................................................................................................. 29
5.2 Empty/Loaded Mile Ratios ..................................................................................................................... 29
5.3 Estimation Procedure .............................................................................................................................. 30
5.4 RTM/G Estimates for Major Commodity Movements ........................................................................... 30
5.5 Validation of Modeling Process .............................................................................................................. 31
6. TRUCK FUEL EFFICIENCY MODEL ......................................................................................................... 33
6.1 Variations in Fuel Economy with Speed ................................................................................................. 34
6.2 Empty Truck Miles ................................................................................................................................. 34
6.3 Fuel Attributable to Empty Truck Miles ................................................................................................. 35
7. WATERWAY FUEL EFFICIENCY MODEL ............................................................................................... 37
7.1 Revenue Ton-Miles per Gallon by River System ................................................................................... 37
7.2 Illustrative Waterway Movements .......................................................................................................... 38
8. COMPARISONS OF MODAL ENERGY EFFICIENCY ............................................................................ 39
8.1 Grain ....................................................................................................................................................... 39
8.1.1 Truck-Rail Comparisons................................................................................................................ 39
8.1.2 Rail-River Comparisons ................................................................................................................ 40
8.2 Flour ........................................................................................................................................................ 41
8.3 Ethanol .................................................................................................................................................... 41
8.4 Iron Ore ................................................................................................................................................... 42
8.4.1 Car Weights and RTM/G Values ................................................................................................... 42
8.4.2 Rail-Truck Comparison ................................................................................................................. 42
8.5 Soda Ash ................................................................................................................................................. 42
8.6 Coal ......................................................................................................................................................... 43
8.6.1 Car Weights ................................................................................................................................... 43
8.6.2 Rail-Truck Comparisons................................................................................................................ 44
8.6.3 Rail-Waterway Comparison .......................................................................................................... 45
9. TRANSPORTATION CIRCUITY .................................................................................................................. 47
9.1 Network Circuity ..................................................................................................................................... 47
9.2 Route Circuity ......................................................................................................................................... 47
9.3 Total Circuity .......................................................................................................................................... 48
9.4 Railroad-Truck Comparisons Adjusted for Circuity ............................................................................... 48
9.5 Effects of Circuity on Railroad and Barge Energy Comparisons ........................................................... 48
10. CONCLUSION ................................................................................................................................................ 49
10.1 Issues with Railway-Waterway Comparisons ........................................................................................ 49
10.2 Additional Issues .................................................................................................................................... 50
10.3 Additional Modeling Needs .................................................................................................................... 50
REFERENCES ....................................................................................................................................................... 51
APPENDIX ............................................................................................................................................................. 53
LIST OF TABLES
Table 2.1 Fuel Consumption Factors Used in State Rail Planning (1978) ............................................................. 3
Table 2.2 Simulated Fuel Efficiency of Unit Grain Trains Movements from Iowa ............................................... 3
Table 2.3 Range of Movement Factors and Fuel Efficiencies of Double-Stack Container Movements ................ 4
Table 3.1 Typical Values of C and Frontal Areas for Freight Equipment .............................................................. 6
Table 3.2 Example of Train Resistance Factors for Lead Locomotive and Gondola Car ...................................... 7
Table 3.3 Illustrations of Locomotive Tonnage Ratings and Power Requirements ................................................ 8
Table 3.4 Equipment Types and Weights Used in Uniform Train Simulations ..................................................... 9
Table 3.5 Cruising Speeds for Unit Trains on Flat Terrain Based on Locomotive Requirements ....................... 12
Table 3.6 Estimated Fuel Consumption at Origin (O) and Destination (D) for Unit Trains................................. 13
Table 3.7 Average and Train Cycle Distances ...................................................................................................... 13
Table 3.8 Predicted and Observed Train Speeds in Miles per Hour in Flat Terrain ............................................. 14
Table 3.9 Calculation of Drayage Fuel for Automobile and Container Trains ..................................................... 15
Table 3.10 Predicted RTM/G of Uniform Train Movements ................................................................................. 16
Table 4.1 Class I Railroads and Geographic Regions in the United States .......................................................... 19
Table 4.2 Variables in Railroad Fuel Regression Model ...................................................................................... 21
Table 4.3 GTMC/G and RTM/G by Region in 2008 ............................................................................................ 22
Table 4.4 Key Model Properties and Indicators.................................................................................................... 22
Table 4.5 Parameter Estimates from Railroad Fuel Model ................................................................................... 23
Table 4.6 Results of Test for Serial or Autocorrelation ........................................................................................ 24
Table 4.7 Predicted versus Observed Gallons of Fuel in 2008 ............................................................................. 24
Table 4.8 Marginal Estimates of Revenue Ton-Miles per Gallon, by Train Type................................................ 25
Table 4.9 Predicted and Observed GTMC/G Values by Train Type and Region ................................................. 26
Table 4.10 Estimated RTM/G in 2008, by Train Type and Region ........................................................................ 26
Table 5.1 Empty/Loaded Car-Mile Ratios by Region .......................................................................................... 30
Table 5.2 Weighted-Average RTM/G for Major Commodity Groups ................................................................. 31
Table 6.1 Average Miles per Gallon by Truck Configuration and Weight ........................................................... 34
Table 7.1 Waterway Fuel Efficiency Factors for Specific River System ............................................................. 37
Table 7.2 Average Revenue Ton-Miles per Gallon for Barge Shipments from Upper Mississippi
River Stations to New Orleans .............................................................................................................. 38
Table 8.1 Grain Car Weights and Net/Gross Ratios of Grain Unit Train Shipments ........................................... 39
Table 8.2 Weights and Capacities of Combination Trucks ................................................................................... 40
Table 8.3 Relative Fuel Efficiency of Railroads to Trucks in Transporting Grain ............................................... 40
Table 8.4 RTM/G Estimates for Grain Shipments Used in Rail-Waterway Comparisons ................................... 41
Table 8.5 Car Weights and Fuel Efficiencies of Iron Ore Movements ................................................................. 42
Table 8.6 Distribution of Coal Mined in the United States, by Destination Mode (2009) ................................... 43
Table 8.7 Coal Car Weights and Net/Gross Ratios by Car Type and Shipment Type .......................................... 44
Table 8.8 RTM/G Estimates for Coal Shipments Used in Rail-Waterway Comparisons ..................................... 44
Table 10.1 Percentages of Tons Transported by Barge on Inland River Systems in 2008 ..................................... 49
LIST OF FIGURES
Figure 1.1 Trend in Railroad Fuel Efficiency in the United States .......................................................................... 2
Figure 3.1 Variations in Resistance with Velocity for Select Car Movements ........................................................ 7
Figure 3.2 Average Train Speeds: Jan. 2009–May 2011 ....................................................................................... 10
Figure 3.3 Profile of Railroad Speed Limits in the United States .......................................................................... 11
Figure 3.4 Acceleration Curve for Unit Grain Train .............................................................................................. 12
Figure 3.5 Elevation Profile in the Pacific Northwest Region of the United States ............................................... 15
Figure 4.1 Plot of Gallons of Fuel versus Gross Ton-Miles of Cars and Contents ................................................ 20
Figure 6.1 Combination Truck Fuel Efficiencies ................................................................................................... 33
i
OVERVIEW
The energy efficiency of freight transportation is a critical issue in the United States in light of the price volatility
of fuels and America’s dependence upon foreign sources of petroleum. Moreover, energy efficiency is important
to environmental policy. The use of fossil fuels for transportation purposes results in the emission of air pollutants
such as nitrogen oxides, particulate matters, and volatile organic compounds. Because emissions increase with
gallons of fuel consumed, energy efficiency is a necessary condition for improved environmental quality.
The U.S. Department of Transportation (USDOT) has identified Environmental Sustainability as a major
goal. The goal calls for improving energy efficiency and reducing greenhouse gas emissions. Recently adopted
benefit/cost guidelines call for the quantification of “expected reductions in emissions of CO2 or fuel
consumption” when transportation projects are evaluated (USDOT, 2010, p. 21700). According to USDOT, a
“projected decrease in the movement of people or goods by less energy-efficient vehicles or systems will be given
priority” (Ibid).
While the sustainability goals are clear, techniques for quantifying energy comparisons are lacking. For want
of better information, system-average modal efficiency factors are often used. At the national level, a great deal of
effort has been devoted to emissions modeling. As a result, models such Mobile6 have emerged and are being
widely used. However, much less attention has been paid to models for quantifying differences in fuel
consumption. Within the multimodal investment framework articulated by USDOT, conservation of fuel is a
critical factor in assessing intercity freight options.
Research Objectives. The purpose of this study is to provide information about railroad fuel efficiency that may
be useful in evaluating transportation energy policies and assessing the sustainability of potential projects. The
specific objectives are to (1) develop railroad energy efficiency models that describe differences in fuel economy
among classes of trains and commodities; (2) apply these models to a wide range of movements to estimate fuel
efficiency ratings for coal, grain, iron ore, food products, and other key commodities; (3) develop comparable
procedures for estimating truck and waterway fuel consumption; and (4) compare rail, truck, and waterway
energy efficiencies. The focus on railroads in this study is appropriate, because many of the alternatives to
highway investment involve railroad transportation or multimodal options.
Energy Efficiency Measures. The two indicators of energy efficiency used in this study are (1) the gross
ton-miles of cars and contents produced with a gallon of fuel (GTMC/G), and (2) revenue ton-miles per gallon
(RTM/G). Gross ton-miles of cars and contents reflect the weights of containers, trailers, freight cars, and cargo.
Only the cargo weight is reflected in revenue ton-miles.
Observed Efficiency Ratings. In 2008, railroads achieved efficiency ratings of 806 GTMC/G and 457 RTM/G.
However, these ratings varied significantly among regions. The observed GTMC/G values were 779, 908, and
809 in the eastern, central, and western regions, respectively. The observed RTM/G values were 431, 501, and
464 in the same regions. These variations reflect differences in terrain, geography, and networks. Railroads in the
central or plains region do not cross mountain ranges. In contrast, western railroads encounter substantial grades
while crossing the Rocky Mountains and coastal ranges. Similarly, eastern railroads operate in the Appalachian
Mountains.
Trends in Energy Efficiency. GTMC/G increased by 45% from 1985 to 2008. In comparison, RTM/G increased
by 61% during the same period which is equivalent to a 2.6% annual growth rate. The growth in GTMC/G
implies that railroad fuel savings are due in part to technological and operational efficiencies that have enabled the
movement of a given weight with fewer gallons of fuel. However, the higher rate of growth in RTM/G indicates
that additional energy gains are attributable to moving more revenue tons in a single car—i.e., increasing the net
to gross weight ratio.
ii
Methods of Estimating RTM/G. Two methods of estimating railroad energy efficiency are considered in this
study (1) an analytical procedure based on train resistance factors, speeds, changes in elevation, and car weights;
and (2) a statistical procedure that utilizes observed data. In the second approach, a model of railroad fuel
consumption is estimated from GTMC, shipment classification, and region. In both procedures, fuel consumption
is estimated from weight and distance—e.g., GTMC. Both procedures are used to estimate the efficiencies of coal,
grain, iron ore, and other shipments. The two methods produce similar (but not identical) results. The analytical
procedure (which is previewed next) illustrates the underlying forces that affect fuel efficiency and provides a
benchmark for the statistical model.
ANALYTICAL PROCEDURE
In the analytical method, the horsepower-hours (hp-hr) required to move a uniform train of like cars over level
terrain at a steady velocity are derived from train resistance equations that account for axle, wheel tread, flange,
and air resistance. While wheel and flange resistance levels vary linearly with speed, air resistance increases with
the square of velocity and varies with the streamline coefficients of the vehicles and their cross-sectional areas. In
some cases, air resistance per ton is much greater for movements of empty cars than for loaded ones. These
differences are especially great for open hopper and gondola cars that haul coal, ore, and aggregates. For example,
the resistance of a loaded gondola car weighing 139 tons traveling at 45 mph is 4.0 lb/ton. In comparison, the
resistance of an empty gondola car weighing 23 tons and traveling at the same speed is 17.1 lb/ton.
Locomotive Resistance and Drawbar Force. Resistance factors are used to estimate the combined resistance of
a train of like cars, which, in turn, is used to estimate the number of locomotives needed as a function of
horsepower, tractive effort, and drawbar force. The latter is the residual tractive effort available to pull cars once
the locomotive resistance has been overcome.
Acceleration and Speed Limit Changes. A train routinely varies speeds as a result of changes in speed limits,
planned stops, and traffic control. Fuel consumption at steady velocities represents only part of the total energy
requirements of a trip. In this study, a speed profile is constructed from timetable speed limits. Using this profile,
the excess fuel consumed during acceleration and traffic control delays is estimated and added to the fuel
consumed at steady speeds.
Lifting Resistance. It takes 200,000 ft-lb of work to lift one ton 100 ft. Since 1.98 million ft-lb are equivalent to 1
hp-hr, it follows that 0.10 hp-hr are needed to lift one ton 100 ft in elevation. To quantify lifting resistance, an
elevation profile is developed for a mountain crossing in the western region. Using this profile, the additional fuel
consumed from changes in elevation is estimated and added to the energy needed to overcome rolling resistance
and acceleration.
STATISTICAL MODEL
While the analytical procedure is insightful, it cannot practicably be used to estimate the energy efficiency of
thousands of rail movements. For this reason, a regression model is formulated that reflects regional variations
and differences in train service. The model is estimated from 24 years of R-1 reports submitted by Class I
railroads to the U.S. Surface Transportation Board (STB). The R-1 database includes the reported gallons of fuel
consumed for freight purposes, as well as the gross ton-miles and revenue ton-miles of operation. The variables
utilized in the model are summarized in Table 1.
Train Service Levels. The type of train service has a major effect on energy efficiency. Single carloads
typically travel in way trains at origin and/or destination. These trains operate primarily between branch-
line stations and railroad yards, stopping frequently to drop off and pick up cars en route. In comparison,
through trains usually move from yard to yard, performing only limited switching en route. Unit trains
are characterized by shuttle service—e.g., the cycling of trains between origin and destination. Because
iii
way and through train movements are linked, these two categories are combined to form “non-unit train”
GTMC. A non-unit train movement may consist of individual carloads, blocks of carloads moving to the
same destination, or an entire trainload. The distinguishing characteristic is that some marshaling or
gathering of cars is required at origin and/or destination, where car blocks may arrive or depart in way
trains, traveling to or from nearby industry locations, ports, or terminals.
Regions. While many railroad mergers have occurred in the United States, railroads can be organized into three
geographic regions that have remained constant over time: east, central, and west. Each current or merged railroad
in the R-1 database is assigned to one of these regions. Once the assignments are made, the GTMC and gallons of
fuel consumed by individual railroads are summed to derive regional totals. Although the model is estimated from
Class I data, the regions are interpreted as geographic entities. While statistics for regional railroads are not
available, the physical relationship between fuel consumption and GTMC is expected to be the same for any
railroad operating the same train over the same route under the same conditions.
Indicator Variables. Each region in the R-1 database is represented by an indicator variable—e.g., Region 3.
When the observation is for the western region, Region 3 equals 1. Otherwise, Region 3 equals 0. To avoid
singularity, only n – 1 indicator variables are included in the model. The signs and magnitudes of the variables are
interpreted in relation to the excluded effect, which is subsumed in the intercept.
Parameter Estimates and Probabilities. The estimates from the fuel model and their corresponding standard
errors are shown in Table 2. As shown in column 3, the standard errors of the variables are small in relation to the
estimated values. The probabilities (or p values) in column 5 are all highly significant (i.e., values of less than
.0001), indicating less than a one in 10,000 chance of observing t values as large as those observed. The signs of
the eastern and central regions are negative in relation to the west, which is characterized by challenging grades
and rough terrain. The parameter estimate of time is positive, meaning that the fuel needed to transport a given
quantity of gross ton-miles is less today than in previous years. The main predictions are (1) after controlling for
time and region, an increase of 1,000 unit-train GTMC results in the consumption of 0.75 gallons of fuel, and (2)
an additional 1,000 GTMC in other (non-unit) trains consumes 1.11 gallons of fuel (ceteris paribus).
Key Model Properties. The model has 66 error degrees of freedom, which should be sufficient to realize large
sample properties. The R-square of 0.998 suggests that the model explains almost all of the variation in fuel
consumption. The coefficient of variation of 3.45% (which is computed as the standard error of the regression
divided by the mean of the dependent variable [gallons of fuel] multiplied by 100) suggests that the model
provides a very precise fit. Nevertheless, to confirm its precision, the model is used to predict the values (i.e.,
gallons of fuel consumed) in each region in 2008. The average prediction error is less than 1.5%.
Marginal versus Average Estimates. Two types of estimates are generated from the statistical model:
marginal and average. The former do not include overhead fuel that cannot be traced directly to GTMC.
The marginal fuel efficiencies derived from the model’s coefficients are 1,343 and 903 GTMC/G for
unit and non-unit trains, respectively. However, these values are applicable only to small changes in
railroad output and do not reflect regional variations. In comparison, the average predicted GTMC/G
values for 2008 (which include overhead and fixed-system fuel) are 699, 819, and 697 for non-unit train
movements in the eastern, central, and western regions, respectively. In comparison, the predicted
GTMC/G levels are 944; 1,188; and 931 for unit train movements in the same regions. Unlike the
marginal estimates, average values are applicable to large changes in traffic and can be used to compare
energy efficiencies among modes. However, because GTMC/G estimates reflect the tare weights of cars
and containers, they must be transformed into RTM/G before meaningful comparisons can be made.
iv
Net/Gross Ratios. GTMC/G factors are transformed into estimates of RTM/G for specific commodities and
service levels using net to gross car weight ratios. These ratios are computed from the 2008 waybill sample using
empty/loaded ratios derived from the R-1 report. The calculation of a net/gross ratio is illustrated for a plain
gondola car used in unit train service (Table 3). In this example, the net/gross ratio is computed as the net car
weight (116 tons) divided by the sum of the gross car weight (139 tons) and the tare weight (23 tons), after the
latter has been multiplied by the empty return ratio (1.0)—i.e., 116 / (139 + (23 × 1.0)) = 0.72. While many
movements are analyzed in this study, commodities of strategic importance (such as coal, grain, and iron ore) are
emphasized.
COAL
Coal is critically important to U.S. energy sufficiency. According to the U.S. Energy Information Administration
(2010), 93% of the coal mined in the United States is used to generate electricity. Approximately 71% of the coal
used by utilities is delivered by railroads. In comparison, trucks deliver less than 14% of the coal needed to
generate electricity.
Environmental Importance of Western Coal. Roughly 42% of the coal mined in the United States originates
from the Powder River Basin (PRB) of Wyoming. Although anthracite and bituminous coals generate more BTUs
per ton, PRB subbituminous coal possesses relatively low levels of sulfur and ash. In order to comply with stricter
sulfur emission regulations, many utilities throughout the United States burn PRB coal. Because the mines are
located far from navigable rivers and lakes, western coal is moved primarily to utilities and transloading facilities
by rail.
Energy Efficiency of Coal Movements. Car weights, distances, and shipment type (i.e., unit versus non-unit) are
key variables affecting fuel efficiency. Values for these parameters are estimated from the 2008 waybill sample
for coal movements within and between regions. As shown in Table 3, 93% of the coal transported by railroads
moves in unit trains of 75 cars or greater. More than half of the coal is moved in gondola cars. The remainder
moves in open-top hoppers. The average trip distance is 661 miles (Table 4). However, there are considerable
variations within and between regions. Because of terrain, the greatest efficiencies are realized within the central
region (848 RTM/G), which reflects movements of PRB coal to Texas and states in the south-central region, as
well as eastward movements to the Mississippi River, Great Lakes, and utilities located in the northern plains. In
comparison, the movement of PBR coal to the eastern United States by rail (which requires an interchange of
shipments) results in 762 RTM/G. Movements within the eastern and western regions (which travel through and
within mountain ranges) average 611 and 632 RTM/G, respectively.
Coal Unit Trains. The average gondola unit train contains 120 cars, while the average unit train of specialized
hopper cars is comprised of 112 cars. Gondola trains have the highest net/gross ratios, followed by unit trains of
specialized hopper cars. On average, coal unit trains are 60% more energy efficient than non-unit train shipments,
averaging 744 RTM/G for all regions and flows. Unit train movements in the central region are particularly
energy efficient, averaging 856 RTM/G. The average for all coal movements (irrespective of shipment type and
region) is 722 RTM/G.
Analytical Estimates of Coal Train Efficiency. The values presented above (computed from the regression
model and waybill sample) are compared to estimates derived from the analytical procedure. When variations in
speed limits and traffic control delays are considered, a coal unit train of gondola cars is expected to achieve 976
RTM/G on flat terrain. Of course, this estimate is purely theoretical, since no route is perfectly flat. For example,
PRB coal must first move north or south from the basin through rugged terrain before heading south or east. The
effects of severe gradients are illustrated in the report for shipments crossing the Rocky and Cascade Mountains.
On this route, the same unit coal train is expected to achieve 676 RTM/G, after the additional fuel needed to
overcome lifting resistance is considered. However, neither of these estimates reflects common or overhead fuel
use. For this reason, the analytical estimates will always exceed the statistical ones (ceteris paribus). In effect, the
v
analytical estimates establish an upper bound for estimates generated from the statistical model. While the
analytical procedure predicts a coal train efficiency of 976 RTM/G in flat terrain, the statistical model predicts
856 RTM/G in the flattest region. While the analytical procedure predicts a coal train efficiency of 676 RTM/G in
the western region, the statistical model predicts an efficiency rating of 654 RTM/G.
OTHER STRATEGIC COMMODITIES
Grain. Railroads transported nearly 152 million tons of grain in 2010, which is equivalent to roughly one-third of
the total tonnage (Association of American Railroads, 2011A). Approximately 48% of the grain transported by
railroads travels in unit trains with a median length of 108 cars. The vast majority of grain tonnage (i.e., 98%) is
transported in covered hopper cars. However, certain specialty crops (for which identity preservation is critical)
travel in containers. Based on 2008 waybill data, the average unit train has a net/gross ratio of 0.63 and achieves
693 RTM/G. However, the average efficiency of unit trains traveling in the central region is 748 RTM/G, versus
598 RTM/G in the west. In comparison, the average non-unit train shipment has a net/gross ratio of 0.60 and
achieves 454 RTM/G. The overall efficiency of grain shipments (irrespective of region and shipment type) is 556
RTM/G. In comparison, the estimated unit train efficiencies using the analytical method are 862 and 649 RTM/G
in flat and mountainous terrain, respectively. The average distance of all grain shipments is 1,450 miles.
Iron ore is critical to the production of steel and durable goods. Approximately 59.1 million tons of iron ore were
mined in the United States in 2008 (U.S. Geological Survey, 2008). Seventy-seven percent of this total originated
from the Mesabi Range of Minnesota. Another 23% originated from the Upper Peninsula of Michigan. According
to the waybill sample, railroads moved 53 million tons of iron ore in 2008, which equals 90% of the tonnage
mined. Approximately 75% of these tons were moved in unit trains. The average unit train consisted of 132 cars,
each weighing 25 tons empty, 116 tons loaded, and hauling 91 payload tons. More than 98% of the ore tonnage
was hauled in open hopper cars. The average trip length was 569 miles. In 2008, the average energy efficiency of
an ore unit train was 740 RTM/G. Unit train shipments in the central region (where most of the traffic is
concentrated) are even more efficient, averaging 765 RTM/G (Table 5). The overall efficiency of ore movements
in the United States (irrespective of train service and region) is 668 RTM/G.
Major Commodity Groups. Additional commodities such as ethanol and soda ash are analyzed later in the
report. However, to provide a holistic overview, the average fuel efficiencies of major commodity groups are
presented next. In this part of the study, net/gross ratios and RTM/G values are estimated at the two-digit
Standard Transportation Commodity Code (STCC) level (Table 6). As noted earlier, coal and ore shipments
generate 722 and 668 RTM/G, respectively. Movements of farm products (including grain) and food and kindred
products (including flour) achieve 543 and 411 RTM/G, respectively. In addition, railroads are an energy-efficient
means of transporting chemicals and allied products (429 RTM/G) and waste and scrap materials (389 RTM/G).
The RTM/G values for containerized freight (STCC 46) and empty containers (STCC 42) are relatively low.
Nevertheless, according to ICF International (ICFI, 2009), movements of double-stack containers by rail are more
energy-efficient than truck movements.
VALIDATION OF MODELING PROCESS
The railroad modeling process described above produces very reasonable results. The overall (predicted) net/gross
ratio using the statistical model and 2008 waybill sample is 0.55. In comparison, the actual net/gross ratio from
the R-1 report is 0.57. The overall fuel efficiency rating predicted from the statistical model and waybill sample is
456 RTM/G. In comparison, the observed 2008 fuel efficiency rating is 457 RTM/G. In effect, the predicted and
actual values are nearly identical. As these comparisons suggest, procedures based on the statistical model and
waybill sample slightly underestimate the observed net/gross ratios and efficiency ratings. However, on a national
scale, the estimates are quite accurate and do not overstate railroad fuel efficiencies. The next step in the process
is to estimate truck fuel efficiency ratings so that railroad estimates can be placed in a multimodal context.
vi
TRUCK FUEL EFFICIENCY
During the nine-year period from 2000 to 2008, the average fuel efficiency of combination trucks (i.e., tractors
pulling trailers or semitrailers) ranged from 5.1 to 5.9 mpg. The average for the period is 5.4 mpg, which is the
same as the 2008 value. However, because of grandfather clauses, many varieties of combination trucks operate in
the United States. The weights of many of these trucks exceed the 80,000-pound federal limit. Consequently, the
use of a simple mile-per-gallon average would misstate the fuel economies of specific trucks.
In an earlier study, the U.S. Department of Transportation estimated fuel efficiency factors for a wide range of
trucks, including longer configuration vehicles (LCVs). While these values are outdated, the relative efficiency
factors are not, since most (if not all) heavy duty tractors have experienced similar improvements in engine
technology to meet stricter emission regulations. Based on this assumption, a table of truck efficiency factors is
developed by indexing USDOT’s original values to account for improvements in tractor fuel economy since the
late 1990s (Table 7). Some of the trucks shown in Table 7 (such as Rocky Mountain Doubles, triple trailers, and
eight- or nine-axle combinations) are limited to turnpike operations and movements in western states. In
comparison, the five-axle tractor-semitrailer and five-axle twin trailer trucks are legal nationwide.
Variations in Fuel Economy with Weight. As shown in Table 7, fuel economy decreases with weight.
According to Delorme, Karbowski, Sharer (2009), the fuel economy of a five-axle tractor semitrailer increases by
roughly 0.06% with each thousand pounds of weight reduction. However, fuel economy varies with tractor
performance, the aerodynamics of the truck configuration, and other factors. Because of these variances,
relationships between LCV and semitrailer fuel economy are not based solely on weight.
Variations in Fuel Economy with Speed. According to Peterbilt Motors Company (2011), the power necessary
to overcome aerodynamic drag at speeds of 50 mph is equal to roughly half the power needed to overcome rolling
resistance and the energy consumed by accessories. At 75 mph, the power needed to overcome aerodynamic drag
is roughly 2.5 times the power needed to overcome rolling resistance and accessory drain (Ibid). According to
Goodyear (2008), each mile per hour increment above 55 mph increases fuel consumption by 2.2%.
Empty Truck Miles. Railroad empty/loaded ratios are reflected in the RTM/G estimates presented earlier.
However, there is no comprehensive source of empty/loaded ratios for trucks, since the Vehicle Inventory and
Use Survey (VIUS) has been discontinued. In its 2009 report, ICFI summarized data from the 2002 VIUS survey
and, in doing so, discovered that the percentages of empty miles attributable to van trailers ranged from 26% at
distances of 200-500 miles to 19% at distances > 500 miles. In comparison, the percentages of empty miles
incurred by tanker trailers ranged from 43% at distances of 200-500 miles to 31% at distances > 500 miles. The
percentages of empty miles for dump trucks ranged from 41% at distances of 200-500 miles to 38% at distances >
500 miles. Because more specific information is not available, van trailers are conservatively assumed to incur
25% empty miles. A similar assumption is made for hopper trailers hauling grain. However, backhauls are much
more difficult to obtain for specialized trailers such as hopper and dump trailers.
Fuel Attributable to Empty Truck Miles. According to simulations in Delorme et al. (2009), the fuel efficiency
of an empty five-axle tractor semitrailer is 23% to 27% greater than the fuel efficiency of the same truck loaded to
80,000 pounds, at speeds ranging from 50 to 70 mph. These estimates seem reasonable, given the fact that a large
percentage of the fuel consumed at higher speeds is attributable to aerodynamic resistance rather than weight.
Based on the aforementioned simulations, a liberal assumption is made that allows for potential energy savings
during acceleration and speed change cycles. Empty truck fuel consumption (in gallons per mile) is assumed to be
70% of loaded consumption. The same percentage of fuel savings is applied to Rocky Mountain Doubles and
twin-trailer trucks, even though the empty aerodynamics of these trucks is worse than the aerodynamics of a
tractor-semitrailer. Because of these liberal assumptions, the allowances made for empty-mile fuel savings in this
study should not understate the benefits of reduced weights.
vii
RAIL-TRUCK COMPARISONS
Modal comparisons are made on an RTM/G basis (i.e., per mile of operation). These values can be used to
estimate the fuel consumed by each mode for a specific origin and destination based on miles traveled. However,
because of differences in network coverage and routes, one mode may require more or less miles than another
mode to move a commodity between the same origin and destination. As discussed later, this inconsistency is
accounted for through the use of circuity factors.
Coal Transportation in the Eastern United States. Most of the competition for coal between railroads and
trucks occurs in the eastern United States, where several states have increased their truck weight limits for
movements on designated highways, allowing trucks to weigh as much as 60 tons, with some overweight
tolerance. Three varieties of trucks are widely used to transport coal in these states (1) three-axle trucks with
tandem rear axles, (2) five-axle tractors with semitrailers, and (3) six-axle tractor-semitrailer combinations with
triple or tridem rear axles. The maximum allowable gross weights for these trucks are 40 tons, 45 tons, and 60
tons, respectively, on designated roads. Of these trucks, the six-axle tractor semitrailer is the most economical
truck for longer hauls from mines to river transfer facilities and utilities. With a tare weight of roughly 15 tons,
this truck can haul 45 tons in a single trip. Because coal trucks often operate in shuttle service and have limited
backhaul opportunities, the empty/loaded ratio is assumed to be 1.0. Because of the six-axle truck’s greater weight
(60 tons instead of 40 tons), its fuel efficiency (4.9 mpg) is less than that of a typical 80,000-lb tractor-semitrailer.
With 50% empty miles, the average efficiency of this truck is 130 RTM/G. In comparison, the average
efficiencies of non-unit and unit train coal shipments in the eastern United States (as predicted from the regression
model and waybill sample) are 460 and 637 RTM/G, respectively. As these comparisons suggest, non-unit and
unit train coal movements in the eastern United States are 3.5 and 4.9 times more fuel efficient than movements of
the heaviest truck approved for travel on designated state routes on a per mile basis.
Coal Transportation in the Western United States. While most western coal is moved to the central and
eastern United States, some movements occur within the western region. In several western states, coal is moved
in LCVs under exceptions to the 80,000-pound vehicle weight limit. In some areas, coal is moved in double-
trailer combinations with gross weights of 129,000 pounds or more. With a tare weight of 22.5 tons, these trucks
can haul 42 tons in a single trip. With a fuel efficiency rating of 5.2 mpg and no backhaul, these trucks can
achieve 128 RTM/G. In comparison, the average fuel efficiencies of non-unit and unit train movements in the
western United States (as predicted from the regression model and waybill sample) are 444 and 654 RTM/G,
respectively. In effect, these shipments are 3.7 to 5.4 times more fuel efficient than movements in the heaviest
truck allowed under grandfather clause exemptions, on a per mile basis.
Grain Transportation. Grain is shipped from elevators to markets, transfer locations, and export facilities in
large commercial trucks. The two most common types are the five-axle tractor semitrailer and the seven-axle
Rocky Mountain Double—which consists of a 40- to 48-foot semitrailer followed by a smaller “pup” trailer with
two single axles. A third truck configuration (which is used primarily in cross-border movements between the
United States and Canada) has four sets of tandem axles and nine axles, altogether. All three trucks are equipped
with hopper trailers with top loading and bottom (gravity) discharging capabilities. The typical weights of these
trucks are shown in Table 8, where the gross weight of the twin-trailer truck is an average of individual state
weight limits that range from 105,500 to 137,800 pounds. Movements of this truck are confined to limited
geographic areas near the northern border. In comparison, Rocky Mountain Doubles are common in the western
United States under grandfather clauses, but are rare east of the Mississippi River. However, the tractor-
semitrailer combination is legal nationwide and is therefore the dominant truck used in grain transportation.
viii
RTM/G Estimates for Grain Trucks. Because of their specialized nature, the empty/loaded mile ratios are
higher for hopper trailers than for van trailers. If half of the truck’s miles are empty (e.g., zero backhaul), the
estimated fuel efficiencies are 86, 109, and 126 RTM/G for the five-axle, seven-axle, and nine-axle trucks,
respectively. If 25% of the truck’s miles are empty, the estimated fuel efficiencies are 119, 150, and 174 RTM/G
for the five-axle, seven-axle, and nine-axle trucks, respectively.
Relative Energy Efficiencies of Railroad and Truck Grain Movements. As noted earlier, unit and non-unit
train movements of grain achieve 693 and 464 RTM/G, respectively. The overall efficiency rating of railroad
grain movements is 556 RTM/G. As these factors suggest, grain movements by rail are 6.5 times more fuel
efficient on a per-mile basis than grain shipments in five-axle tractor-semitrailer trucks with no backhaul (column
2 of Table 9), and 5.1 and 4.4 times more fuel efficient than movements in Rocky Mountain Doubles and twin-
trailer trucks, respectively, with no backhaul. The relative fuel advantages are even greater for grain unit trains,
which are 8.1, 6.4, and 5.5 times more efficient than highway movements in five-axle tractor-semitrailers, Rocky
Mountain Doubles and twin-trailer trucks, respectively (column 3). If trucks incur 25% empty miles, grain unit
trains are 5.8, 4.6, and 4.0 times more fuel efficient than highway movements in five-axle tractor-semitrailers,
Rocky Mountain Doubles and twin-trailer trucks, respectively (column 5). The energy advantage of railroads is
even greater in the central region, where grain unit trains achieve their maximum fuel efficiency of 748 RTM/G.
In this region, a railroad unit train is five times more fuel-efficient than a Rocky Mountain Double, on a per mile
basis.
Energy Efficiencies of Iron Ore Movements by Rail and Truck. Minnesota does not allow LCVs. So, the
heaviest vehicle that can operate with a routine permit is a six-axle tractor-semitrailer that weighs 90,000 pounds.
With a tare weight of 13.5 tons and 25% empty miles, this truck can achieve 138 RTM/G. The maximum weight
of a double trailer combination in Michigan with typical axle spacing and no special permit is 109,000 pounds. At
25% empty miles, this truck can achieve 157 RTM/G. As noted earlier, the efficiency of an ore unit train in the
central region is 765 RTM/G. Thus, on a per mile basis, an ore unit train is 4.9 and 5.5 times more fuel efficient
than highway movements in Michigan and Minnesota, respectively, in the heaviest combination trucks routinely
allowed in those states. With these ratios, it is not surprising that iron ore in the central region moves almost
exclusively by rail and water.
Flour and Grain Mill Products. Railroads transported more than 10 million tons of flour and related grain mill
products in 2008. All of these shipments moved as individual carloads. About 86% of this tonnage moved in
covered hopper cars with an average load factor of 93 tons. The average trip length was 784 miles. The overall
efficiency rating was 429 RTM/G. Because final demand sites (such as bakeries) are distributed throughout the
United States and because of the protective needs of the cargo, truck transportation is the only feasible alternative
to rail. While it is possible for bulk flour to be transported in LCVs, flour is more likely to move in five-axle
tractor semitrailers. With 25% empty miles, the typical efficiency of a flour movement by truck is 118 RTM/G.
On average, railroad movements of bulk flour are 3.6 times more fuel efficient than highway movements on a per
mile basis. However, this ratio may not be applicable to bagged flour movements in van trailers and boxcars, or
other bagged or packaged grain mill products.
Soda Ash. Sodium carbonate or soda ash is an important industrial compound that is essential to the manufacture
of glass, detergents, paper, and various chemicals. In 2008, approximately 12.5 million tons of soda ash were
produced in the United States, mostly in Wyoming and California. Most of this product is transported long
distances (e.g., an average of 1,300 miles) to manufacturers located throughout the United States by railroads in
covered hopper cars with an average load factor of 104 tons and an average efficiency rating of 479 RTM/G.
Because of the west to east orientations of these movements, river transportation is infeasible on a broad scale.
Moreover, because of the nationwide distribution of soda ash, the only practical truck option is the five-axle
tractor semitrailer which, with 25% empty miles, can achieve 118 RTM/G. In effect, soda ash movements by rail
are 4.1 times more energy efficient than movements by truck on a per mile basis.
ix
Ethanol. Movements of ethanol are important to clean fuel programs in several states, including California.
According to the American Association of Railroads (AAR, 2011B), railroads account for 70% to 75% of ethanol
transportation. Ethanol is transported primarily in large tank cars with an average load factor of 93 tons. With a
net/gross ratio of 0.58, these movements realize an efficiency rating of 443 RTM/G and are 3.9 times more fuel
efficient than movements in five-axle tractor semitrailers with 25% empty miles. Because of the east-west
orientations of these movements, river transportation is infeasible on a broad scale. However, waterway
transportation is critically important to coal and grain logistics.
WATERWAY FUEL EFFICIENCY MODEL
Waterway fuel consumption rates are estimated from the River Efficiency Model (REM) developed by the
Tennessee Valley Authority (TVA). REM uses vessel and lock performance models to estimate horsepower and
speed for each river segment, as well as average processing and delay time at each lock. Fuel usage is derived
from speed, horsepower, lock time, and other performance data. The individual computations for river segments
are added to determine the total fuel consumption for each waterway. While REM cannot be used to analyze fuel
consumption for specific commodity movements, average fuel efficiency factors can be estimated for particular
waterways. These factors (shown in Table 10) reflect both loaded and empty barge movements and fuel consumed
during lock transits and queuing.
RTM/G Estimates by River Segment. While railroad energy efficiency has increased consistently over time,
waterway fuel efficiency has fluctuated from year to year. Moreover, there are considerable differences in fuel
efficiency among river segments (Table 10). In 2008, the average efficiency from Minneapolis to the mouth of the
Missouri River was 348 RTM/G. The average efficiency on the Illinois River was 287 RTM/G. However, down-
river fuel efficiencies were much greater—e.g., 656 RTM/G for the open river segment from the mouth of the
Ohio River to Baton Rouge, LA. In 2009, the estimated fuel efficiencies were 482 and 771 RTM/G for the
Minneapolis to Missouri River and Ohio River to Baton Rouge segments, respectively. Moreover, the estimated
efficiency of the Illinois River was much greater than in 2008—i.e., 395 RTM/G instead of 287. These differences
are partly explained by the major flood that occurred on the Mississippi River in 2008, resulting in the second
highest crest ever recorded for stations such as Quincy, IL. Because of the flood, a three-year average of
waterway fuel efficiency (centered around 2008) is used in the comparison (column 3, Table 10). While averaging
has a minimal effect on Ohio River fuel efficiency, it significantly alters the values used for the Mississippi and
Illinois Rivers.
Illinois and Upper Mississippi Rivers. In addition to floods, the fuel efficiencies of the Upper Mississippi and
Illinois Rivers are constrained by small aging locks, many of which are programmed for replacement in the future.
At many of these locks, barge operators must separate a tow into two components, and move the blocks through
separately. Afterward, the tow must be reassembled. This process may be repeated at several locks as a shipment
travels on the Upper Mississippi River or Illinois Waterway. Queuing and double locking maneuvers significantly
increase fuel consumption. The estimated RTM/G ratings for shipments originated from upper river stations
(which are computed by weighting the RTM/G estimates on each river segment by the distance traveled) are
shown in Table 11. A shipment from Minneapolis (the head of navigation) to the port of New Orleans yields 619
RTM/G, while a shipment from Quincy, IL (located only 130 miles from the mouth of the Missouri River) to New
Orleans yields 702 RTM/G.
Ohio River System. The average revenue ton-miles per gallon for shipments on the Ohio River (592 RTM/G) is
higher than on the Upper Mississippi River (436 RTM/G). Ohio River fuel efficiency is illustrated for a barge
shipment from Greenup, KY, (located 640 miles from the mouth of the Ohio River) to the port of New Orleans.
The estimated efficiency of this shipment (which travels a total of 1,492 river miles) is 678 RTM/G.
x
General Inferences Regarding River Transportation. Several patterns are apparent from the data. (1) Open
river fuel efficiencies are the greatest. Propulsive energy is minimized when loaded barges travel downstream
with the current. (2) Shipments that travel the shortest distances on the Upper Mississippi or Illinois Rivers are
more efficient than shipments originated farther upstream. Movements originated in the northernmost reaches of
the river must transit more locks en route to New Orleans and, thus, consume more fuel in queuing and lock
transits.
Rail versus Barge Energy Efficiency in Coal Transportation. As shown in Table 10, the average fuel
efficiency of shipments on the Ohio River is 592 RTM/G. In comparison, the average efficiency of coal shipments
in the eastern United States is 611 RTM/G. A precise comparison is difficult because the comparative distances
from mines to rail and river transfer facilities are unknown. Nevertheless, it appears that the energy efficiency of
coal shipments in the eastern region is at least comparable to barge shipments on the Ohio River. As noted earlier,
coal mined in the western United States is often railed from the Powder River Basin to the Great Lakes or Upper
Mississippi River Valley. Because these movements are complementary, direct modal comparisons are not
considered. Nevertheless, it can be said that unit coal train movements in the central region are more energy-
efficient than barge movements on the Illinois or Upper Mississippi Rivers.
Grain Movements from Upper Mississippi River Stations to the Gulf. As noted earlier, the weighted average
fuel efficiency of a waterway movement from Minneapolis to New Orleans is 619 RTM/G. This is higher than the
average for a non-unit train movement within the central region (497 RTM/G). However, the typical waterway
fuel efficiency is less than that of a grain unit train (748 RTM/G) traveling in the same region. Moreover, a unit
train movement in the central region is more fuel efficient than movements from all of the Upper Mississippi
River stations shown in Table 11.
Grain Movements from Upper Ohio River Stations to the Gulf. As noted earlier, the average fuel efficiency of
barge shipments from Greenup, KY, to New Orleans is 678 RTM/G. In comparison, a non-unit train movement
within the eastern region yields 431 RTM/G, while a grain unit train movement generates 596 RTM/G. In this
example, Ohio River barge movements offer greater energy efficiencies than railways for movements of grain to
the Gulf of Mexico at average regional efficiency levels.
TRANSPORTATION CIRCUITY
The comparisons presented thus far have not been adjusted for circuitous routing. As noted earlier, the
comparisons reflect a one-mile haul by each mode. If one mode requires fewer miles than another to move the
same product between origin and destination, such comparisons could misstate the energy requirements of a trip.
In surface transportation, out-of-line movements and meandering routes create additional mileage known as
“circuity,” which is the difference between the most direct (as the crow flies) route and the actual miles traveled
between origin and destination. There are essentially two types of circuity: (1) network and (2) route. The former
is a function of the built transportation network and how much it deviates from great circle distances. The latter is
a measure of the extent to which shipments do not take the most direct path between origin and destination. In
some cases, a longer alternative path with a shorter travel time may be selected. In other cases, a longer path may
be followed to avoid congestion, severe gradients, or hazards. The consolidation and distribution of freight at
origin and destination may contribute to circuity, resulting in way train and truck flows counter to the prevailing
direction of movement. Total circuity is the sum of network and route circuity.
Circuity Factors. A circuity factor or multiplier is equal to 1.0 plus the percentage (or proportion) of circuitous
miles. The latest factors were estimated in the early 1980s. According to the Congressional Budget Office (CBO,
1982), the average network circuity factors for intercity carriers are 1.83 for barges, 1.32 for railways, and 1.15
for trucks. In comparison, route circuity multipliers are 1.15 and 1.06 for railroads and trucks, respectively (Ibid).
Barge route circuity is assumed to be zero, since there is rarely more than one river path between an origin and
destination. While CBO’s truck estimate is dated, it may still be applicable, because most of the interstate and
xi
intercity arterial highway network was in placed by 1982. However, the latest railroad estimate of 1.135 was
developed in 1983 (STB, 2009). This circuity factor reflects both unit and non-unit train traffic, which have
substantially different operating characteristics. Because out-of-line routing for purposes of yard classification is
not required, unit trains typically travel the shortest path between origin and destination. In the Uniform Railroad
Costing System, the STB assumes a circuity factor of 1.0 for unit trains (2009). After adjusting for the substantial
increase in unit train traffic that has occurred over time, the weighted-average route circuity factor of 1.09
(estimated for 2008) reflects a non-unit train circuity factor of 1.16 and a unit-train circuity factor of 1.0. Using
these factors, the overall ratio of railroad to truck circuity for non-unit train movements is 1.22. For unit train
movements, the ratio of rail to truck circuity is 1.09. For unit train movements, the ratio of barge to rail circuity is
1.24.
Railroad-Truck Comparisons Adjusted for Circuity. The ratios of railroad to truck and railroad to barge
energy efficiencies discussed earlier are summarized in Table 12 (columns 3-5), along with a set of ratios that
have been adjusted for circuity (columns 6-8). As expected, the railroads’ energy advantage over trucks drops
after the adjustments. However, railroads still hold substantial energy advantages for all of the commodities
analyzed. For example, coal unit trains are still 4.5 and 5.0 times more energy efficient than truck movements in
the largest vehicles allowed under grandfather clauses in the eastern and western regions respectively. Unit grain
train movements in the central region are still 4.6 times more fuel efficient than movements in Rocky Mountain
Doubles.
Effects of Circuity on Railroad and Barge Energy Comparisons. While railroad to truck efficiency ratios drop
when circuity is considered, railroad to barge ratios increase. The Minneapolis to New Orleans movement is an
interesting example. The shortest railroad distance is 1,273 miles. In comparison, the river distance is 1,706 miles.
In this example, the actual ratio of barge to rail circuity is 1.34, instead of the 1.24 factor computed earlier.
However, this ratio reflects only network circuity. If another route to New Orleans is selected, the rail distance
will be greater. As a result, the barge-to-rail circuity ratio will be much smaller.
ISSUES AND LIMITATIONS
Access or Drayage Energy Consumption. Truck movements are necessary to move grains from farms to
elevators located on rail lines and rivers. However, because of the diversity of these trips, it is difficult to make
generalizations about the truck portions of railroad and waterway movements. If there are substantial differences
in the distances attributable to connecting truck movements, the comparisons presented earlier may be unrealistic.
Drayage fuel consumption for container-on-flatcar (COFC) movements is discussed in the main report.
Issues with Railway and Waterway Comparisons. In previous studies, system-average railroad and waterway
RTM/G values were compared and inferences drawn that waterways are more energy efficient than railways
(Kruse, Protopapas, Olson, and Bierling, 2007). While this is true overall, the comparison is not completely
relevant. The inland waterway traffic mix consists largely of bulk commodities such as coal, ore, petroleum,
chemicals, aggregates, and farm products. In contrast, the railroad traffic base includes automobiles, auto parts,
loaded and empty containers, and a multitude of neobulk and manufactured goods for which railways and trucks
compete. For these reasons, comparisons of railway and waterway averages do not necessarily result in “apples to
apples” comparisons. A different picture emerges when the commodities moved by rail are organized into two
categories (1) bulk commodities and heavy manufactured goods that also move via waterways and (2) low-
density, containerized, and/or time-sensitive goods for which railways and trucks compete. If the second group is
excluded from the calculation, the railroads’ energy efficiency increases from 457 RTM/G to 581 RTM/G. For
purposes of comparison, the average river efficiency during the 2007-2009 period was 592 RTM/G. While this
comparison is approximate, it indicates that railroad energy efficiency is approaching waterway efficiency levels
for similar commodities. However, the comparison is lacking in several respects. There are significant differences
in the ways that railroads and waterways make investment decisions. Railroads make decisions based on market
priorities. In contrast, waterway investments are approved (or disapproved) by Congress through a lengthy
xii
political process. While there is no certainty that railway investments are optimized, many of the locks on the
Upper Mississippi and Illinois Rivers are outmoded. The modernization of locks on these segments could affect
forecasts of modal energy efficiency. Nevertheless, future waterway investment levels will not alter the fact that
railroads are an efficient mode of transporting strategic bulk commodities and are critically important to national
energy goals.
Additional Issues. The methods, data, assumptions and limitations of the study are discussed in greater detail in
the main report. However, some interpretive issues are highlighted next. (1) The outcomes and implications of
fuel efficiency comparisons may be different, depending upon whether marginal or average estimates are used.
Most studies do not explicitly say which factors are being used and why. (2) Comparisons of railway and truck
energy efficiencies may vary from state to state depending on the types of grandfathered trucks allowed and
overweight permitting processes. (3) While fuel consumption for way train movements on Class I railways is
reflected in this study, specific estimates for local and regional railroad operations are not. The fuel consumption
rates for these movements are assumed to be similar to those experienced by Class I railroads within the same
region. Because local railroad movements are short in duration, they should not have a strong influence on the
conclusions. (4) While railroad energy efficiencies are specifically developed for individual commodities, river
efficiency ratings reflect the average of all commodities moving on a river segment. While most of these
movements are comprised of bulk commodities, there may be substantial differences in the net tonnage per barge
among commodity groups. These potential differences are not reflected in the comparisons. As a result, railroad-
barge comparisons for specific commodities may reflect averaging errors. (5) Railroad and waterway comparisons
are further complicated by the periodic flooding of the Mississippi River basin. Because floods are a reality of
waterway transportation, they must be reflected in the comparisons. In this study, a three-year waterway average
(centered on 2008) is used to compare waterway and railway fuel efficiencies. Overall, the use of a three-year
average yields a waterway efficiency rating of 592 RTM/G, which seems reasonable, given that the most recent
five-year average is 588 RTM/G. In essence, the use of a longer averaging period has little or no effect on the
comparisons presented in this report.
Table 1. Variables in Railroad Fuel Regression Model
Variable
Description
Type
GAL
Gallons of fuel consumed in train and yard freight service
Response
UGTMC
Thousands of gross ton-miles of cars and contents moved in unit trains
Structural
NGTMC
Thousands of gross ton-miles of cars and contents moved in non-unit
trains
Structural
T
Time in years before 2008 (2008 = 0)
Control
REG
A set of regional variables {1, 2, 3}
Indicator
Table 2. Parameter Estimates from Railroad Fuel Model
Parameter
Estimate
Standard Error*
t-value*
Probability of > |t|*
Intercept
713,465,017
67,216,343
10.61
<.0001
UGTMC
0.74449
0.10696
6.96
<.0001
NGTMC
1.10767
0.11788
9.40
<.0001
T
3,339,195
581,239
5.74
<.0001
REG1
-476,171,943
29,754,620
-16.00
<.0001
REG2
-701,468,788
57,404,842
-12.22
<.0001
*Heteroscedasticity consistent estimates
xiii
Table 3. Weights of Railcars Used in the Transportation of Coal
Shipment
Type
Car Type
Percent
of Cars
Net
Weight
(Tons)
Tare
Weight
(Tons)
Net/
Gross
Ratio
Non-Unit
Plain Gondola
1.7%
102
30
0.64
Open Hopper- General Service
2.5%
98
30
0.62
Open Hopper- Special Service
3.4%
112
27
0.68
Unit
Plain Gondola
49.7%
116
23
0.72
Open Hopper- General Service
10.4%
105
29
0.65
Open Hopper- Special Service
32.3%
115
25
0.70
Table 4. Energy Efficiency of Coal Movements
Average Distance
Revenue Tons Miles per Gallon
Shipment Type
Region
(Miles)
Non-Unit
Unit Train
Average
East (Internal)
368
460
637
611
Central (Internal)
960
511
856
848
West (Internal)
497
445
654
632
Central-East
1,280
462
767
762
East-West
1,733
448
629
629
West-Central
966
472
750
737
United States
661
465
744
722
Table 5. Car Weights and Fuel Efficiencies of Iron Ore Movements
Shipment Type
Region
Net/Gross
RTM/G
Non-Unit
East
0.61
425
Central
0.63
513
West
0.59
410
Unit
East
0.63
595
Central
0.64
765
xiv
Table 6. Fuel Efficiencies of Major Commodity Movements at Two-Digit STCC Level
STCC
Product Name
Net/Gross
RTM/G
01
Farm Products
0.61
543
10
Metallic Ores
0.63
668
11
Coal
0.70
722
14
Nonmetallic Minerals, except Fuels
0.63
524
20
Food or Kindred Products
0.54
411
24
Lumber or Wood Products
0.54
397
26
Pulp, Paper or Allied Products
0.49
356
28
Chemicals or Allied Products
0.57
429
29
Petroleum or Coal Products
0.51
385
32
Clay, Concrete, Glass or Stone Products
0.62
459
33
Primary Metal Products
0.56
412
37
Transportation Equipment
0.24
177
40
Waste or Scrap Materials
0.54
389
42
Empty Containers or Shipping Devices
0.23
165
46
Miscellaneous Mixed Shipments
0.33
246
Table 7. Average Miles per Gallon by Truck Configuration and Weight
Gross Vehicle Weight (pounds)
Truck Configurations
60,000
80,000
100,000
120,000
Five-Axle Semitrailer
6.2
5.5
4.9
Six-Axle Semitrailer
6.1
5.4
4.9
Five-Axle Double Trailer
6.8
6.0
5.5
Seven-Axle Rocky Mountain Double
5.8
5.2
5.0
Eight Axles (or more)
5.8
5.5
5.2
Triple-Trailer Combination
6.0
5.7
5.4
U.S. Department of Transportation. Western Uniformity Scenario Analysis: A Regional Truck Size and Weight Scenario Requested by
the Western Governors’ Association, April 2004. Original values indexed by 1.14.
Table 8. Weights of Combination Grain Trucks
Truck Configuration
5-Axle 48-ft
Semitrailer
7-Axle Rocky
Mountain Double
9-Axle
Twin Trailers
Gross Weight (lb)
80,000
105,500
127,400
Tare Weight (lb)
26,800
34,200
45,000
Net Weight (tons)
26.6
35.7
41.2
xv
Table 9. Ratio of Railroad to Truck Energy Efficiency in Transporting Grain
Truck Configuration/Train Type
50% Empty Truck Miles
25% Empty Truck Miles
Average
Unit Train
Average
Unit Train
Five-Axle Semitrailer
6.5
8.1
4.7
5.8
Seven-Axle Rocky Mountain Double
5.1
6.4
3.7
4.6
Nine-Axle Twin 48-ft Trailers
4.4
5.5
3.2
4.0
Note that Rocky Mountain Doubles operate primarily in the western United States, while nine- and ten-axle grain trucks operate
largely in cross-
border movements to and from Canada. A simple average of the values in any column of the table may be misleading,
since most of the shipments occur in five-axle units.
Table 10. Waterway Fuel Efficiency Factors for Specific River Systems
Waterway Name
Revenue Ton-Miles per Gallon
2008
Average 2007-2009
Mississippi River, Mouth of Ohio to Baton Rouge
656
742
Mississippi River, Mouth of Missouri to Mouth of Ohio
584
703
Mississippi River, Minneapolis to Mouth of Missouri
348
436
Illinois River (IL)
287
348
Ohio River
603
592
Tennessee River (TN, AL and KY)
445
453
Tennessee Tombigbee Waterway
354
377
Table 11. Average Revenue Ton-Miles per Gallon for Barge Shipments from Upper Mississippi River
Stations to New Orleans: 2007-2009
Station
Distance to Mouth of Missouri
River
Weighted RTM/G
Minneapolis, MN
659
619
Minnesota City, MN
543
633
Dubuque, IA
388
654
Clinton, IA
327
664
Keokuk IA
169
693
Quincy, IL
130
702
xvi
Table 12. Ratio of Railroad-to-Alternate Mode Fuel Efficiency, Adjusted for Circuity
Commodity and Region
Alternate
Mode
Unadjusted Ratios
Adjusted Ratios
Non-
Unit
Average
Unit
Train
Non-
Unit
Average
Unit
Train
Coal
East Region
Truck
3.5
4.7
4.9
2.9
4.0
4.5
East Region
Barge
0.7
0.9
0.9
0.9
1.2
1.3
West Region
Truck
3.7
5.2
5.4
3.0
4.4
5.0
Grain
All Regions
5-axle
3.8
4.7
5.8
3.1
4.0
5.3
All Regions
7-axle
3.0
3.7
4.6
2.5
3.2
4.2
All Regions
9-axle
2.6
3.2
4.0
2.1
2.7
3.7
Central Region
7-axle
3.3
4.1
5.0
2.7
3.5
4.6
Minneapolis to Gulf
Barge
0.8
1.0
1.2
1.0
1.3
1.7
Greenup, KY, to Gulf
Barge
0.6
0.7
0.9
0.7
0.9
1.0
Iron Ore
Michigan
Truck
3.3
4.6
4.9
2.7
3.9
4.5
Minnesota
Truck
3.7
5.2
5.5
3.0
4.4
5.0
Flour (All Regions)
Truck
3.6
3.0
0.0
Soda Ash (All Regions)
Truck
3.9
4.1
5.3
3.2
3.5
4.9
Ethanol (All Regions)
Truck
3.7
3.9
5.2
3.0
3.3
4.8
1
1. INTRODUCTION
The fuel efficiency of freight transportation is a critical issue in the United States in light of the price
volatility of fuels and America’s dependence upon foreign sources of petroleum. Moreover, fuel
efficiency is important to environmental policy. The burning of fossil fuels for transportation purposes
results in the emission of air pollutants such as nitrogen oxides, particulate matters, and volatile organic
compounds. According to the U.S. Environmental Protection Agency (EPA), transportation is responsible
for 27% of the total greenhouse gas emissions in the United States. Moreover, transportation is the
fastest-growing source of greenhouse gas emissions, accounting for 47% of the net increase in U.S.
emissions since 1990. Today, transportation is the largest end source of carbon dioxide (CO2), the most
prevalent greenhouse gas. Because emissions increase with gallons of fuel consumed, fuel efficiency is a
necessary condition for improved environmental quality.
Perhaps the most common measure of transportation fuel efficiency is revenue ton-miles per gallon
(RTM/G), which is an indication of the transportation output produced by a gallon of fuel. It measures
how many miles one ton of cargo can be moved with a gallon of fuel. On the basis of this measurement,
U.S. railroads increased their fuel efficiency rating from 283 RTM/G in 1985 to 457 RTM/G in 2008
(Figure 1.1), which represents a 61% increase over 24 years, or a 2.6% annual growth rate. A
complementary indicator of fuel efficiency is the gross ton-miles of cars and contents (GTMC) moved
with a gallon of fuel. In addition to revenue or cargo tons, this indicator reflects the weights of containers,
trailers, and freight cars. As shown in Figure 1.1, the gross ton-miles of cars and contents produced with
each gallon of fuel (GTMC/G) has increased from 558 in 1985 to 806 in 2008, a 45% increase in 24
years, or a 1.9% annual growth rate.
As these trends suggest, there are two primary sources of energy efficiency gains. (1) Railroad fuel
savings are due in large part to technological and operational efficiencies that have enabled the movement
of a given quantity of weight with fewer gallons. (2) Higher levels of revenue ton-miles per gallon are
attributable, at least in part, to moving more revenue tons in a single car—i.e., increasing the net to gross
weight ratio. By either measure (GTMC/G or RTM/G), railroad fuel savings have increased dramatically
in the United States since 1980, when the railroads were deregulated—a policy change that has allowed
greater operational flexibility and spurred innovations leading to fuel savings.
With a few notable exceptions, most of the comparisons of modal fuel efficiencies have utilized system
averages. While insightful, these generalizations do not distinguish among different types of operations,
such as railroad unit trains versus individual carload shipments. The average railroad revenue ton-miles
per gallon reflects all types of train movements (unit trains, mixed freight trains, and way or local trains),
as well as commodities (bulk, neobulk, containers, etc.). As shown later, there are wide disparities in fuel
efficiencies within these classes.
2
Figure 1.1 Trend in Railroad Fuel Efficiency in the United States
The main objectives of this study are to (1) conduct an analytical evaluation of movements using train
resistance and fuel consumptions equations, (2) estimate a railroad fuel efficiency model that describes
differences in fuel economy among classes of movements, (3) compare the results of the two methods to
each other and to other studies, and (4) compare rail, truck, and waterway fuel efficiencies. Although
container shipments are analyzed, the primary focus is on bulk commodity movements. The rest of the
report is organized as follows. (1) The most relevant previous studies are reviewed. (2) Resistance
equations are used to estimate the horsepower-hour requirements (and gallons of fuel consumed) for
specific train movements. (3) Statistical equations are estimated from fuel and operational data reported
by railroads to the U.S. Surface Transportation Board (STB) in the R-1 report. (4) The results of the
analytical and statistical methods are compared. (5) The relative fuel efficiencies of the primary
commodities transported by rail are analyzed using the waybill sample. (6) The energy efficiencies of
railroads, trucks, and barges are compared for bulk commodity movements.
-
100
200
300
400
500
600
700
800
900
1985 1990 1995 2000 2005
Gallons of Fuel per Ton-Mile
GTMC/G RTM/G
3
2. BRIEF REVIEW OF RAILWAY ENERGY STUDIES
A series of studies were conducted in the 1970s to provide simplified estimation techniques and
generalized values for use in state rail planning in the United States (Table 2.1). Although outdated, these
factors illustrate two key points. (1) Railroad fuel consumption is greatest for branch line or way train
operations (e.g., short hauls). (2) The fuel efficiencies of unit trains are significantly greater than those of
other trains. In another study from the same era, the Congressional Budget Office (CBO, 1982)
synthesized the results of individual analyses conducted by railroads and concluded that the typical fuel
efficiencies were 139 RTM/G for trailer-on-flatcar (TOFC) trains and 375 RTM/G for unit coal trains.
Table 2.1 Fuel Consumption Factors Used in State Rail Planning (1978)
Type of Train Service Gallons of Diesel Fuel per
1,000 Net Ton Miles
Revenue Ton Miles
per Gallon
Short-Haul Rail 24.90 40
Through Train 5.05 198
Unit Train
2.38
420
Trailer-on-Flatcar (TOFC) 5.77 173
Source: Federal Railroad Administration (FRA). Rail Planning Manual, Volume II, 1978.
In 1999, Gervais and Baumel used computer simulations provided by two Class I railroads to analyze unit
grain train movements from Boone, IA, to New Orleans and Los Angeles. The hypothetical movements
consisted of 100-car trains for two different car load factors: 100 and 110 tons (Table 2.2). A 110-car
movement from Sioux City, IA, to Tacoma, WA, was also evaluated in the study. Each trip was simulated
with three different types of locomotives.
Table 2.2 Simulated Fuel Efficiency of Unit Grain Trains Movements from Iowa
Origins-Destinations Simulations
Cars per
Train
Loads per
Car
RTM/G
Range
Boone, IA, to Los Angeles 6
100
100, 110
513-585
Boone, IA, to New Orleans 6
100
100, 110
688-802
Sioux City, IA, to Tacoma
2
110
110
554-664
Source: Gervais and Baumel.
Fuel Consumptions for Shipping Grain Varies by Origin, Destination. Feedstuffs, Vol. 71(36).
All of the values shown in Table 2.2 reflect the empty movements of covered hopper cars before and/or
after the loaded movements. While all movements are energy efficient, there are substantial variations in
the results. For the most part, rail shipments from Iowa to New Orleans travel over flat or gentle terrain.
As a result, the estimated fuel efficiencies of these movements (688 to 802 RTM/G) are significantly
greater than the efficiencies of movements from Iowa to the Pacific Coast, which must cross the Rocky
Mountains.
A recent study by ICF International for the Federal Railroad Administration (ICFI, 2009) focused on the
comparative efficiencies of rail and truck transportation for a select set of truck-competitive commodities
hauled by railroads. The 2009 report is actually an update of a 1991 study by Abacus Technology
Corporation. In the ICFI study, truck fuel efficiencies for 23 movements were analyzed using the Physical
Emission Rate Estimator (PERE) model developed by the U.S. Environmental Protection Agency. Line-
haul railroad fuel consumption was estimated for the movements by “two participating railroads with in-
house train simulators” (ICFI, 2009, p 56). The simulations were based on “actual rail routes with their
4
grade profile, curvature, and speed limit changes” (ICFI, 2009, p 57). However, because of data
inconsistencies, the effects of speed limit changes and curvature were not reflected in the analysis.
The estimates for the 12 intermodal movements analyzed in the ICFI study are shown in Table 2.3. For
the most part, these are heavy double-stack container-on-flatcar (COFC) shipments with a median lading
of 49 tons. The estimated fuel efficiencies of the movements range from 226 RTM/G to 512 RTM/G.
According to ICFI, these movements are 3.5 to 5.1 times more fuel efficient than comparable truck
movements.
Table 2.3 Range of Movement Factors and Fuel Efficiencies of Double-Stack Container
Movements
Variable
N
Minimum
Median
Mean
Maximum
Cars per Train
12
27
66
60
85
Distance (miles)
12
294
1,225
1,231
2,232
GTMC
12
4,243
5,629
5,922
8,107
Tare Weight (tons)
12
21
31
32
48
Net Weight (tons)
12
15
49
47
70
GTMC/G
12
523
646
669
849
RTM/G
12
226
379
384
512
ICF International. Comparative Evaluation of Rail and Truck Fuel Efficiency on Competitive Corridors. Published by
Federal Railroad Administration, Washington, D.C., 2009
Motorized vehicles and TOFC shipments were also analyzed by ICFI. The estimated fuel efficiencies of
the three selected automobile shipments are 156, 157, and 164 RTM/G, while the estimated fuel
efficiency of the lone TOFC simulation is 273 RTM/G. The authors conclude that the rail TOFC
movement is 3.2 times more efficient than a comparable truck movement, while the automobile shipments
are 1.9 to 2.2 times more fuel efficient than highway trailer movements.
Although useful, the ICFI study addresses only one element of freight flows: truck-rail competitive
movements in select corridors. Bulk commodity movements are not analyzed. While the truck fuel
consumption estimates from the PERE model can be independently verified, the estimates from the in-
house train simulations cannot. With the exception of the two previous quotes, no descriptions are
provided of the methods used in estimating railroad fuel consumption or the fundamental underlying
relationships.
5
3. ANALYTICAL MODELS OF RAILROAD FUEL CONSUMPTION
Analytical methods of modeling railroad fuel consumption are described in this section of the paper. In
section 3.1, forces and resistance equations are introduced. In section 3.2, procedures for estimating
locomotive requirements and the energy expended during train movements are outlined. In section 3.3,
the procedures are applied to uniform train movements of bulk, neobulk, and container shipments. The
usefulness and limitations of the approach are summarized in section 3.4.
3.1 Resistance Forces and Equations
The energy requirements of a freight movement are a function of speed, weight, and resistance, which
includes (1) axle (bearing) resistance, (2) wheel (rolling) resistance, (3) flange resistance, (4) air
resistance, (5) track resistance, (6) curve resistance, and (7) grade resistance. Rolling resistance results
mostly from friction between wheel treads and the running surfaces of rails, while flange resistance
results from contacts between wheel flanges and the inside heads of rails. Track resistance is a function of
deflection. Collectively, the first five forces are referred to as train resistance (R). Train resistance (which
is measured in pounds per ton) refers to rolling, flange, axle, track, and air resistance on level tangent
track. Curve resistance represents the additional work needed to overcome wheel-rail friction in curves.
Without lubrication, curve resistance is assumed to be 0.8 pounds/ton per degree of curvature (AREMA,
2008). Grade resistance is 20 pounds/ton per percent of grade. A general model of train resistance is
(1) R = A + BV + CV2
Where
R = Train resistance in pounds per ton
V = Train speed or velocity in mph
A = Resistance component that is independent of speed (e.g., axle resistance)
B = Resistance that varies with speed
C = Resistance that varies with the square of speed
Axle resistance is primarily reflected in A, while B reflects flange friction and dynamic flange impacts,
which increase with speed. Rolling and track resistances are reflected in A and B. C is a streamlining
coefficient that captures many aerodynamic effects such as frontal air pressure, rear drag, the swirling of
air in open top cars, and turbulence between cars. Curve and grade resistance are added to the value of R
to derive total train resistance (TR) for a specific segment.
Several resistance equations are used in train performance simulators, including the Davis equation, the
modified Davis equation, and the Canadian National (CN) equation. These equations are defined in
Volume 4, Chapter 16, Section 2.1 of the Manual for Railway Engineering, published by the American
Railway Engineering and Maintenance of Way Association (AREMA, 2008). According to this manual,
the CN formula “has given reliable results in train performance calculator programs or similar
applications” (AREMA, 2008, p. 16-2-4). Therefore, it is used in this paper to analyze a broad mix of
equipment types. In the CN formula (which is shown in Equation 2), B is equal to 0.03.
(2) = 1.5 + 18
+ 0.03+
10000
6
Where
N = Number of axles
W = Total weight in tons of locomotive or car
a = Cross-sectional frontal area of vehicle in square feet
The streamline coefficient (C) is 24 for a lead freight locomotive. However, it drops to 5.5 for trailing
power units and to 5.0 for a mixture of freight cars. The C coefficients and dimensions of the equipment
analyzed in this study are shown in Table 3.1. These values can be used to estimate train resistance for
unit trains of like cars—i.e., coal, grain, COFC double stack, and multilevel (“autorack”) trains.
Understandably, empty open top cars generally have the poorest aerodynamics. The aerodynamic
resistance of intermodal trains varies with the composition. Uniform double-stack trains are
aerodynamically superior to intermodal trains with mixtures of containers and trailers, and to trains with
unfilled slots—i.e., mixed double-stack and single container-on-flatcar movements.
Table 3.1 Typical Values of C and Frontal Areas for Freight Equipment
Type of Equipment
C*
Area (Sq. Ft.)
Gondola (loaded)
4.2
105
Gondola (empty)
12.0
105
Covered Hopper
7.1
125
Loaded Intermodal Flatcar (average)
5.0
125
Enclosed Multilevel Flatcar (autorack)
7.1
170
Leading Freight Locomotive
24.0
160
Source: American Railway Engineering and Maintenance of Way Association. Manual for Railway Engineering, 2008.
Volume 4, Chapter 16, Section 2.1.
*Streamline coefficient
3.2 Estimating Procedures
3.2.1 Car and Locomotive Resistance Factors
Calculating fuel consumption from train resistance models is a multi-step process. In the initial step,
resistance factors for locomotives and freight cars on tangent level track are computed for a specific
velocity. Two examples are shown in Table 3.2 — a lead locomotive and a loaded gondola car. Both are
traveling at 40 mph. The resistance of the loaded car is approximately 3.7 lb/ton, while the resistance of
the empty car is 15.2 lb/ton. The resistance of the locomotive is 6.4 lb/ton. In the final step of the
calculation, grade and curve resistance are added to the initial train resistance to arrive at the total
resistance for the route. For example, the calculation of a total resistance factor (TR) for a 5-degree curve
is illustrated in Table 3.2.
Figure 3.1 illustrates the relationships between velocity (in miles per hour) and resistance (in pounds per
ton) for empty gondola cars, loaded gondola cars, and double-stack flatcars. The resistance factor for a
loaded gondola car increases from 2.95 lb/ton at 25 mph to 4.92 lb/ton at 60 mph. Over the same interval,
the resistance of an empty gondola car increases from 9.1 to 27.3 lb/ton, while the resistance of a double-
stack car increases from 3.8 lb/ton to 7.5 lb/ton.
7
Table 3.2 Example of Train Resistance Factors for Lead Locomotive and Gondola Car
Input/Result
Locomotive
Loaded Car
Empty Car
W (tons)
195
143
21.9
N (axles)
6
4
4
a (cross-section frontal area)
160
105
105
V (velocity in mph)
40
40
40
B
0.03
0.03
0.03
C
24
4.2
12
Resistance (R from Equation 2)
6.4
3.7
15.2
Degree of curvature
5
5
5
Curve factor (lb/ton)
0.8
0.8
0.8
Curve resistance per ton
4
4
4
Total resistance (TR)
10.4
7.7
19.2
Figure 3.1 Variations in Resistance with Velocity for Select Car Movements
3.2.2 Locomotive Requirements
Once the resistance factors are calculated, the number of locomotives required for the movement is
estimated. There are six major steps in this process. (1) The resistance factors for the freight car and
locomotive at the fastest (target) speed are determined. (2) The tractive effort of the locomotive (i.e., its
propulsive force in pounds) is computed for the same speed as 308 × horsepower / speed (in miles per
hour). However, the locomotive’s tractive effort is constrained to the adhesion limit. (3) The tractive
effort required to overcome the locomotive’s own resistance is computed. (4) From this result, the
drawbar pull of the locomotive is estimated—i.e., the tractive effort available to move the train after the
tractive effort needed to move the weight of locomotive is considered. (5) Using the drawbar pull and the
resistance factor of the freight car, the number of trailing tons that can be moved by the locomotive (e.g.,
its tonnage rating) is computed. (6) In the final step, the tonnage rating is divided into the trailing tons of
0
5
10
15
20
25
30
25 30 35 40 45 50 55 60
Resistance (lb/ton)
Miles Per Hour
Empty Gondola Double Stack Loaded Gondola
8
the train (e.g., the total weight of all freight cars) to arrive at the number of locomotives needed for the
speed, equivalent grade, and train composition.
The calculations needed to determine the locomotive requirements of a train of loaded 143-ton gondola
cars are illustrated in Table 6 for a 0.25% grade and a velocity of 25 mph. The illustrations are based a
locomotive with 4,400 horsepower (hp) and 82% efficiency. The total resistance of the freight car (in
pounds per ton) is shown in line 5 of Table 3.3, while the total resistance of the locomotive (in pounds) is
calculated in lines 6-10. The tractive effort of the locomotive is computed in line 12 as 308 × 4400 / miles
per hour (mph). The drawbar pull of the locomotive is shown in line 13, while the tonnage rating is
computed in line 14. In this example, three 4,400-hp locomotive units will be needed if the target speed
on the grade is 25 mph.
Table 3.3 Illustrations of Locomotive Tonnage Ratings and Power Requirements
Line
Factor
Denomination/
Source
Input or
Computed Value
1
Target Speed
Mph
25
2
Equivalent Grade
Percent
0.25
Freight Car Resistance
3
Rolling Resistance
lb/ton
2.95
4
Grade Resistance / Ton
20*L2
5
5
Total Resistance/ Ton
L3+L4
7.95
Locomotive Resistance
6
Rolling Resistance
lb/ton
4.03
7
Locomotive Weight
Tons
195
8
Locomotive Rolling Resistance
L6*L7
786.8
9
Locomotive Grade Resistance
20*L2*L7
975
10
Total Locomotive Resistance
L8+L9
1,762
11
Locomotive Horsepower
per unit
4,400
12
Tractive Effort
308*L11/L1
54,208
13
Locomotive Drawbar Force
L12-L10
52,446
14
Tonnage Rating per Unit
L13/L5
6,600
15
Car Weight
Tons
143
16
Cars per Train
Units
110
17
Trailing Tons
L15*L16
15,730
18
Locomotives Units
Ceiling(L17/L14)
3
3.2.3 Total Train Resistance
Once the locomotive requirements are estimated, the resistance of the entire train (including the
locomotives) is calculated. First, the total locomotive resistance is computed for the leading and trailing
units. This calculation considers the fact that the aerodynamic resistance of the trailing units is less than
the lead. Once computed, locomotive resistance is then added to the total resistance of all freight cars to
derive the total resistance for the train, including grade and curve resistance. The next step is to determine
the energy required to overcome total resistance, and convert these requirements to gallons of fuel.
Initially, railroad energy requirements are expressed in horsepower-hours (hp-hr).
9
3.2.4 Horsepower-Hours and Gallons of Fuel Consumed
The energy required to move each ton of the loaded gondola car described in Table 3.2 one mile over
straight level track at 40 mph is 3.7 lb × 5,280 ft / 1.98E6 ft-lb = .01 hp-hr per ton-mile. In this
expression, 1.98 million ft-lb is a measure of the work performed during a horsepower-hour (i.e., 33,000
ft-lb per minute times 60 minutes). Continuing this example, it requires 1.41 hp-hr to move a loaded
gondola car one mile (0.01 × 143) and 155 hp-hr to move a train of 110 gondola cars one mile. The hp-hr
needed to overcome grade and curve resistance are derived in a similar manner. In the final step of the
calculation, hp-hr is converted to gallons of fuel using conversion factors published by the U.S.
Environmental Protection Agency (EPA, 2009). These conversion factors are 20.8, 18.2, and 15.2
horsepower-hours per gallon for large line-haul locomotives, small line-haul locomotives, and switching
locomotives, respectively.
3.3 Uniform Train Simulations
The methods described previously are used to estimate gallons of fuel consumed and RTM/G for uniform
trains of like cars. The car weights used in the analysis are shown in Table 3.4. The first scenario consists
of a coal gondola car with an empty (tare) weight of 23 tons and a gross weight of 139 tons. The second
scenario involves a grain hopper car with a gross weight of 138 tons and a tare weight of 31 tons. The
third scenario reflects a double-stack COFC movement of articulated cars. The weights shown in Table
3.4 correspond to one unit of an articulated car (e.g., a well). The intermodal train consists of 100 such
units. The gross weight of the intermodal flatcar includes the weights of two marine containers
(approximately 9 tons). Each container is assumed to be loaded with manufactured goods weighing 12.5
tons per container. This load factor is representative of commodities such as computers and electronic
components, which have densities of 16 to 20 lb/ft3. It is also the average from the waybill sample (STB,
2008).
Table 3.4 Equipment Types and Weights Used in Uniform Train Simulations
Equipment Type
Tare
Net
Gross
Coal Gondola Cars*
23
116
139
Covered Hopper Cars*
31
107
138
Intermodal (COFC) Flatcars*
31
25
56
Multi-Level Flatcars*
53
20
73
* Values computed from the 2008 waybill sample (STB, 2008)
In these scenarios, the railcars are assumed to travel the average number of empty miles for the car type.
For example, intermodal flatcars average 13% empty miles, while the empty-loaded ratio of a covered
hopper car is approximately 1.0 (STB, 2009). The fuel consumed during the empty movement is reflected
in each calculation. The analysis entails six major steps: (1) determine the number of locomotives needed
for the train’s cruising speed, (2) construct a speed profile based on track speed limits, (3) estimate the
fuel consumed at each steady-state speed corresponding to a speed limit, (4) estimate the additional fuel
consumed by accelerating from limit to limit within the profile, (5) estimate the fuel consumed during
switching at origin and destination, and (6) include fuel consumed during idling or waiting time as a
result of traffic control delays and/or congestion. A set of observed train speeds serves as a starting point
for the analysis.
10
3.3.1 Average Train Speeds
Class I railroads publish weekly performance measures on their websites, including the average speeds of
various types of trains. The average speeds for a 95-week period starting in January of 2009 and ending in
May of 2011 are shown in Figure 3.2. These speeds (which represent the average of all seven Class I
railroads) reflect train priorities based on commodity values. The average speeds for container and auto
trains are the highest, while coal trains are the slowest.
Figure 3.2 Average Train Speeds: Jan. 2009-May 2011
3.3.2 Train Speed Profile
While average speeds are insightful, train resistance varies in a non-linear manner with speed (Equation
2). For this reason, deviations in fuel consumption attributable to speed variations are important.
Although the speed limits of individual track segments are not published in a central directory, the grade
crossing inventory (which includes 65,535 public crossings) lists the maximum timetable speed of the
track segment on which the crossing is located. While train speeds through grade crossings may be
restricted to less than the timetable speed, the inventory provides a cross-section of railroad speed limits
throughout the United States. According to the inventory, only 3% of track segments have timetable
speeds of 80 mph or greater. For trains to travel at these speeds, an advanced track-train communication
system must be installed that allows distant signals to be displayed in the cabs of locomotives.
Because the intent of this analysis is to focus on movements that reflect a large percentage of the train
population, a speed profile is developed that excludes segments with speed limits > 79 mph. As shown in
Figure 3.3, 14.6% of the remaining segments have speed limits of 79 mph. Other important limits are
evident in the figure. The peaks centered at 10, 25, 40, and 60 mph denote the maximum speeds of FRA
track classes 1-4, respectively. Similarly, the peak at 49 mph is a function of regulation. For trains to
operate at speeds of 50 mph or faster, the line must be equipped with a block signal system.
21.4 22.7 23.1
25.8
31.2
0
5
10
15
20
25
30
35
Coal Unit Grain Unit Manifest Multilevel Intermodal
Miles per Hour
Train Type
11
Figure 3.3 Profile of Railroad Speed Limits in the United States
Computed from the National Grade Crossing Inventory by including segments with at least one main track
and one through train per day with a maximum timetable speed ≤ 79 mph
For reasons illustrated in Figure 3.1, freight trains may be restricted to speeds beneath the limit when
incremental locomotive and fuel requirements become prohibitive. This is especially true of heavy bulk
trains of low-value commodities, for which additional speeds are relatively unimportant. To reflect cost-
effective operations, maximum practical cruising speeds are estimated that result in realistic assignments
of locomotive power (Table 3.5). Increasing these speeds significantly will result in the allocation of
additional locomotives to the trains.
3.3.3 Fuel Consumed During Acceleration
Train acceleration is a function of horsepower, tractive effort, and train resistance. The additional tractive
effort required to accelerate a train from an initial velocity (Vi) to a final or new velocity (Vf) is shown in
Equation 3 (Hay, 1980).
(3) =95.6
Where
W= Weight of train in tons
V= Velocity in miles per hour (mph)
A= Rate of acceleration (mph/sec)
Fa = Tractive effort in pounds required to accelerate from the lower to higher velocity
0
2
4
6
8
10
12
14
010 20 30 40 50 60 70 80
Percent
Limit (mph)
12
Table 3.5 Cruising Speeds for Unit Trains on Flat Terrain Based on Locomotive Requirements
Train
Top Cruising
Speed (mph)
Number of 4,400-hp
Locomotives
Coal gondola (empty) – 120 cars
45
2
Coal gondola (loaded) – 120 cars
45
3
Covered hopper (empty) – 110 cars
47
2
Covered hopper (loaded) – 110 cars
47
3
Double-stack COFC (empty) – 100 cars
75
2
Double-stack COFC (loaded) – 100 cars
70
4
Multilevel autorack (empty) – 50 cars
45
1
Multilevel autorack (loaded) – 50 cars
60
2
From Equation 3, it can be seen that 95.6 pounds of force are required to accelerate one ton one mile per
hour per second. Since Vf – Vi = ΔV, and ΔV/t is the rate of acceleration (A) in mph/sec, Equation 3 can
be reduced to 95.6WA. It follows from Equation 3 that =/95.6. However, this formula is of no
practical use because train resistance varies with velocity. For this reason, acceleration cannot be
represented as a constant. Instead, the distance and time required to accelerate from Vi to Vf must be
calculated in an iterative (computational) manner for small mile-per-hour intervals (e.g., 0.25 mph), and
the fuel consumed during each acceleration interval must be summed to arrive at the total fuel consumed
during the acceleration event. A speed-distance curve for a 110-car grain train calculated in this manner is
depicted in Figure 3.4.
Figure 3.4 Acceleration Curve for Unit Grain Train
3.3.4 Origin-Destination Switching
According to the Surface Transportation Board (2009) the average industry switching time in the United
States ranges from 7.82 minutes per car in the east to 5.24 minutes per car in the west. However, average
switching times are much lower for unit train and large multicar shipments. Because a true unit train is
switched as a single block, the switching time per car is 25% of the system average (STB 2009). Because
large multicar blocks necessitate fewer switching moves or cuts, the average multicar switching time is
50% of the system average (STB 2009).
0
0.5
1
1.5
2
2.5
3
3.5
010 20 30 40 50
Miles
Speed (mph)
13
Container and automobile shipments are trainloads rather than unit trains. A trainload consists of several
blocks of cars that are integrated into a single train at a marshaling or classification yard near the origin
(e.g., a port or automobile manufacturing plant). At the marshaling yard, several blocks of cars are
switched individually as the train is assembled. As this description suggests, the switching efficiencies of
container and automobile shipments are more consistent with multicar operations than with unit-train
loading. Nevertheless, once assembled, the container and auto trains move from origin to destination with
little or no switching en route.
With the exception of intermodal trains, the spotted/pulled ratio is 2.0—i.e., the switching of a loaded car
necessitates the switching of an empty car before or after. Because switching operations occur in the
lower throttle positions (e.g., 2-5), a weighted fuel consumption rate of 35 gallons per hour is used. Based
on this factor, the estimated gallons of fuel consumed for each train are shown in Table 3.6, using the
greater of the two regional switching averages.
Table 3.6 Estimated Fuel Consumption at Origin (O) and Destination (D) for Unit Trains
Line
Input or Calculated Value
Source
Train Type
Grain
Coal
COFC
Auto
1
Average Minutes per Car
STB
7.82
7.82
7.82
7.82
2
Percent of Avg. Switching Time
STB
0.25
0.25
0.50
0.50
3
Adjusted Minutes per Car
L.1 x L.2
1.955
1.955
3.91
3.91
4
Cars per Train
Assumed
110
120
100
75
5
Hours per Switch: O or D
(L.3 x L.4)/60
3.6
3.9
6.5
4.9
6
Gallons per Hour
Assumed
35
35
35
35
7
Gallons per Switch: O or D
L.5 x L.6
125.4
136.9
228.1
171.1
8
Spotted/Pulled Ratio
STB
2
2
1
2
9
Gallons Consumed: O and D
L.7 x L.8 x 2
501.8
547.4
456.2
684.3
3.3.5 Average Trip Distance and Cycle Length
The origin and destination fuel estimates shown in Table 3.6 are allocated to shipments based on the
average loaded trip distance for each train. These values (which are shown in column 2 of Table 3.7) are
computed from the waybill sample (STB 2008). Fuel consumption during acceleration is allocated to
shipments based on an assumed cycle distance (column 3, Table 3.7). Each cycle reflects acceleration of a
train through the speed profile from 0 mph to cruising speed, and deceleration from cruising speed to 0
mph again.
Table 3.7 Average and Train Cycle Distances
Train
Average Shipment Distance (miles)1
Cycle Length (miles) 2
Coal
743
50
Grain
1,013
75
Containers
1,007
250
Autos
911
150
1 Computed from 2008 waybill sample
2
Based on approximate distance between rail yards and train priorities between yards
14
The basis for the cycle is the approximate distance between rail yards or classification points—e.g., 250
miles. Each train must either slow to pass through a classification yard at reduced speed, stop for fuel, or
stop to change crews. In between yards, lower-order trains are diverted to side tracks to allow trains with
higher priorities to pass. The typical order of priorities is (1) intermodal and passenger, (2) auto, (3) grain,
and (4) coal. In addition, manifest trains of mixed freight may take priority over coal trains. The cycle
lengths of lower-order trains reflect the combined probability that a train will be bumped by higher-
echelon trains. For example, a coal train may be diverted to a side track by a container, passenger, auto, or
manifest train.
3.3.6 Estimated Speeds and Idle Time
The train speeds estimated using the methods and assumptions described above are shown in column 3 of
Table 3.8. After rounding, the predicted average speed for intermodal trains matches the observed average
speed of 31.2 mph (column 2). However, the predicted speeds of other trains are significantly greater than
the observed speeds. This is because time spent waiting on side tracks while other trains pass is not
reflected in the predicted speeds. On high-traffic lines, a lower-echelon train may be delayed for some
time while higher-priority trains move through the subdivision. For auto, grain, and coal trains, the
differences between the predicted and observed speeds are assumed to reflect waiting times due to traffic
control delays. Fuel consumed during these intervals is estimated using an idling rate of 4 gallons per
hour.
Table 3.8 Predicted and Observed Train Speeds in Miles per Hour in Flat Terrain
Train Type
Observed Average Speed
Predicted Average Speed
Coal Unit
21.4
25.6
Grain Unit
22.7
26.9
Intermodal
31.2
31.2
Auto
25.8
29.5
3.3.7 Fuel Consumed During Drayage
Automobile and container shipments often require drayage at origin and/or destination. For example, an
imported container that terminates in a Chicago rail yard is delivered by truck to its final destination.
Similarly, automobiles are drayed from a destination rail yard to dealers in the surrounding area. Because
of these linked movements, drayage fuel requirements are estimated for automobile and containers
shipments.
In the drayage analysis, each double-stack flatcar is equivalent to two individual trucks. Additional empty
movements are assigned to COFC shipments to allow for the return of empty containers to the rail yard or
port area. Drayage trucks are assumed to operate in shuttle service, returning to the yard for the next load
until the train is empty. However, trucks may occasionally backhaul containers from industry locations to
yards. In ports with on-dock rail access, origin or destination drayage may be unnecessary for import or
export containers. Given these considerations, average drayage hauls of 10 and 20 miles are assumed for
automobile and container shipments, respectively. Drayage is assumed at both origin and destination.
The assumptions and calculations in the drayage analysis are shown in Table 3.9. In the final calculation,
484 and 1,935 gallons of drayage fuel are added to the total energy requirements of automobile and
container trains, respectively. However, no such adjustments are made for coal shipments, because
loading and unloading typically occur at or near the mine and utility. Similarly, the fuel consumed in
moving grain from farms to elevators is not considered because the farm-to-elevator movement occurs
regardless of whether the grain moves from the elevator to market by rail, truck, or waterway.
15
Table 3.9 Calculation of Drayage Fuel for Automobile and Container Trains
Line
Input or Calculated Value
Source
Train
Auto
Container
1
Drayage Events per Loaded Move (Orig. & Dest.)
Assumed
2
2
2
Ratio: Total to Loaded Truck Movements
Assumed
2
1.5
3
Truck/Railcar Ratio
Assumed
1.5
2
4
Empty & Loaded Truck Movements per Railcar
L.1 × L.2 × L.3
6
6
5
Average Distance of Truck Movement (Mi.)
Assumed
10
20
6
Truck Miles per Railcar
L.4 × L.5
60
120
7
Average Truck Fuel Efficiency (Mi. per Gallon)
Assumed
6.2
6.2
8
Gallons of Fuel per Railcar
1/L.7 × L.6
9.7
19.4
9
Cars per Train
Assumed
50
100
10
Drayage Fuel per Train (Gallons)
L.8 × L.9
484
1,935
3.3.8 Grade Profile
Railroad grades vary widely and there is no public file containing the gradients of line segments in the
United States. However, an elevation profile of the Pacific Northwest (PNW) is used to illustrate the
effects of grades on fuel consumption (Figure 3.5). In this profile, trains moving from Seattle to the
interior must cross two mountain ranges (the Cascade and the Rocky Mountain ranges). The grades on
these routes are 2.2% and 1.8% respectively, making them severe gradients in terms of railway
transportation. Because the detailed route profiles are unknown, a simplified method is used to analyze
incremental energy requirements.
Figure 3.5 Elevation Profile in the Pacific Northwest Region of the United States
0
1
2
3
4
5
6
Eelvation (1,000 ft)
16
It takes 200,000 ft-lb of work to lift one ton 100 ft. As noted earlier, 1.98 million ft-lb = 1 horsepower-
hour. Thus, 0.10 hp-hr are needed to lift one ton 100 ft in elevation. These incremental energy
requirements are independent of speed and track profile.
A train traveling from Seattle to the interior will rise approximately 2,800 ft in elevation by the time it
reaches the Cascade Tunnel. From there, it drops to an elevation of roughly 650 ft at Wenatchee, WA.
However, the gradient is so steep that most or all of the potential energy of the train at the top is lost when
braking down the mountain, dissipated as friction heat. From Wenatchee, the train begins to rise again as
it approaches the Rocky Mountains. After a beneficial negative grade between Harrington, WA, and
Bonner’s Ferry, ID, (where some of expended energy is recovered), the train begins to rise again and
climbs an additional 3,433 ft to the crest of Marias Pass in Montana. On the eastbound route, the total rise
in train elevation with no energy recovery on the downhill gradients is approximately 6,400 ft. However,
the route is kinder to westbound movements because of the beneficial negative grade from Harrington to
Wenatchee. In this case, the total rise in train elevation without offsetting recoveries is 4,400 ft.
Loaded coal and grain trains are assumed to move west along this route, while empty trains travel east.
The opposite flow is assumed for 75% of the auto and container trains, which are assumed to represent
imports via Seattle which are destined for interior locations. Based on these assumptions, the incremental
fuel consumption due to changes in elevation along this route are 4,704; 4,545; 1,892; and 1,407 gallons
for coal, grain, COFC, and auto trains, respectively.
3.3.9 Predicted Results
Based on the methods and assumptions previously described, a uniform coal train is expected to achieve
976 RTM/G on flat terrain (Column 2 of Table 3.10), while a grain train generates 862 RTM/G. In
comparison, an intermodal train yields 311 RTM/G. A train of automobile transporters experiences the
worst fuel economy, realizing only 177 RTM/G on flat terrain.
Table 3.10 Predicted RTM/G of Uniform Train Movements
Train
Flat Terrain
PNW Route Profile
Auto
177
139
Intermodal
311
252
Grain Unit
862
649
Coal Unit
976
676
As shown in column 3 of Table 3.10, the same coal train is expected to achieve 676 RTM/G on the PNW
route, after the additional fuel needed to overcome lifting resistance is added to the baseline fuel
consumption for the trip. Similar estimates are shown for auto, intermodal, and grain trains.
17
3.4 Summary and Limitations of Analytical Method
Unlike previous studies, the methods used in this analysis have been explicitly documented and variations
in speed limits and origin-destination switching have been considered. However, there are limitations to
the analysis. (1) The effects of track curvature on speeds and resistance are not considered. In territories
with substantial curves, trains may be unable to attain the same speeds as on tangent track. Alternatively,
maintaining the same speeds in curved sections will result in additional fuel consumption. Either way, an
effect is missing. (2) While realistic, the PNW route profile is one of many in the United States and may
overstate or understate the energy requirements of other routes. (3) The effects of interim grades and track
profiles are not considered, including train momentum and velocity head. (4) Simplified assumptions are
made regarding the percentages of potential energy lost to braking on downhill grades.
While the analytical method is useful, it cannot practicably be used to develop a comprehensive picture of
railroad fuel efficiency via selective application. Moreover, estimates generated from this method are not
based on observed data. Rather, they are generated from deterministic resistance equations which,
themselves, are generalizations of a vast number of contextual and operational factors that may affect the
resistance of a train. For these reasons, an alternative method is pursued.
18
19
4. STATISTICAL MODEL OF RAILROAD FUEL CONSUMPTION
A regression model estimating railroad fuel consumption is described in this part of the paper and reflects
differences in train services and terminal operations. The model is estimated from 24 years of data
derived from R-1 reports submitted by Class I railroads to the U.S. Surface Transportation Board from
1985 through 2008. The database includes the reported gallons of fuel consumed for freight purposes by
each Class I railroad, as well as the gross ton-miles and revenue ton-miles of operations. While many
railroad mergers have occurred during this period, railroads can be organized into three geographic
regions that have remained constant over time: East, Central, and West. Each railroad in the database is
assigned to one of these regions (Table 4.1).
Table 4.1 Class I Railroads and Geographic Regions in the United States
Railroad
Code
Region
Atchison, Topeka, & Santa Fe
ATSF
West
Burlington Northern
BN
West
Burlington Northern-Santa Fe
BNSF
West
Chicago & Northwestern
CNW
West
Conrail
CR
East
CSX Transportation
CSX
East
Grand Trunk Corporation
GTC
Central
Grand Trunk Western
GTW
Central
Illinois Central Gulf
ICG
Central
Kansas City Southern
KCS
Central
Norfolk Southern
NS
East
Soo Line
SOO
Central
Southern Pacific
SP
West
Union Pacific
UP
West
While revenue ton-miles are reported for each railroad, no distinctions are made among types of service.
However, gross ton-miles of cars and contents are reported by train type. As noted earlier, GTMC include
the revenue ton-miles of cargo, as well as the tare ton-miles of the cars, trailers, and containers needed to
transport the cargo. The overall relationship between GTMC and fuel consumption is illustrated in Figure
4.1, in which each square represents an individual railroad and year. As the trend line suggests, there is a
strong linear relationship between fuel consumption and GTMC. Indeed, a simple regression of gallons of
fuel against GTMC yields an R-Square of 0.986.
4.1 Model Formulation
4.1.1 Main Explanatory Variables
Generally, railroads operate three types of trains: way, through, and unit. Single cars usually travel in way
trains at origins and/or destinations. Way trains operate primarily between branch-line stations and
railroad yards, stopping frequently to drop off and pick up cars en route. Through trains typically move
from yard to yard, performing only limited switching en route. According to the STB (2010), unit trains
are characterized by “shuttle-type service in equipment (railroad or privately owned), dedicated to such
service, moving between origin and destination.” As the definition suggests, unit trains do not require
intermediate yard switching. Moreover, as detailed in 3.3.4, unit train origin-destination switching is very
efficient.
20
Figure 4.1 Plot of Gallons of Fuel versus Gross Ton-Miles of Cars and Contents
In 2008, way train activity comprised less than 3% of the gross ton-miles of cars and contents in the
United States, while through trains accounted for approximately 55% of GTMC. The movement of a car
in a way train typically precedes or follows a movement in a through train. Because way and through train
movements are linked and the percentage of way train GTMC is very small, these two categories are
combined to form “non-unit train” GTMC.
A non-unit train movement may consist of individual carloads, blocks of carloads moving to the same
destination, or trainloads. The distinguishing characteristic is that some marshaling or gathering of cars is
required at origin and/or destination, where car blocks may arrive or depart in way trains, traveling to or
from nearby industry locations. In port areas, cars loaded with import or export goods may be shuttled
between docks and inland classification yards. According to the STB (2009), the typical way-train trip is
14 miles in the western United States and 25 miles in the east.
4.1.2 Model Statement
Based on train definitions, a multiple regression model is formulated in which the dependent variable is
gallons of fuel and the primary independent variables are unit train and non-unit train GTMC. The general
form of the model is GAL = f(UGTMC, NGTMC, T, REG). Each variable is described in Table 4.2, along
with its source—i.e., the schedule of the R-1 report from which the measure is derived. The specific form
of the model is
(4) =+ + ++++
y = 1.3164x + 26.718
R² = 0.9856
0
200
400
600
800
1000
1200
1400
1600
0200 400 600 800 1000 1200
Gallons of Fuel (millions)
GTMC (billions)
21
Table 4.2 Variables in Railroad Fuel Regression Model
Variable
Description
Source
Type
GAL
Gallons of fuel consumed in freight service during
the year
R-1, 750
Response
UGTMC
Thousands of gross ton-miles of cars and contents
moved in unit trains
R-1, 755
Structural
NGTMC
Thousands of gross ton-miles of cars and contents
moved in non-unit trains
R-1, 755
Structural
T
Time in years before 2008 (2008 = 0)
N/A
Control
REG
A set of regional variables {1, 2, 3}
N/A
Indicator
In Equation 4, i denotes the region, j the year, and eij the error term. The predicted response is gallons of
fuel consumed in region i during year j, where j ranges from 1 to 24 and i from 1 to 3. There are three
possible combinations of values for REG1 and REG2: REG1=1 and REG2=0, REG1=0 and REG2=1, and
REG1=0 and REG2=0. The last combination defines Region 3.
Fuel consumed in freight service includes both train and yard fuel. Thus, the dependent variable reflects
three distinct operational components: (1) fuel consumed in train running or line-haul operations, (2) fuel
consumed in yard switching, and (3) fuel consumed in train switching. The latter activity occurs when
cars are switched at industries or tracks located outside of classification yards. While all trains perform
some switching, most origin-destination switches are attributable to way and unit trains. Way trains
switch cars at dispersed industries in small blocks, often less than five cars at a time. On the other hand,
unit trains switch cars at one origin and destination, usually in large quantities—e.g., 100 cars or more.
Because unit trains are characterized by very efficient origin-destination switching and little or no yard
switching, the variable UGTMC is expected to have a smaller positive coefficient than NGTMC.
4.1.3 Time Variable
Time (T) represents the number of years prior to 2008. For example, t=1 for 2007 and 23 for 1985. This
variable is important because the relationship between fuel consumption and GTMC is expected to
change over time, as railroads purchase more fuel efficient locomotives and adopt energy saving
practices. Because of T’s inverse relationship to time, the variable is expected to have a positive sign.
Because t=0 in 2008, it vanishes from the equation when values are predicted for that year.
4.1.4 Regional Variables
Each region in the R-1 database is represented by an indicator variable—e.g., Region 3. When the
observation is for the western region, Region 3 equals 1. Otherwise, Region 3 equals zero. To avoid
singularity, only n – 1 indicator variables are included in the model. The signs and magnitudes of the
variables are interpreted in relation to the excluded effect, which is subsumed in the intercept. In a linear
model, indicator variables shift the intercept of the regression, creating separate predictive equations for
each level or classification.
The regional variables capture differences in terrain, geography, and networks that affect fuel economy.
Railroads in the central or plains region do not encounter mountains. Operating primarily in gentle terrain,
these railroads exhibit a decidedly north-south orientation. Heavy cars (such as grain traffic) move south
via negative gradients to the Gulf. This pattern (of loads moving downgrade and empties moving
upgrade) enhances fuel economy. In contrast, western railroads (which have east-west orientations)
encounter substantial grades while crossing the Rocky Mountains and coastal ranges. Eastern railroads
22
operate in the Appalachian Mountains. These geographic differences are reflected in the observed values
of GTMC per gallon (GTMC/G) listed in Table 4.3.
Table 4.3 GTMC/G and RTM/G by Region in 2008
Region
GTMC/G
RTM/G
East
779
431
Central
908
501
West
809
464
4.2 Regression Results
4.2.1 Key Model Properties
The model has 71 total and 66 error degrees of freedom (DF), which should be sufficient to realize large
sample properties (Table 4.4). The R-square of 0.998 suggests that the linear model explains almost all of
the variation in fuel consumption. The coefficient of variation of 3.45% (which is computed as the
standard error of the regression divided by the mean of the dependent variable [gallons of fuel] multiplied
by 100) suggests that the model provides a very precise fit.
Table 4.4 Key Model Properties and Indicators
Observations
72
Error Degrees of Freedom
66
F Value
6,736
Prob. > F
<.0001
Coefficient of Variation
3.45
R-Square
0.9980
Adjusted R-Square
0.9979
4.2.2 Parameter Estimates and Standard Errors
The estimates from the fuel model and their corresponding standard errors are shown in Table 4.5. As
shown in column 3, the standard errors of the variables are small in relation to the estimated values. This
is a desirable outcome. However, the standard errors may be suspect unless the variance of the regression
is consistent over the entire range of the dependent variable.
23
Table 4.5 Parameter Estimates from Railroad Fuel Model
Variable
Parameter
Estimate
Standard
Error
t
Value
Prob.
> |t|
Heteroscedasticity Consistent
Standard
Error
t
Value
Prob.
> |t|
Column 1
Column 2
Column 3
Column 4
Column 5
Column 6
Column 7
Column 8
Intercept
713465017
58015819
12.30
<.0001
67216343
10.61
<.0001
UGTMC
0.74449
0.09710
7.67
<.0001
0.10696
6.96
<.0001
NGTMC
1.10767
0.09849
11.25
<.0001
0.11788
9.40
<.0001
T
3339195
1071017
3.12
0.0027
581239
5.74
<.0001
REG1
-476171943
29227568
-16.29
<.0001
29754620
-16.00
<.0001
REG2
-701468788
46887449
-14.96
<.0001
57404842
-12.22
<.0001
The issue of non-constant variance or heteroscedasticity is common in regression analysis. In most
instances, the form of heteroscedasticity is unknown and cannot be ascertained from the data. In such
cases, the variance is said to be inconsistent, meaning it is not a function of an independent variable and
does not increase or decrease monotonically. The regression coefficients (i.e., the parameter estimates)
are not biased by heteroscedasticity. However, there are two potential issues. (1) Regression coefficients
estimated from sample data may no longer be efficient (e.g., minimum variance estimators). (2) The
standard errors may be affected. As a result, hypothesis tests may be unreliable.
The first issue is not really a concern for this study because the parameters are estimated from population
data. Nevertheless, as recommended by Hayes and Cai (2007), heteroscedasticity-consistent errors are
used to assess the potential effects of inconsistent variance. These standard errors shown in column 6 of
Table 4.5) are computed under the assumption that the variance is not constant. A comparison of the t
values in columns 4 and 7 (which are computed by dividing the parameter estimates in column 2 by the
appropriate standard errors) shows only modest differences, suggesting only mild inconsistency. None of
the hypothesis tests are affected.
4.2.3 Probability Values and Inferences
In this study, the R-1 database constitutes the population of Class I railroads in the United States. Because
population data are available, sampling variability is not an issue. However, as railroad revenues change,
carriers may rise above (or fall below) the Class I revenue threshold, altering the population. For this
reason, it is beneficial to envision the R-1 database as a large sample of Class I railroads that do (or could)
exist. This visualization allows hypothesis tests that provide intuitive insights regarding the statistical
significance of effects.
For each variable, the null hypothesis is that the partial effect attributable to the variable is statistically
insignificant. For indicator variables, this means that the intercept shift attributable to the variable is not
significantly different from zero. For quantitative variables, the null hypothesis is that the partial slope
coefficients are not significantly different from zero. The probabilities (or p values) associated with the t
statistics in column 8 of Table 4.5 are all highly significant (i.e., values of less than .0001), indicating less
than a 1 in 10,000 chance of observing t values as large as those observed. The main predicted effects are
(1) after controlling for time and region, an increase of 1,000 unit-train GTMC results in the consumption
of 0.75 gallons of fuel, and (2) an additional 1,000 GTMC in other (non-unit) trains consumes 1.11
gallons of fuel (ceteris paribus).
The signs of the eastern and central regions are negative in relation to the west, which is characterized by
challenging grades and rough terrain. The parameter estimate of time is positive, meaning that the fuel
needed to transport a given quantity of gross ton-miles is less today than in previous years. There are
several reasons for this trend. (1) The computerization of locomotives has resulted in the optimization of
24
throttle settings, traction motor performance, and slip control. (2) The practice of distributed power (i.e.,
placing remotely controlled locomotive units at strategic locations throughout the train) has reduced slack.
(3) The manufacture of higher horsepower units has reduced the number of locomotives that must be used
to haul a given tonnage. With fewer locomotive axles pulling an equivalent weight, the axle resistance per
ton is reduced, as well as resistance posed by the locomotives’ total weights. (4) Emission control
standards have encouraged manufacturers to produce more fuel efficient and cleaner locomotives.
4.2.4 Test for Serial Correlation
Serial correlation is a potential issue in panel datasets that include time-series observations. In serial
correlation, the errors associated with particular time periods are related—e.g., the error term in the
current period (e.g., year) may be a function of the error term in the previous year. In regression analysis,
the Durbin-Watson (DW) test is used to detect potential first-order autocorrelation. This test compares the
residuals (or errors) of the regression at various times (t). If the difference between the errors for
observations t and t−1 is small in relation to the error at t, the errors may be correlated. As shown in Table
4.6, the DW test is not significant, supporting the notion that the error terms of the fuel model are not
correlated across time.
Table 4.6 Results of Test for Serial or Autocorrelation
Durbin-Watson Test Statistic
1.692
Prob. < DW
0.1177
Prob. > DW
0.8823
Number of Observations
72
1st Order Autocorrelation Statistic
0.124
Note: Prob. < DW is the p-value for testing positive autocorrelation, and Prob. > DW is the p-value for testing negative
autocorrelation.
4.3 Model Predictions
4.3.1 Level of Precision
To confirm its precision, the model is used to predict the values (i.e., gallons of fuel consumed) in each
region in 2008. Because t=0 in 2008, the predictive equation for the western region is simply b0 +
b1*UGTMC + b2*NGTMC = gallons of fuel. The alternative equations are b0 + b1*UGTMC +
b2*NGTMC + b4 and b0 + b1*UGTMC + b2*NGTMC + b5 for the eastern and central regions,
respectively. As shown in Table 4.7, the prediction errors are 0.2%, 1.7%, and 2.5% for regions 1-3,
respectively. The average prediction error for the three regions is less than 1.5%. This high level of
precision is due to the very high R2 and very low coefficient of variation.
Table 4.7 Predicted versus Observed Gallons of Fuel in 2008
Region
Gallons of Fuel Used
Predicted Gallons of Fuel
Prediction Error
East
1,027,912,283
1,029,539,077
0.16%
Central
212,893,103
209,225,598
-1.72%
West
2,645,378,417
2,712,295,930
2.53%
25
4.3.2 Marginal Estimates Derived from Coefficients
The coefficient of UGTMC represents the change in gallons per thousand unit train GTMC, holding all
else constant—i.e., the gross ton-miles of non-unit trains, time, and region. The coefficient of NGTMC
has an analogous meaning.
The gross ton-miles of cars and contents per gallon (GTMC/G) for each type of train are computed from
the regression parameters as 1 / bn * 1,000, where n takes values of 1 or 2, corresponding to UGTMC and
NGTMC, respectively. The resulting estimates are 1,343 and 903 GTMC/G for unit and non-unit trains,
respectively. Note that the predicted marginal fuel efficiencies are greater than the average efficiency of
806 GTMC/G for 2008.
The GTMC/G values are transformed into estimates of RTM/G for specific trains using the net-to-gross
car weight ratios shown in Column 2 of Table 4.8. For example, the net/gross ratio of a grain shipment is
computed as the gross car weight (138 tons), plus the tare weight (31 tons) times the empty return ratio
(1.0), divided by the net car weight (107 tons)—i.e., 107/(138+31 * 1.0)=0.63.
Table 4.8 Marginal Estimates of Revenue Ton-Miles per Gallon, by Train Type
Train
Net/Gross Weight
Ratio
RTM/G by Train Type
Non-Unit
Unit
(1)
(2)
(3)
(4)
Coal
0.71/0.72*
638
962
Grain
0.63
572
850
COFC
0.42
376
N/A
Auto
0.21
193
N/A
*Difference due to higher empty/loaded ratio for non-unit train movements
As shown in Table 4.8, the marginal fuel efficiency of a unit coal train added to the railroad network is
962 RTM/G, while the marginal fuel efficiency of a unit grain train is 850 RTM/G. These interpretations
presume that the addition of a single train will not substantially affect existing traffic on the network and
change operating conditions or the distribution of traffic among regions. These assumptions are realistic,
because the ratio of the additional GTMC of a single train to total GTMC is nearly zero. However, all fuel
consumed in providing freight services is not reflected in the marginal estimates. Specifically, fuel that is
not traceable to GTMC and regional variations in fuel efficiency are not considered. For these reasons,
average fuel consumption factors are also considered.
4.3.3 Average Fuel Efficiency Factors
Average fuel efficiency factors for each type of train and region are estimated from the model. For Region
3, the predictive equation for unit trains is (b0 + b1*GTMC) / GTMC, where GTMC is the total GTMC in
Region 3 during 2008. The predictive equation for non-unit trains is (b0 + b2*GTMC) / GTMC. Similarly,
the predictive equations for unit trains are (b0 + b1*GTMC + b4) / GTMC and (b0 + b1*GTMC + b5) /
GTMC for Regions 1 and 2, respectively. In each calculation, the fixed fuel requirement is added to the
variable fuel component computed from the observed GTMC. In the first set of calculations, the observed
GTMC is assumed to move in unit trains. In the second set of calculations, the observed GTMC is
assumed to move in non-unit trains.
26
The predicted values are validated by computing a weighted estimate of gallons per GTMC for each
region (Equation 5), and comparing these estimates to the observed regional averages. In these
calculations, the estimated gallons per unit train GTMC (Gal./UGTMC) are weighted by the proportion of
unit train GTMC (U), while the estimated gallons per non-unit train GTMC (Gal./NGTMC) are weighted
by the proportion of non-unit train GTMC (N).
(5) .
=.
+.
The mean difference between the predicted and observed values is 2% for the three regions. While the
predicted and observed values are very close, the GTMC/G estimates are scaled to the observed values in
each region using the ratio shown in Equation 6. The results are depicted in Table 4.9, along with the
observed values for each region.
(6) .
.
Table 4.9 Predicted and Observed GTMC/G Values by Train Type and Region
Region
Non-Unit Train GTMC/G
(Predicted)
Mean GTMC/G
(Observed)
Unit Train GTMC
(Predicted)
East (region1)
699
779
944
Central (region 2)
819
908
1,188
West (region 3)
697
809
931
Several points are noteworthy. (1) The average values are substantially lower than the marginal values of
903 and 1,343 GTMC/G for non-unit and unit trains, respectively. This is because all gallons of fuel
consumed in freight services are reflected in the average estimates. (2) The predicted GTMC/G estimates
are substantially greater in Region 2 than in Region 1 or 3. (3) The estimated fuel efficiencies of unit
trains are substantially greater than the average for all trains, irrespective of region. (4) The estimated fuel
efficiencies of unit trains are 25% to 31% greater than non-unit trains.
Average revenue ton-miles per gallon (RTM/G) are derived from GTMC/G using the net/gross ratios
shown in Table 4.8. The results are depicted in Table 4.10. Instead of truck drayage, way train drayage is
reflected in the auto and COFC estimates. As the table shows, the fuel efficiencies of coal unit trains
range from 667 to 851 RTM/G, while non-unit train coal shipments yield 493 to 579 RTM/G. The fuel
efficiencies of unit grain trains range from 589 to 752 RTM/G, while other grain shipments generate 446
to 524 RTM/G. Double-stack COFC and auto shipments average 310 and 155 RTM/G, respectively, for
all regions. However, the use of way train instead of truck drayage may substantially understate the total
fuel requirements of these movements.
Table 4.10 Estimated RTM/G in 2008, by Train Type and Region
Region
Train Type
Coal
Grain
COFC
Auto
Non-Unit
Unit
Non-Unit
Unit
Non-Unit
Non-Unit
East
494
676
443
598
291
149
Central
579
851
519
752
341
175
West
493
667
441
589
290
149
27
4.4 Comparisons of Predictions
In this section, estimates from the analytical method and statistical model are compared to each other and,
where possible, to estimates derived from other studies.
4.4.1 Grain Shipments
The estimated revenue ton-miles per gallon of a grain unit train using the analytical method are 862 and
649 RTM/G in flat and mountainous terrain respectively. In comparison, the marginal estimate from the
fuel model is 850 RTM/G, ignoring regional effects. The average fuel efficiency factors derived from the
statistical model are 598, 752, and 589 RTM/G for regions 1, 2, and 3 respectively.
The only real source of comparison is the Gervais and Baumel study of 1990. There, the fuel efficiencies
of 141-ton car movements in the western United States that cross the Rocky Mountains range from 585 to
664 RTM/G, while the maximum fuel efficiency of movements from Iowa to New Orleans is 802 RTM/G
(Table 2.2). Presumably, these are marginal values that exclude fixed and overhead fuel requirements. If
so, the most comparable estimates to the Iowa-New Orleans movement are the marginal (statistical) and
flat terrain (analytical) estimates, which are 6% to 7% greater than the 1990 Gervais-Baumel estimate. In
comparison, the average predicted efficiency of 752 RTM/G in the central Region (Table 4.10) is 6% less
than the 1990 Iowa-New Orleans simulation. This relationship is expected, because the latter estimate
excludes overhead fuel.
The analytical estimate of 649 RTM/G for the PNW route (Table 3.10) is close to the maximum value of
664 from the Gervais-Baumel study for movements from Iowa to Washington state. Moreover, the
average RTM/G in the western region predicted by the statistical model (589 RTM/G) lies within the
range of estimates from that study.
4.4.2 Auto Shipments
The average fuel efficiency of the three automobile shipments analyzed in the ICFI study (2009) is 159
RTM/G. According to the waybill sample (STB 2008), roughly 20% of the automobiles shipped via
railroads in the United States cross the Rocky Mountains. Using this factor, the weighted marginal fuel
efficiency derived from the analytical method is 169 RTM/G. The difference between the analytical
results and the ICFI simulations are largely explained by differences in lading weights. The average net
weight used in the analytical procedure (which was computed from the waybill sample) is 20 tons per car,
while the mean lading in the three ICFI simulations is 18.7 tons. In comparison, the mean of the average
efficiency ratings from the fuel model (Table 4.10) is 158 RTM/G. As these comparisons suggest, the
estimates from this study are quite consistent with those of the ICFI study. As noted earlier, ICFI
concluded that automobile shipments by rail are 1.9 to 2.2 times more fuel efficient than movements in
highway trailers.
28
4.4.3 Coal Shipments
Coal shipments are not analyzed in the ICFI study or in any recent study, for that matter. Therefore,
comparisons can only be made among estimates developed using different methods—i.e., analytical,
marginal, and average.
According to the waybill sample (STB, 2008), roughly 10% of coal shipments move within the Rocky
Mountain-Pacific Coast region. Using this percentage, the weighted estimate from the analytical method
is 946 RTM/G, as compared to the marginal estimate of 962 RTM/G from the model. The average
efficiency factors of 676, 851, and 667 RTM/G (from Table 4.10) reflect variations among regions due to
gradients, networks, and operational factors.
4.4.4 General Inferences
The marginal estimates from the statistical model are quite similar to the estimates from the analytical
procedure. However, only one elevation profile is reflected in the analytical model. More comparisons are
needed before it can be concluded that the two approaches yield essentially the same results.
Nevertheless, the initial comparison is encouraging.
The average fuel efficiency measures generated from the model are significantly less than the marginal
estimates. However, they indicate significant differences among regions and are consistent with observed
values. Moreover, the train efficiency factors within each region show substantial differences between
unit and non-unit train movements and agree very closely with observed values when weighted.
29
5. NATIONAL ANALYSIS OF RAILROAD FUEL EFFICIENCY
Thus far, railroad fuel efficiencies have been estimated for a limited number of movements by specifying
car tare and net weights. While these illustrations are realistic and descriptive of actual movements, they
represent only a few of the many potential net/gross ratios of movements in the United States and do not
necessarily reflect the average for any particular commodity. In this section of the paper, average fuel
efficiency factors are estimated for a broader set of movements using the coefficients of the regression
model and the 2008 railroad waybill sample.
5.1 Overview of Waybill Sample
The public waybill sample identifies commodity flows within and among five territories. Two of these
territories lie in the eastern region. Two lie in the central region. The fifth lies in the western region, and
includes the Rocky Mountains and Pacific Coast.
In the waybill sample, the commodity represented by each sample movement is identified by the Standard
Transportation Commodity Code (STCC). In addition to the origin and destination territories and the
commodity, the car type, net weight, and tare weight are identified for each sample movement, as well as
the expanded (or population) carloads represented by the movement. While the waybill sample has
limitations, it is the only comprehensive source of railroad movement information in the United States.
Using this sample, six intra- or inter-regional flows are analyzed: (1) flows within the eastern region, (2)
flows within the central region, (3) flows within the western region, (4) flows between the eastern and
central regions, (5) flows between the eastern and western regions, and (6) flows between the western and
central regions.
The waybill sample does not indicate whether a movement is a true unit train or a large multicar
shipment. In costing the waybill sample, the STB assumes that 50-car shipments have the characteristics
of unit trains. However, the minimum tariff volume for a shuttle train often exceeds 100 cars. For this
reason, a higher threshold is used. A shipment of 75 cars or more is assumed to move as a unit train that
requires no marshaling of cars at origin, or intermediate yard switching. This is obviously a blunt
delineation that mischaracterizes some shipments.
5.2 Empty/Loaded Mile Ratios
Although empty/loaded car-mile ratios cannot be derived from the waybill sample, they can be calculated
from the R-1 report for each car type and region, and merged with the waybill sample. The empty/loaded
ratios used in this analysis are shown in Table 5.1. By definition, a shuttle train has an empty/loaded ratio
of 1.0, since it cycles between an origin and destination.
30
Table 5.1 Empty/Loaded Car-Mile Ratios by Region
Car Code
Description
Region
East (1)
Central (2)
West (3)
37
Plain Boxcar
0.819
0.791
0.561
38
Equipped Boxcar
0.961
0.876
0.755
39
Plain Gondola
0.926
1.017
1.134
40
Equipped Gondola
0.931
1.044
0.953
41
Covered Hopper
0.976
1.044
1.006
42
Open Hopper – General Service
0.961
0.969
1.025
43
Open Hopper – Special Service
0.975
0.996
1.046
44
Mechanical Refrigerator
0.751
3.632
0.734
45
Non-mechanical Refrigerator
0.965
0.625
0.801
46
Flatcar – TOFC/COFC
0.133
0.099
0.132
47
Flatcar – Multi-Level
0.547
0.353
0.298
48
Flatcar – General Service
1.276
1.864
1.365
49
Flatcar – Other
0.990
1.026
0.994
50
Tank Car < 22,000 Gallons
1.018
1.074
1.072
51
Tank Car ≥ 22,000 Gallons
1.014
1.049
1.096
52
All Other Freight Cars
0.061
0.289
0.735
Computed from 2008 R-1 Reports
5.3 Estimation Procedure
The calculation process is essentially as follows: (1) The average net/gross weight ratio is computed for
each combination of commodity type, regional or interregional flow, and shipment type (e.g., a unit train
of coal moving within the eastern region). In each calculation, the tare weight of the car is multiplied by
(1.0 + the empty/loaded ratio). This interim value is added to the net weight of the car to derive the gross
weight per loaded mile. This value, in turn, is divided into the net tons to yield the net/gross ratio. (2)
Once calculated, the net/gross ratio is multiplied by the GTMC/G values (by train type and region) from
Table 4.9 to yield RTM/G estimates. For interregional flows, the empty/loaded and GTMC/G inputs are
averaged for the origin and termination regions. The results for major commodities which comprise at
least 1% of the total tons moved in 2008 are shown in Table 5.2.
5.4 RTM/G Estimates for Major Commodity Movements
As Table 5.2 shows, movements of coal, metallic ores, farm products, non-metallic minerals, and building
materials are the most energy efficient flows. In particular, railroads are very efficient in moving coal
from mines to utilities, averaging 722 RTM/G. This finding is important to national energy policy,
because the energy consumed in transporting coal to power plants must be subtracted from the energy
output of utilities to estimate net energy gains to the economy. Other key findings are: (1) Ore
movements, which are critical to steel-making and heavy manufacturing, generate 668 RTM/G.
(2) Movements of farm products and food and kindred products, which are critical to food supplies and
prices, average 543 and 411 RTM/G, respectively. (3) Movements of chemicals and allied products
average 430 RTM/G. In addition, railroads are an energy-efficient mode of transporting waste and scrap
materials (averaging 390 RTM/G). The fact that some of these materials are used in recycling or in place
of raw materials has an additional environmental benefit.
31
Table 5.2 Weighted-Average RTM/G for Major Commodity Groups
STCC
Product Description
Net/Gross
RTM/G
11
Coal
0.70
722
10
Metallic Ores
0.63
668
01
Farm Products
0.61
543
14
Nonmetallic Minerals, except Fuels
0.63
524
32
Clay, Concrete, Glass or Stone Products
0.62
459
28
Chemicals or Allied Products
0.57
430
33
Primary Metal Products
0.56
412
20
Food or Kindred Products
0.54
411
24
Lumber or Wood Products
0.54
397
40
Waste or Scrap Materials
0.54
390
29
Petroleum or Coal Products
0.51
385
26
Pulp, Paper or Allied Products
0.49
356
37
Transportation Equipment
0.24
177
Computed using the 2008 Waybill Sample
5.5 Validation of Modeling Process
The railroad modeling process described above produces very reasonable results. The overall (predicted)
net/gross ratio using the statistical model and 2008 waybill sample is 0.55. In comparison, the actual
net/gross ratio from the R-1 report is 0.57. The overall fuel efficiency rating predicted from the statistical
model and waybill sample is 456 RTM/G. In comparison, the observed 2008 fuel efficiency rating is 457
RTM/G. In effect, the predicted and actual values are nearly identical. As these comparisons suggest,
procedures based on the statistical model and waybill sample slightly underestimate the observed
net/gross ratios and efficiency ratings. However, on a national scale, the estimates are quite accurate and
do not overstate railroad fuel efficiencies. The next step in the analysis process is to estimate truck and
waterway fuel efficiency ratings so that railroad estimates can be placed in a multimodal context.
32
33
6. TRUCK FUEL EFFICIENCY MODEL
During the nine-year period from 2000 to 2008, the average fuel efficiency of combination trucks (i.e.,
tractors pulling trailers or semitrailers) ranged from 5.1 to 5.9 mpg (Figure 6.1). The average for the
period was 5.4 mpg. However, there is no apparent trend. Fluctuations in fuel economy during the period
may be due to factors other than engine technology and emission controls (i.e., demand, highway
congestion, tractor selection, and variations in truck travel patterns).
According to Kruse, et al. (2008), the average fuel economy of a heavy duty diesel truck with a gross
weight of more than 60,000 pounds is 6.2 mpg. This value is derived from the Environmental Protection
Agency’s MOBILE6 model. However, because a particular weight is not specified, the estimate may be
applicable to trucks weighing more than 60,000 pounds. In a comparative sense, the value derived from
the EPA model is greater than the largest observed value in recent history—i.e., 5.9 mpg in 2003 and
2004 (Figure 6.1)—and substantially greater than the average of 5.4 mpg for the 2000-2008 period.
While single efficiency factors are useful, a wide range of truck configurations are used to transport
freight in the United States. Because of grandfather clauses, the weights of many trucks exceed the federal
80,000-pound gross vehicle weight limit. In a 2004 study that focused on the uniformity of regulations in
the western United States, the U.S. Department of Transportation estimated fuel efficiency factors for a
wide range of trucks, including longer configuration vehicles (LCVs). The factors used in that study were
originally developed in the late 1990s and do not reflect the more stringent 2004 emission regulations of
the Environmental Protection Agency. While these values are outdated, the relative fuel efficiencies
should be similar to those that existed in 2008, since improvements in engine technology and fuel
economy are applicable to most (if not all) heavy duty tractors.
Figure 6.1 Combination Truck Fuel Efficiencies
Based on this assumption, a table of fuel efficiency factors is developed (Table 6.1) by taking the 2008
value from the EPA model (6.2 mpg) and attributing it to a five-axle tractor-semitrailer operating at a
gross weight of approximately 60,000 pounds. The rest of the table is filled in by assuming that the ratio
of the fuel economy of this truck to all other trucks has remained the same since the USDOT’s study. This
calculation is equivalent to indexing each value in the original table by 1.14 to account for improvements
in tractor fuel economy since the late 1990s.
4.4
4.6
4.8
5.0
5.2
5.4
5.6
5.8
6.0
2000 2001 2002 2003 2004 2005 2006 2007 2008
MPG
34
As shown in Table 6.1, fuel economy decreases as weight increases. According to Delorme et al., the fuel
economy of a five-axle tractor semitrailer increases by roughly 0.06% with each thousand pounds of
weight reduction (2009). However, fuel economy also varies with tractor performance and the
aerodynamics of truck configurations. Because of these variances, relationships among LCV and
semitrailer fuel economies are not based solely on weight.
Table 6.1 Average Miles per Gallon by Truck Configuration and Weight
Gross Vehicle Weight (pounds)
Truck Configurations
60,000
80,000
100,000
120,000
Five-Axle Semitrailer
6.2
5.5
4.9
Six-Axle Semitrailer
6.1
5.4
4.9
Five-Axle Double Trailer
6.8
6.0
5.5
Seven-Axle Rocky Mountain Double
5.8
5.2
5.0
Eight Axles (or more)
5.8
5.5
5.2
Triple-Trailer Combination
6.0
5.7
5.4
U.S. Department of Transportation. Western Uniformity Scenario Analysis: A Regional Truck Size and Weight Scenario
Requested by the Western Governors’ Association, April 2004. Original values indexed by 1.14.
6.1 Variations in Fuel Economy with Speed
According to Peterbilt Motors Company (2011), the power necessary to overcome aerodynamic drag at
speeds of 50 mph is equal to roughly half the power needed to overcome rolling resistance and the energy