ESTIMATION, VALIDATION, AND FORECASTS OF
REGIONAL COMMERCIAL MARINE VESSEL
INVENTORIES
Final Report
Submitted by
James J. Corbett, P. E., Ph. D.
Jeremy Firestone
University of Delaware
Coauthored by
Chengfeng Wang, Ph. D.
ARB Contract Number 04- 346
CEC Contract Number 113.111
Prepared for
the California Air Resources Board
and the California Environmental Protection Agency
and for
the Commission for Environmental Cooperation of North America
5 April 2007
i
Disclaimer
The statements and conclusions in this Report are those of the contractor and not necessarily
those of the California Air Resources Board. The mention of commercial products, their source,
or their use in connection with material reported herein is not to be construed as actual or implied
endorsement of such products.
ii
Acknowledgments
This work received partial funding from the Commission for Environmental Cooperation
( CEC) project number 113.111. This work also benefited from significant in- kind support from
members of the North American SOx Emission Control Area ( SECA) team and their contractors.
In particular, this work was shared with Research Triangle Institute ( RTI), at the direction of the
United States Environmental Protection Agency ( U. S. EPA) and ARB. RTI developed a trade-based
model forecasting energy demand from commercial ships, and the forecasts described in
this work are a product of coordination with this U. S. EPA- funded effort. Necessarily, we cite
these communications in this study, although future work may cite the RTI report to U. S. EPA
when that report can be referenced.
In particular, we thank John Callahan of the Research & Data Management Services of
the University of Delaware for his help applying the GIS tools. We acknowledge with
appreciation the significant review comments, and contributions by ARB staff ( including
Dongmin Luo, Todd Sax, Andy Alexis, Michael Benjamin, and Kirk Rosenkranz); the work
greatly benefited from their guidance. We acknowledge collaborative discussions with the North
American SECA team, including representatives of Environment Canada ( Joanna Bellamy,
Naomi Katsumi ( who provided the Canada data), Patrick Cram, Andrew Green, Veronique
Bouchet, and Morris Mennell), the Commission for Environmental Cooperation ( Paul Miller,
now at NESCAUM, who obtained the Mexico data on our behalf), and the U. S. Environmental
Protection Agency ( Barry Garelick, Penny Carey, and others). In addition, we acknowledge the
work and review of fellow SECA contractors, including Brewster Boyd at Ross and Associates,
Louis Browning at ICF, and Chris Lindhjem at Environ. We acknowledge the good work
products and collaborative discussions regarding forecast results with Mike Gallagher and
Martin Ross at RTI and with Dave St. Amand at Navigistics Consulting.
This Report was submitted in fulfillment of contract number 04- 346, Estimation,
Validation, and Forecasts of Regional Commercial Marine Vessel Inventories, by the University
of Delaware under the partial sponsorship of the California Air Resources Board ( ARB). Tasks
1 and 2 received partial funding from the Commission for Environmental Cooperation ( CEC)
project number 113.111, and significant in- kind support from member of the North American
SOx Emission Control Area ( SECA) team and their contractors. Work on Tasks 1 and 2 of the
project was completed as of March 2006. Work on Tasks 3 and 4 of the project was completed
as of October 2006. Project work was completed as of January 2007.
iii
Table of Contents
Disclaimer ............................................................................................................................... ...................... i
Acknowledgments................................................................................................................ ........................ ii
Table of Contents....................................................................................................................... ................. iii
List of Figures ............................................................................................................................... .............. iv
List of Tables ............................................................................................................................... ............... iv
ABSTRACT....................................................................................................................... .......................... v
EXECUTIVE SUMMARY ......................................................................................................................... vi
1.0 INTRODUCTION................................................................................................................... ........... 1
1.1 Purpose and Scope....................................................................................................................... 1
1.2 Project Background and Assumptions......................................................................................... 1
1.3 Previous Work ............................................................................................................................. 2
1.3.1 Inventory Development ...................................................................................................... 3
1.3.2 Trends and Forecasting ....................................................................................................... 4
2.0 MATERIALS AND METHODS........................................................................................................ 7
2.1 Baseline Conditions: STEEM description................................................................................... 8
2.2 Rates of Change: Installed power as first- order trend indicator for CMV emissions................ 12
2.2.1 Evaluating coupled growth in cargo and energy............................................................... 13
2.3 Patterns of Change: First- order consideration at North American scale ................................... 14
3.0 RESULTS ............................................................................................................................... ......... 16
3.1 Baseline Emissions Estimates ................................................................................................... 16
3.2 Producing Spatially Resolved Emissions Inventories for Various Pollutants ( Task 1)............. 18
3.3 Inventory Summary by Vessel Type ......................................................................................... 20
3.4 Comparison with Other Emissions Studies ( Task 2)................................................................. 20
3.5 Forecasting principles................................................................................................................ 23
3.6 Activity- based modeling of freight growth ............................................................................... 24
3.5.1 Growth Rates .................................................................................................................... 26
3.5.2 Growth Patterns ................................................................................................................ 29
3.7 Future Emissions without SECA region ( Task 3) ..................................................................... 30
3.8 Future Emissions with Potential SECA ( Task 4) ...................................................................... 31
4.0 DISCUSSION ............................................................................................................................... ... 33
4.1 Comparison with global forecast trends .................................................................................... 33
4.2 Uncertainty and Bounding......................................................................................................... 33
5.0 SUMMARY AND CONCLUSIONS................................................................................................ 35
5.1 Baseline Inventory..................................................................................................................... 35
5.2 Forecast Trends ......................................................................................................................... 36
6.0 RECOMMENDATIONS................................................................................................................ . 39
6.1 Improve precision...................................................................................................................... 39
6.2 Reduce Base- Year Uncertainty ................................................................................................. 39
6.3 Improve Trend Extrapolation .................................................................................................... 39
6.4 Incorporate additional detail among drivers affecting change................................................... 41
6.5 Incorporate planned or proposed signals to modify technological change trends ..................... 41
6.6 Model fleet behavior in response to potential action................................................................. 41
6.7 Extend voyage data or analytical detail ..................................................................................... 41
7.0 REFERENCES..................................................................................................................... ............ 42
LIST OF ACRONYMS .............................................................................................................................. 48
Appendix: Summary of North American Ports and Waterways................................................................ 49
iv
List of Figures
Figure 1. Illustration of Waterway Network Ship Traffic, Energy and Environment Model
( STEEM) as applied to emission estimation........................................................................... 9
Figure 2. Illustration of spatial distribution of SO2 from North American shipping; shaded areas
represent approximate delineation of coastal exclusive economic zones ( EEZs). ............... 19
Figure 3. Comparison of the inventories produced with Waterway Network- STEEM and top-down
approach using ICOADS. ........................................................................................... 22
Figure 4. Illustration of domains of regional/ port emission inventories studies........................... 22
Figure 5. Comparison of emissions inventories of different approaches; emissions for Houston &
Galveston are NOx, emissions for the other areas are SO2................................................... 23
Figure 6. Container statistics from U. S. Maritime Administration and American Association of
Port Authorities.................................................................................................................... 25
Figure 7. South Coast ( South Pacific) growth rates derived from historic data ( 1997- 2003),
showing upper- bound ( exponential), lower- bound ( linear), and average trends. ................. 28
Figure 8. US container growth trends from data extrapolation ( 1997- 2003) and from unpublished
draft RTI trade- energy model. .............................................................................................. 28
Figure 9. Model domain showing hypothetical with- SECA region and baseline 2002 model
results. ............................................................................................................................... ... 30
Figure 10. Illustration of 2020 ship SOx emissions without SECA reductions............................ 31
Figure 11. Illustration of 2020 ship SOx emissions with hypothetical SECA region. ................. 32
Figure 12. Global indices for seaborne trade, ship energy/ fuel demand, installed power............ 34
Figure 13. Forecast reduction in 2020 of annual SOx emissions due to hypothetical SECA....... 36
Figure 14. Forecast increases from base- year inventory in SOx emitted in 2020 with SECA..... 37
Figure 15. Trends with and without IMO- compliant SECA, and with 0.5% SECA .................... 38
Figure 16. Uncertainty in model output from input parameters scaled by contribution to output
variance. ............................................................................................................................... 40
List of Tables
Table ES- 1. Baseline 2002 inventory of emissions and fuel use in North American Domain
( metric tonnes)………………………………………………………………………..… vii
Table 1. Summary of engine power and at- sea load profile ......................................................... 16
Table 2. Emission Factors............................................................................................................. 17
Table 3. Summary of auxiliary engine SO2 emissions factor ....................................................... 17
Table 4. Baseline 2002 inventory of emissions and fuel use in North American Domain1 ......... 19
Table 5. Estimated Domain Emissions by Vessel Type .............................................................. 20
Table 6. Estimated Percent of Total Emissions from Auxiliary Engines ( AEs).......................... 20
Table 7. Power- based growth rate summary for commercial ships 2002 - 2020 ( CAGR)............ 29
Table A- 1. State- by- state Summary of Ports and Port Calls .…………………………………... 50
Table A- 2. U. S. Port and Waterway Summary from USACE Foreign Commerce Data ...…..… 51
Table A- 3. Canadian Port and Waterway Summary from LMIU Movement Data …………..… 57
Table A- 4. Mexican Port and Waterway Summary from LMIU Movement Data .…………..… 61
v
ABSTRACT
This report presents results of a project to develop and deliver commercial marine
emissions inventories for cargo traffic in shipping lanes serving U. S. continental coastlines. A
regional scale methodology consistent with port- based inventory methods was applied for
estimating commercial marine vessel ( CMV) emissions in coastal waters. Geographically
resolved inventories were produced for a 2002 baseline year ( Task 1). Several port- based
inventories were evaluated to validate the regional inventory ( Task 2). Using average growth
trends describing trade and energy requirements for North American cargo and passenger
vessels, an unconstrained forecast was developed to describe a business as usual ( BAU) scenario
without sulfur controls ( Task 3), and a with- SECA scenario assuming IMO- compliant reductions
in fuel sulfur to 1.5% by weight for all activity within the Exclusive Economic Zone ( 200
nautical miles) of North American nations ( Task 4). This work contributes to better regional
inventories of commercial marine emissions for North America that supports the California Air
Resources Board ( ARB), Commission for Environmental Cooperation of North America ( CEC),
western regional states, United States federal, and multinational efforts to quantify and evaluate
potential air pollution impacts from shipping in U. S, Canadian, and Mexican coastal waters.
vi
EXECUTIVE SUMMARY
Background: Current best practices for marine vessel emissions inventories have not
been applied to spatially and temporally describe North American interport shipping activity
until now. ( Interport shipping is ship activity voyaging between ports; it does not include
dockside hotelling.) We produced a baseline ( 2002) emissions inventory for ships engaged in
foreign commerce arriving at U. S. ports, and for ship activity in Canada and Mexico by
commercial cargo and passenger vessels ( excluding ferries). We forecast inventories for
business- as- usual ( BAU) and for a hypothetical SOx Emission Control Area ( SECA) including
the Exclusive Economic Zone ( EEZ) of North American nations ( i. e., 200 nautical miles). The
base- year inventory and forecasts assist the California Air Resources Board ( ARB) in evaluating
air quality and health impacts in California, and help evaluate national impacts, providing part of
the required information to request a North American SECA ( or SECAs) on behalf of the United
States, Canada, and Mexico at the International Maritime Organization ( IMO).
Methods: We use a network model, the Waterway Network Ship Traffic, Energy and
Environment Model ( STEEM), to quantify and geographically represent inter- port vessel traffic
and emissions for North America, including the United States, Canada, and Mexico. The model
estimates main and auxiliary engine emissions from nearly complete historical North American
shipping activities and individual ship attributes, applying activity- based emissions estimates in a
GIS platform using an empirically derived network of shipping routes.
We evaluate various sources of growth projections for commercial marine activity and
energy use, ultimately choosing an adjusted extrapolation scenario from historic trends in
installed power on ships calling on North American ports. Use of installed power trends depends
on the following assumptions: 1) commercial marine vessels in cargo service design power
systems to satisfy trade route speed and cargo payload requirements; 2) commercial marine
vessels operate under duty cycles that are well understood, especially at sea speeds; 3) installed
power trends for ships calling on North American ports directly reveals the trend in speed and
size for these routes. Trend extrapolations for installed power reveal the correlated trend in
energy use by ships, although different extrapolations approaches yield different forecasts. An
unconstrained exponential fit may be overly optimistic given economic cycles in shipping and
technological change in the fleet; a linear fit may be unrealistic with regard to fundamental work-energy
principles and economic drivers for global trade. These define bounding limits for
expected change in ship activity. We average these to describe a BAU growth trend that
implicitly reflects a mix of positive and negative drivers for ship energy requirements.
Results for Baseline Inventory: North American shipping consumed about 47 million
tons of heavy fuel oil and emitted ~ 2.4 million tons of SO2 in 2002, with approximately 30
million tons fuel and 1.6 million tons SO2 within the North American domain for this project.
Comparison of our results with port and regional studies shows good agreement, and improved
accuracy over existing top- down methods. Shipping activity within the domain, defined for this
project by consensus with the North American SECA team. Table ES- 1 summarizes the
interport inventory estimates for the baseline year of 2002. The table presents results for coastal
regions ( defined as the 200 nautical mile EEZ) by nation, and the total for all domain areas
outside coastal regions. Comparison of our results with five inventories from other regional and
port emissions inventories studies ( including Great Lakes, Western Canada, the Port of Los
Angeles, Houston & Galveston area, and the Port of New York and New Jersey) showed no bias
and better accuracy using STEEM than top- down emissions inventories.
vii
Results for Forecasts: We estimate a growth trend for North America ( including United
States, Canada, and Mexico) of about 5.9%, compounded. We produce two classes of forecasts:
1) a business as usual ( BAU) forecast applying a common growth trend without sulfur controls
( but with existing IMO NOx requirements); and 2) a with- SECA scenario assuming IMO-compliant
reductions in fuel sulfur to 1.5% by weight for all activity within the Exclusive
Economic Zone ( 200 nautical miles) of North American nations. Our BAU scenario compares
reasonably well with available energy and fuel usage trends and with trends describing growth in
trade volume; our growth trends are lower than have been reported since 2002 by major US
ports. We identify no systemic bias in our forecasts. Various trends agree under BAU scenarios
that energy used by ships bringing global trade to and from North America will double by or
before 2020. Forecasts show that implementing a North American SECA region reducing fuel-sulfur
content from 2.7% to 1.5% ( whether through fuel changes or through control technology)
will reduce future SOx emissions ( as SO2) by more than 700 thousand metric tons (~ 44%) from
what they may otherwise grow to be in 2020. However, our 2020 inventory with an IMO-compliant
SECA represents an increase over emissions in the 2002 base- year of more than 2
million metric tons of SOx emissions throughout the North American domain. At a growth rate
of 5.9% from the baseline year 2002, trade growth offsets emissions under a 1.5% fuel- sulfur
SECA by 2012; using alternative growth rates of 3.6% ( separate work presented to the West
Coast SECA team), emissions within a North American SECA return to 2002 levels by 2019.
Conclusions: Baseline ( 2002) inventory results are being used by ARB, the U. S.
Environmental Protection Agency ( U. S. EPA), Environment Canada, and others to model
atmospheric fate and transport of pollution, evaluate air quality impacts, and assess potential
health effects attributed to ships. Health and environmental impacts evaluated using these
inventories may merit emissions control beyond current IMO standards to maintain emissions
targets despite trade growth. Future work could improve precision of near- port inventories
through improved network or vessel activity details.
Table ES- 1. Baseline 2002 inventory of emissions and fuel use in North American Domain ( metric tonnes) 1
NOx as NO2 SO2 CO2 HC PM CO Fuel Use
United States EEZ2
West Coast 135,000 80,200 4,817,000 4,470 11,300 10,500 1,480,000
East Coast 255,000 151,000 9,095,000 8,440 21,300 19,900 2,800,000
Gulf Coast 174,000 103,000 6,201,000 5,750 14,500 13,600 1,910,000
Great Lakes 16,200 9,620 578,000 540 1,350 1,260 178,000
Alaska 63,300 37,600 2,260,000 2,100 5,300 4,940 697,000
Hawaii 20,500 12,200 732,400 680 1,720 1,600 226,000
Canada EEZ2,3
West Coast 21,900 13,000 781,000 720 1,830 1,700 241,000
East Coast 96,200 57,200 3,440,000 3,190 8,050 7,500 1,060,000
Great Lakes 10,100 5,980 359,000 330 840 800 111,000
Mexico EEZ2
West Coast 99,400 59,100 3,550,000 3,290 8,320 7,800 1,090,000
Gulf Coast 107,000 63,700 3,827,000 3,550 8,970 8,000 1,180,000
Total Coastal regions 998,000 593,000 35,640,000 33,100 83,500 77,900 10,980,000
Non- coastal regions4 1,740,000 1,040,000 62,200,000 57,700 146,000 136,000 19,170,000
Total in Domain 2,740,000 1,630,000 97,800,000 90,800 229,000 214,000 30,160,000
1. Values are rounded to three significant figures for presentation; sums may vary as a result of rounding.
2. National estimates of EEZ boundaries use an ArcGIS buffer of 200 nautical miles and informal national divisions.
3. Western Canada summaries include emissions in the Northwestern part of the domain; Eastern Canada summaries
include emissions in the Northeastern part of the domain.
4. Non- coastal regions are areas in the Domain not within the EEZ of Canada, United States or Mexico.
1
1.0 INTRODUCTION
This report is intended to assist the role of the California Air Resources Board ( ARB) and
other agencies evaluating the feasibility and extent of a North American Sulfur Emissions
Control Area ( SECA) as defined by the International Maritime Organization ( IMO) in terms of
potential impact to air quality and human health by oceangoing commercial marine vessels in
transit.
1.1 Purpose and Scope
A primary objective of this project is to describe a regional scale methodology for
estimating commercial marine vessel ( CMV) emissions in coastal waters ( i. e., the Exclusive
Economic Zone or EEZ) that is consistent with port- based inventory methods. There are several
tasks that follow from this objective, including:
Task 1 Provide a baseline inventory of CMV emissions at a regional scale appropriate for modeling
impacts relevant to potential SECA designation. Using this methodology, this work produced a
spatially resolved inventory of CMV emissions for North America for a baseline year of 2002.
This represents a distance larger than the Exclusive Economic Zone for the continental United
States and Canada and Mexico, a legal area beyond and adjacent to the territorial sea that
provides certain federal authority to protect and preserve the marine environment ( 1).
Task 2 Evaluate several port- based inventories in terms of their potential agreement and validation of the
regional inventory. We conclude that different assumptions, inputs, or methods applied in port-based
inventories produce expected differences reflecting more detailed local information at the
port level that cannot be easily reflected at the regional scale. Based on our results, we offer
recommendations to improve regional inventory methods or otherwise reconcile differences with
port- based inventories.
Task 3 Forecast how baseline emissions may change in future years. Future emissions will be dependent
in part upon the changes in emission factors ( due to MARPOL Annex VI, other policy, and other
changes in engine characteristics), changes in vessel size and number. Additionally, changes may
occur in vessel activity patterns and trade routes, and changes in fuel quality ( especially sulfur
content) – from a mix of technology, economic, and/ or policy drivers.
Task 4 Forecast future- year ship emissions under a potential SECA designation. Modification of future-year
baseline emissions are made using MARPOL Annex VI requirements that requires the sulfur
content of marine fuel used by marine engines within a SECA be equal to or less than 1.5% S by
weight.
This project supports ARB efforts to understand the significance of ship emissions, by
providing forecasts of CMV emissions under assumptions that describe trade- driven fleet
growth, technological changes, and potential designation of special areas under the IMO’s
MARPOL Annex VI convention, called SOx Emission Control Areas ( SECAs).
1.2 Project Background and Assumptions
ARB is participating in a collaborative effort to understand and quantify potential impacts
of CMV activity on North American pollutant emissions, air quality, and public health. This
collaboration is led by the U. S. EPA, with agency support also from Environmental Canada, and
ARB, and with funded participation by various university researchers and consulting firms.
Similarly, the California Goods Movement Action Plan and related efforts to improve freight
transportation infrastructure and environmental performance are multi- scale and multi-
2
dimensional interests that depend on a good understanding of international freight movement
through major U. S. ports, including but not limited to California ports.
While ARB may be most interested in how CMV emissions and their mitigation may
affect California, the international nature of shipping and multi- jurisdictional nature of policy
alternatives established a scale of interest that includes all North America. According to the
World Shipping Council’s container cargo rankings of U. S. ports ( 2), the ports of Los Angeles
and Long Beach together accounted for more than 36% of all U. S. containerized imports and
exports in 2003; together with Oakland, CA ports handle nearly half of all U. S. waterborne
containerized cargoes.
This report presents inventory methodology, results, and validation for ships engaged in
foreign commerce arriving at U. S. ports, and for ship activity in Canada and Mexico by
commercial cargo and passenger vessels ( excluding ferries). We produce a spatially- resolved,
activity- based inventory of North American shipping activity derived from 172,000 port calls in
2002 to Canada, Mexico, and the United States, employing activity- based methods in a GIS
network of empirical shipping routes. We derive emissions forecast trends directly from
aggregated installed power of ships calling on North American ports; this is because emissions
are directly proportional to engine power and load, which for at- sea conditions is highly
correlated with total installed power on commercial ships; this direct proportionality of stack
emissions to engine power is implicit in the use of power- based emissions factors in activity-based
inventory best practices. We then adjust base- year inventory to estimate emissions from
commercial marine vessels for 2010 and 2020. Using observed trends in installed power by
cargo and passenger vessels calling on North America, we produce two classes of forecasts: 1)
an unconstrained forecast applying a common growth trend to forecast a business as usual
( BAU) scenario without sulfur controls; and 2) a with- SECA scenario assuming IMO- compliant
reductions in fuel sulfur to 1.5% by weight for all activity within the North American nations.
1.3 Previous Work
Air pollutants from marine vessels account for a non- negligible portion of the emissions
inventory and contribute to air quality, human health and climate change issues at local, regional
and global levels ( 3- 25). According to the U. S. EPA, heavy duty truck, rail, and water transport
together account for more than 25% of U. S. CO2 emissions, about 50% of NOx emissions, and
nearly 40% of PM emissions from all mobile sources ( 26, 27). In Europe, freight modes together
generate more than 30% of the transportation sector’s CO2 emissions ( 28). In California, marine
vessel ship emissions are a significant concern with regard to state implementation of federal air
quality requirements ( http:// www. arb. ca. gov/ msprog/ offroad/ marinevess/ marinevess. htm),
particularly for air districts ( 21, 29)) and for major ports ( http:// www. portoflosangeles. org/ and
http:// www. polb. com/).
Better estimation of current and future emissions inventories, including spatial
representation, is needed for atmospheric scientists, pollution modelers, and policy makers to
evaluate and mitigate the impacts of ship emissions on the environment and human health. In
fact, understanding the nature of commercial marine ( e. g., cargo) vessel activity and energy use
serves both environmental and goods movement goals for the State of California and the nation.
This is particularly true for major ports which represent nodes connecting imported and exported
ship cargoes with road and rail freight transportation serving the U. S. and global economies.
3
1.3.1 Inventory Development
Although emissions estimates and fuel use are related to the energy used by ships, recent
studies call into question the validity of relying on the statistics of marine fuel sales ( 4, 30- 33).
Best practices of estimating emissions from transportation overall, and marine vessel emissions
inventories specifically, have focused on activity- based estimation of energy and power demands
from fundamental principles ( 4, 30, 32, 34). These approaches have shown that fuel allocated to
international fuel statistics is insufficient to describe total estimated energy demand of
international shipping. Even if marine fuel sales statistics were perfect, ships may consume fuel
far from where they purchase it. At best, regional statistics provide limited insight into the
spatial and temporal characteristics of ship energy consumption.
Principle existing approaches for producing spatially- resolved ship emissions inventories
generally can be categorized as either top- down or bottom- up. The fundamental difference
between these is that in bottom- up approaches emissions are directly estimated within a spatial
context, whereas in top- down approaches emissions are calculated without respect to location at
an aggregate level and may later be associated with spatial characteristics. In this work, a mixed
approach is developed. First, we associate port arrival- departure data with ship characteristics
data to identify more than 170,000 voyages for North America and to allow for activity- based
inventory methods of estimating emissions for each voyage. Second, we assign routes to voyage
origin- destination pairs using an empirically derived routing network in the Ship Traffic Energy
and Environmental Model ( STEEM); this is a top- down analytical approach in the sense that we
are not directly observing actual voyage routes, but modeling them according to a least- distance
algorithm intended to approximate a least- cost voyage. Third, we apply activity- based
assumptions about vessel speed, power, energy, and emissions directly within the voyage routing
network to produce spatially resolved emissions estimates.
Using a top- down approach, Corbett, et al. produced the first global spatial representation
of ship emissions using a shipping traffic intensity proxy derived from the Comprehensive
Ocean- Atmosphere Data Set ( COADS), a data set of voluntarily reported ocean and atmosphere
observations with ship locations ( 3, 11). They assumed that the reporting ship fleet is
representative of the world fleet, spatial distribution of ship reporting frequencies represents the
distribution of ship traffic intensity, and emissions are proportional to traffic intensity. Endresen,
et al. improved the global spatial representation of ship emissions by using ship size ( gross
tonnage) weighted reporting frequencies from the Automated Mutual- assistance Vessel Rescue
system ( AMVER) data set ( 5). They implicitly assumed that ship energy consumption and
emissions are proportional to ship size, which is not true for some types of ships, and they
observed that COADS and AMVER lead to highly different regional perturbations ( 5). Wang, et
al. addressed the potential statistical and geographical sampling bias of the International
Comprehensive Ocean- Atmosphere Data Set ( ICOADS, current version of COADS) and
AMVER data sets, the two “ best” global ship traffic intensity proxies, and made four
advancements to improve the accuracy of the top- down approach using ICOADS as spatial
proxy ( 35): i) trimming over- reporting vessels to mitigate geographic and statistical sampling
bias; ii) increasing sample size by using multiple- year ICOADS data; iii) weighting ship
observations with installed ship power to reflect emissions variability among different sizes and
types of vessels; and iv) smoothing the inventory with GIS tools.
The quality of top- down approaches is limited by the accuracy of global emissions
estimates, and inventory precision is limited by the representativeness of spatial proxies.
4
Significant differences exist among the various global ship emission inventories ( 4, 5, 30, 31).
Activity- based energy consumption and emissions in the updated inventory by Corbett and
Koehler roughly doubled the results of earlier studies ( 4). Uncertainty exists in the updated
inventory such that the upper bound is about 60% higher than the lower bound ( 4). Discrepancies
among different studies and the range between lower and upper bound of the same study can be
explained by the uncertainties of marine engine load factor, time in operation, and fuel
consumption rates, which vary by ship type, size, age, fuel type, and market situation ( 30, 31).
Variation in these inputs represents first- order barriers to improving the accuracy of the global
ship inventory. Second, since both ICOADS and AMVER data sets rely on voluntary reporting
and neither of them is randomly sampled, both of them are statistically and spatially biased ( 35).
Bottom- up approaches were applied by Lloyd’s register and Entec UK Limited to
produce regional ship emissions inventories for the European Monitoring and Evaluation
Programme ( EMEP) area, the Baltic Sea, and the Mediterranean Sea ( 17, 24, 25). In this type of
approach, ship and route specific emissions are estimated based on historical ship movements,
ship attributes, and ship emissions factors. The locations of emissions are determined by the
locations of the most probable navigation routes, which are great- circle ( i. e., radius) routes
between transoceanic origins and destinations, adjusted where prohibited by land, ice, or depth;
the Lloyds and Entec work was more regional ( not transoceanic) and generally followed straight-line
routes. Streets, et al. estimated emissions from international shipping in Asian waters based
on commodity flow associated with major sea routes ( 7, 8). The accuracy of this method, which
can be categorized as a bottom- up approach using trade as a proxy for emissions, is limited by
the assumed relationships between the volume of trade flow and emissions, which are more
closely related to ship installed power, load profile, etc., and by the aggregation of individual
voyage routes into major shipping lanes.
Although bottom- up approaches appear more precise than top- down methods, large- scale
bottom- up inventories also are uncertain because they must estimate engine workload, ship
speed, and most importantly, the speculative locations of the routes which determine the spatial
distribution of emissions. Given the large number of ship movements and potentially dynamic
shipping routes, the accuracy of regional annual inventories in bottom- up approaches is limited
when selected periods within a calendar year studied are extrapolated to represent annual totals
( 17, 24).
1.3.2 Trends and Forecasting
Trend analyses are useful in describing changes that may have occurred in the past or
how changes may occur in the future. While past trends can often be observed without an
understanding of underlying causes, they are useful when exploring relationships among
correlating histories to evaluate causal drivers or correlated indicators of change. Developing
future trends ( forecasts) represents an uncertain extrapolation of past observations considering
explicit or implicit assumptions about how the trend may be affected by sustained or modified
drivers or indicators of change.
Forecasts differ depending on their purposes and scales. Some forecasts look to reveal
where timely investment and action at a local scale or by a single firm can produce the most
benefit ( e. g., profit). Validity of insights is determined by whether recommended actions produce
expected outcomes for a given decision, not whether the forecast trend or future value is realized.
Other forests are intended to be conservative or aggressive; that is, they intend to be biased to
serve the decision makers’ value and tolerance for risk and surprise. This may describe large
5
scale forecasts such as emissions or trade trends. One challenging class of forecasts may be
considered “ difference” forecasts, where alternative scenarios illustrate how “ a path taken” may
differ from “ a path not taken” rather than to determine which is most probable. These kinds of
forecasts are common in policy domains, such as energy, environment, and economics ( e. g.,
IPPC scenarios). Certainly, freight forecasting presents one challenging example, especially at
the international or multinational scales, and especially when considering policy actions like a
SOx Emissions Control Area ( SECA) under IMO MARPOL Annex VI ( 36).
Previous studies described global growth rates for maritime shipping energy and
emissions based on fleet size, trade growth, and/ or cargo ton- km, mostly calibrated to linear or
conservative extrapolations of historic data. The IMO Study on Greenhouse Gas Emissions from
Ships ( 37) used fleet growth rates based on two market forecast principles, validated by historical
seaborne trade patterns: 1) World economic growth will continue; and 2) Demand for shipping
services will follow the general economic growth. The IMO study correctly described that
growth in demand for shipping services was driven by both increased cargo ( tonnage) and
increased cargo movements ( ton- miles), and considered that these combined factors make
extrapolation from historic data difficult. Nonetheless, their forecast for future seaborne trade
( combined cargoes in terms of tonnage) was between 1.5% and 3% annually. The IMO study
applied these rates of growth in trade to represent growth in energy requirements. The ENTEC
study ( 38) adopted growth rates from the IMO study.
Eyring et al. ( 39) estimated “ future world seaborne trade in terms of volume in million
tons for a specific ship traffic scenario in a future year” using a linear fit to historical gross
domestic product ( GDP) data. Interestingly, this represents one of the only studies to forecast
growth in seaborne trade for energy and emissions purposes at rates faster than GDP. The
TREMOVE maritime model ( 40, 41) estimates fuel consumption and emissions trends derived
from forecast changes in ship voyage distances ( maritime movements in km) and the number of
port calls. According to the TREMOVE report, maritime “ fleet and vehicle kilometres grow
annually by 2.5% for freight and 3.9% for passengers,” while “ port callings grew by 8%
compared to the previously used input figures.”
For national CMV emissions, U. S. EPA’s 2003 forecast methodology improved the
similarity between economic and emissions forecasts from earlier analyses ( 23, 42- 44), although
emissions forecasts represent a compound annual growth rate ( CAGR) of about 3.4% ( range of
2.8% to 3.8%, depending on pollutant). While shipping growth rates accounted for the effect of
increased tonnage in a newer fleet, they do not consider the effect of faster speeds – specifically
the additional installed power to meet combined size and speed requirements. Correcting for
these factors brings the forecasts for international marine activity into closer agreement with
trucking growth rates ( especially when rail cargo volume increases are considered), and better
describes the role of imports growth on the intermodal freight system.
Freight energy use is correlated to increased goods movement, unless substantial energy
efficiency improvements are being made within a freight mode ( e. g., U. S. rail) or across the
logistics supply network. Even assuming that efficiency improvements from economies of scale
reduce energy intensity and emissions rather than being directed to larger and faster ships ( e. g.,
containerships), compounding increases in trade volumes outstrip energy conservation efforts
unless technological or operational breakthroughs in goods movement emerge. However, except
for the Eyring et al. work, these linear extrapolations appear to present growth rates slower than
the economy; these linear extrapolations are likely biased underestimates, because shipping and
trade activity has grown ( and is forecast to grow) faster than the economy. Freight
6
transportation, particularly international cargo movement, is an important and increasing
contributor to global and national economic growth, as well as state and regional economic
growth in and around major cargo ports. If growth in GDP and trade volumes is compounded as
forecast by economic and transportation demand studies, then growth in energy requirements
should be non- linear also. The U. S. Bureau of Transportation Statistics ( BTS) recently released
a report that describes North American freight activity and trends ( 45). This document reports
growth rates for North America above 7.4% for international trade and above 7.2% across all
measures of value, and states that:
“ Since 1994, the value of freight moved among the three countries has averaged
almost 8 percent annual growth in both current and inflation- adjusted terms,
compared with about 7- percent growth for U. S. goods trade with all countries
( table 1). In 2005, both goods trade and gross domestic product ( GDP) grew in
inflation- adjusted terms. Except in 2001 and 2002, during the past decade, U. S.
trade with Canada and Mexico has increased at a faster rate than U. S. GDP.”
Growth in goods movement by dollar value may be expected to differ from growth in the
volume of goods moved, and in the change in activity by the multimodal fleets ( ships, trucks,
trains, and aircraft) moving cargo. We confirmed that the contribution of international trade is
increasing as a proportion of U. S. gross domestic product ( GDP) – i. e., freight transportation is
growing faster than U. S. GDP ( 45, 46). Economic activity related to imports and exports
together contribute about 22% of recent U. S. GDP in recent years; whereas, goods movement
contributed only about 10% of GDP in the 1970s. Moreover, the dominance of containerized
cargoes in seaborne trade suggests that truck and containerized shipments may double by 2025 or
sooner ( 47). GDP in the U. S. is growing at ~ 3.7% CAGR since 1980, and the freight sector is
growing at ~ 6.4% CAGR over the same period ( 46). This freight- sector growth rate in terms of
dollar value is reflected in the observed ~ 6.3% to 7.2% annual growth rates of “ high- value”
containerized trade volumes, particularly from Asia ( 48).
California studies also describe significant growth expected in commercial marine
emissions. The recent Clean Air Action Plan for Southern California ports estimates that
emissions of NOx and PM from oceangoing vessels will increase at baseline rates between 5.5%
and 6% CAGR, respectively, unless measures are taken to reduce emissions ( 49). 1 These growth
rates are consistent with trade growth rates, perhaps modified for IMO- compliant NOx
reductions in new vessels expected to call on California ports and descriptive of modest
improvements in fuel efficiency through fleet modernization and economies of scale. Studies for
Southern California ( San Pedro Bay) ports agree that growth in cargo volumes equivalent to 6-
7% compounding annual growth rates is expected ( 50- 53). However, increased cargo may not
produce a corresponding increase in port calls, as some studies interpret ( 51). Historic data on
port calls to San Pedro Bay have shown the number of ship calls remained between 5,000 and
7,000 calls per year since the 1950s ( 54). Furthermore, proportional relationships between
environmental impacts and goods movement trends are reflected in recent port and regional
studies of goods transport and economic activity, particularly for California ports ( 50, 55- 57).
1The Clean Air Action Plan shows emissions control measures may offset near- term growth ( at least through 2011)
if fully implemented.
7
2.0 MATERIALS AND METHODS
This section describes principles, methods, and data used to produce baseline inventories
and future emissions inventory scenarios for North America. This project represents one of the
first applications of a network model developed to evaluate ship activity characteristics on large
regional and global scales using best- practice assumptions and methods comparable to the latest
port- based inventories of ship activity. The Ship Traffic Energy and Environmental Model
( STEEM) enables emissions inventory analyses that are not scaled from studies of a subset of
ports or smaller regions or patched together from separate inventory efforts ( 58, 59). Starting
with a global empirical network of observed shipping lanes, commercial cargo and passenger
ship arrivals and departures from all ports in North America are routed along coastal and
transoceanic shipping lanes. Vessel engine, speed, and size data for these vessels are applied to
estimate emissions from these vessels in both spatial and temporal domains.
In general, materials for this work include the global network developed at the University
of Delaware primarily by Dr. Chengfeng Wang ( 60), vessel activity data for the United States
from the U. S. Army Corps of Engineers ( 61), vessel movement data for Canada and Mexico
from Lloyds Maritime Intelligence Unit ( LMIU) provided by Environment Canada and the
Commission for Environmental Cooperation, respectively ( 62, 63). Ship characteristics were
also obtained from Lloyd’s ship registry data ( 64). Inventory assumptions and other model
inputs were primarily derived from earlier ARB reports and published work by Dr. Corbett ( 4,
30, 65), modified through discussion with U. S. EPA contractors and review of port- based best
practices ( 34).
Emissions trends are derived from a pluralistic evaluation of historic time series of the
above data and forecast studies that together describe: a) growth expected in international goods
movement in economic terms ( e. g., seaborne trade); and b) correlated trends in energy required
to move more goods in service of global trade in terms of ship fleet characteristics ( e. g., vessel
type and installed power). For cargo activity, we reviewed studies at port, regional, national, and
global scales, all of which document strong growth trends and/ or forecast similar rates of
continued growth ( 50- 53, 66- 71). For vessel activity specific to North American ports, we were
able to construct detailed trend characteristics information including vessel type, power, size, and
speed characteristics for the period between 1997 and 2003; at the global scale, we developed
longer time- series trends in ship characteristics by year of build and from related global studies
( 39, 64).
Three critical questions for understanding freight activity and environmental impacts
defined two phases of the project:
1. Baseline Conditions: What are freight energy and activity patterns?
2. Rates of Change: What is forecast trend in energy needed?
3. Patterns of Change: Where is future freight activity located?
While interrelated, these questions may be evaluated with some independence, and were
separated into phases combining Tasks 1 and 2 and combining Tasks 3 and 4, described above.
The first phase evaluated baseline conditions by applying STEEM, a model that integrates a GIS
routing algorithm allocating North American voyage data to empirically derived global ocean
routes with activity- based methodology to estimate emissions. The second phase analyses
considered rates of change in energy and emissions, demonstrating that installed power was not
only a direct input to estimating baseline emissions, but that installed- power trends described
8
rates of change in fleet energy requirement. These phases are described in detail in earlier
technical memoranda, and summarized below.
2.1 Baseline Conditions: STEEM description
By applying advanced GIS tools and using better data sets, STEEM adopts the strengths
of both top- down and bottom- up approaches and attempts to overcome the weaknesses in each
approach and improves ship emissions inventory both mathematically and theoretically. First, the
model builds an empirical waterway network based on shipping routes revealed from observed
historical ship locations. The spatial allocation approaches the accuracy of a bottom- up approach
by assigning routes from a historically accurate network of actual routes, and is more accurate
than a top- down approach, which uses biased spatial proxies. Second, as in a bottom- up
approach, this model estimates energy use and emissions using complete historical ship
movements, ship attributes, and the distances of routes. Best- practices applied to baseline
inventories include identification and use of installed power characteristics, current power- based
emissions factors, engine load service corrections, and engine operating time ( 34, 72, 73).
STEEM improves baseline emissions inventories for North American shipping in the following
ways:
1. STEEM employs an emprical global waterway network derived from 20- year
International Comprehensive Ocean- Atmosphere Data Set ( ICOADS) data;
2. The model estimates emissions from nearly complete historical North American
shipping activities ( some 172,000 trips in U. S. Foreign Commerce Entrances and
Clearances data set and Lloyds’ Movement data set) and individual ship attributes
while a top- down approach estimates emissions based on statistical analysis;
3. The model is constructed using advanced GIS network analyst technology to solve the
most probable route for each individual trip on a global scale; 2
4. STEEM establishes explicit mathematical relationships among trips, ships, routes, pairs
of ports, and segments of the waterway network using a matrix approach;
5. STEEM uses actual lengths of routes, together with service speed of each individual
ship, to calculate hours of operation while top- down approaches estimate annual hours
of operation based on fleetwide statistics;
6. STEEM follows best practice to estimate emissions based on ship installed power,
service speed, and traveling distance for each trip;
7. STEEM assigns emissions based on the locations of solved routes while earlier bottom-up
approaches drew straight lines between origins and destinations manually and top-down
approaches allocate global emissions based on biased proxies;
8. STEEM captures transit traffic which contributes to local air quality problems in some
areas like Santa Barbara, CA, while port- wide inventories have often ignored or been
unable to quantify these effects.
Figure 1 illustrates the ship traffic module of STEEM, which can geographically and
temporally characterize ship traffic based on an empirical waterway network, historical ship
movement data, and ship attributes data set. The lower boxes in Figure 1 illustrate how we
applied ship attributes data to produce activity- based, spatially- resolved emissions inventories.
2 A summary of ~ 400 North American ports and waterways is provided in the Appendix; these ports connect about
with ~ 1,300 foreign ports in the 2002 U. S. Entrances and Clearances data set; about 950 ports are in the 2002
Lloyd’s movement data set, with some overlapping ports among Canada, Mexico, and the United States.
9
The empirical waterway network built in this model not only aligns the shipping lanes
with actual shipping activity, but also defines the relationships among routes, segments and
nodes with ArcGIS Network Analyst tools. In the empirical waterway network, intersections of
shipping lanes and ports are defined as nodes, and shipping lanes between two immediate nodes
are defined as segments. Traffic can only flow in and out of segments through nodes. A route is
defined as an actual non- stop path ships take between one origin and one destination port. We
next describe the model when applied to ship energy, fuel use, or emissions. With minor
modifications to account for different attributes, the model is generalizable to the other
catergories specified in the lower part of Figure 1.
Figure 1. Illustration of Waterway Network Ship Traffic, Energy and Environment Model
( STEEM) as applied to emission estimation.
The distance of each route can be determined by multiplying the transposition of matrix
A with matrix E and is denoted as matrix F where, A ´ is the transposition of matrix A, and dn is
the distance of route n. Energy, fuel use, or emissions per unit of length for route n can be
determined by dividing the emissions en for each route by its length dn and can be denoted as un.
Enery and emissions per unit of length for all routes are denoted as matrix G.
Total energy, fuel use, or emissions from each segment within one period can be obtained
by summing up the calculations from all trips on that segment during that period. Energy, fuel
use, or emissions per unit of length for all segments are denoted as matrix H, where hm indicates
the distribution of energy, fuel use, or emissions per unit of length for segment m. Total energy,
fuel use, or emissions for segment m can be calculated by multiplying each segment length lm by
its per- unit fuel use or emissions hm and can be denoted as km. Total energy, fuel use, or
emissions for each segment can be further allocated to each grid to produce spatially- resolved
inventories per gridded area if the segment was established as a polygon.
Ship Traffic Spatial Proxy Ship Movement Data Set
Shipping Routes Individual Trips
Ship Attributes Data Set Routes & Network Segments Relationship
Spatial Distribution of Shipping Activities
Network Analysis Routes & Trips Relationship
Emissions from Shipping Activities Spatial Distribution of Emissions
Empirical Waterway Network
10
Matrix A describes the many- to- many relationships across m segments and n routes in
the empirical waterway network, where, bm, n is a binary variable that shows whether segment m
is part of route n ( value of “ 0” if no, “ 1” if yes).
m m m m n
n
n
n
b b b b
b b b b
b b b b
b b b b
A
,1 ,2 ,3 ,
3,1 3,2 3,3 3,
2,1 2,2 2,3 2,
1,1 1,2 1,3 1,
L
M M M M M
L
L
L
= ( 1)
Relationships between routes and trips can be denoted as matrix B where, tn is the
number of trips on route n within one period. The actual number of trips on each route in any
temporal period, where trips are defined as a one- way movement on one route, can be derived
from ship movement data set, where, tn is the number of trips on route n within one period.
n t
t
t
t
B
M
3
2
1
= ( 2)
Depending on need and data availability, we can either assume ships are identical ( as one
group or in subsets by vessel type, fuel properties, etc.) or incorporate individual ship
characteristics into the model. The number of trips or the indicator of traffic volume weighted by
ship attributes on each segment can be denoted as matrix C, where, vm is the number of trips or
the indicator of traffic volume of segment m in one period.
m v
v
v
v
C A B
M
3
2
1
= × = ( 3)
To estimate fuel use and air emissions out of port areas, we assume ships travel at a
typical cruising speed, which appears true in most cases. Fuel use and air emissions from
individual trips can be estimated with current best- practice models based on route distance, ship
characteristics, and ship operating profile. Total emissions en on route n in one period in which
there were tn trips is estimated by equation ( 4), and fuel use fn can be estimated by equation ( 5).
( , , , , , , )
1
i i i m a p L
t
i
n n e f d s m a l l e
nΣ=
= ( 4)
( , , , , , , )
1
i i i m a f L
t
i
n n f f d s m a l l sfoc
nΣ== ( 5)
Where, dn is the length of route n, s is vessel speed, m is main engine power, a is
auxiliary engine power, lm and la are load factors for main and auxiliary engines, and ep
represents emission factor for pollutant p; sfoc in equation ( 5) represents specific fuel oil
11
consumption ( energy rate factor) for fuel type f. Equations ( 4) and ( 5) denote that total emissions
en or fuel use fn on route n in one period is a function of the length of route, the characteristics of
the ships on that route, the operating profile of the ships, and other variables concerned like the
quality of fuel, etc. Where vessel- specific estimates are not required, average vessel values can
be assigned by vessel type ( e. g., tankers, containerized vessels, bulk carriers) to estimate energy,
fuel use, or emissions by route.
Energy, fuel use, or emissions from each route can be denoted as matrix D.
n e
e
e
e
D
M
3
2
1
= ( 6)
Fuel use and emissions per unit of length are determined by dividing the total emissions
on one route by the length of that route, which is the sum of the lengths of all segments of the
route. The length of each segment can be obtained by GIS tools and can be detonated as matrix
E, where, lm is the length of segment m.
m l
l
l
l
E
M
3
2
1
= ( 7)
The distance of each route can be determined by multiplying the transposition of matrix
A with matrix E and is denoted as matrix F, where, A ´ is the transposition of matrix A, and dn is
the distance of route n.
n d
d
d
d
F A E
M
3
2
1
= ' × = ( 8)
Energy, fuel use, or emissions per unit of length for route n can be determined by
equation ( 8) and can be denoted as un.
n
n
n d
u = e ( 9)
Enery and emissions per unit of length for all routes are denoted as matrix G.
12
n u
u
u
u
G
M
3
2
1
= ( 10)
Total energy, fuel use, or emissions from each segment within one period can be obtained
by summing up the calculations from all trips on that segment during that period. Energy, fuel
use, or emissions per unit of length for all segments are denoted as matrix H, where, hm is
energy, fuel use, or emissions per unit of length for segment m. hm indicates the distribution of
emissions over the waterway network.
m h
h
h
h
H A G
M
3
2
1
= × = ( 11)
Total energy, fuel use, or emissions for segment m can be calculated by equation ( 12) and
can be denoted as km.
m m m k = l × h ( 12)
Total energy, fuel use, or emissions for each segment can be further allocated to each grid
to produce spatially- resolved inventories per gridded area if the segment was established as a
polygon.
2.2 Rates of Change: Installed power as first- order trend indicator for CMV emissions
Given that energy used and emissions produced during goods movement increases at a
rate correlated to growth in activity, a number of proxies may be used to estimate inventory
growth rates. These include: economic activity ( GDP and imports/ exports value), trade activity
( tons and ton- miles), and fuel usage ( sales and estimates). All of these are indirect proxies
( second or higher order) of the activity that produces emissions. Except for complete and
accurate fuel usage statistics, none directly describe power requirements for shipboard power
plants ( propulsion and auxiliary engine systems). Best practices for ship emissions inventories
typically use power- based ( or fuel- based) emissions factors, because of the implicit
proportionality between engine load and pollutant emissions – especially for uncontrolled
sources ( 34, 72). Therefore, we derive emissions trends directly from installed power data for
cargo ships in the world fleet.
Assumptions we must make to use trends in installed power are rather simple: 1)
international vessels in cargo service generally design power systems to satisfy trade route speed
and cargo payload requirements; in other words, there is no economic reason to design
propulsion systems for containerships, tankers, etc., with more power than their cargo transport
operation requires; 2) international vessels operate under duty cycles that are well understood,
especially at sea speeds, which for most vessel types utilize the majority of installed power as
reflected in best practice methodologies for activity based inventories of energy and emissions
from ships; and 3) ships in commercial cargo service on major trade routes reflect the best fit of
13
ship design to service requirements; in other words, the trends revealed in installed power of
ships reveals fleet trends in speed and size. With these assumptions, trends in installed power
reveal the correlated trend in energy use by ships.
We evaluated installed power data associated with port calls from USACE and Lloyds
Registry ( for U. S. activity) and from LMIU data ( for Canada and Mexico). Where data were
missing in the installed power field for some vessels, we used linear regression statistics within
each vessel type associating gross registered tonnage ( GRT) and rated power to fill data gaps.
Over a period from 1997 to 2003, we observed the trend in total ship calls, their collective cargo
capacity ( tonnage), and aggregate installed power. Observations provided further confirmation
that ship calls change over time differently than cargo capacity; we also observed the expected
relationship between growth in cargo capacity and installed power. Based on this analysis
( performed for major ports in the U. S. and Canada using 1997- 2003 data), related evaluation of
trends in world fleet propulsion back to 1970, and discussions with the North American SECA
team and with ARB, we used installed power trends to develop emissions forecast growth rates.
2.2.1 Evaluating coupled growth in cargo and energy
A variety of curves could be fit to the multi- year installed- power data. We believe that
the underlying driver for growth in energy and emissions for CMVs is economic trade, which has
and is expected by all accounts to grow at compounding rates. In theoretical terms, if the
underlying functional form driving growth is non- linear, we see no justification for fitting a
linear growth curve to the available data points. In practical terms, work and energy to move
goods by ship are coupled fundamentally unless operational or technological change occurs.
Compounding growth in goods movement could not be associated with a linear trend in energy
or emissions unless that decoupling is dramatic. Air emissions control in onroad mobile sources
provides examples where this has occurred; emissions trends of CO2 and NOx from heavy- duty
trucks were decoupled, because regulatory action required new technologies that reduced NOx
emissions substantially despite increased energy use over the same period ( 26, 27).
An important question is whether forecasts that directly apply seaborne trade growth rates
to energy and emissions trends should assume any change in the fleet- average energy intensity
over time. In international shipping, economies of scale and a shift to thermally efficient slow-speed
diesels over the past three- to- five decades have served as the major drivers for
technological change; ship air emissions remain the least regulated mobile source, and IMO
regulations do not compare with the stringency of onroad standards. A common belief is
technological change improves energy efficiency in ocean freight transportation ( i. e., reduces
energy intensity) over time; rationale for this belief may extend from two historical facts about
shipping and energy use: 1) shipping has traditionally been less energy intensive than other
freight modes ( especially trucking), and 2) marine propulsion engineering developments over the
past century produced what are arguably the most fuel- efficient internal combustion ( diesel)
engines in the world ( 74).
Our hypothesis was that these conditions may, at best, result in a less aggressive
compounding growth in installed power, not a decoupling of work and energy significant enough
to justify a linear fit to installed- power data. Depending on change in energy intensity and/ or
emissions through investments in economies of scale, fuel conservation measures, or emissions
control measures, the rate of change in energy and emissions could be a modified growth curve
from the growth in cargo activity. If so, one indication would be different rates of change for
installed power on ships providing goods movement compared to changes in cargo volume. In
14
other words, if a fleet of ships can carry more cargo without a proportional increase in installed
power, then it must be adopting improved technologies ( e. g., hull forms, engine combustion
systems, plant efficiency) or innovating its cargo operations ( e. g., payload utilization).
In fact, the opposite trend is observed in the world fleet over the past 20 to 30 years,
where fleet installed power has grown at rates faster than global trade growth. Fleetwide
improvements in fuel economy ( indicated for marine engines by in- service specific fuel oil
consumption averages and/ or thermal efficiency) have been much smaller than growth in
seaborne trade and CMV installed power. The compound annual growth rate ( CAGR) for
installed power since 1985 is ~ 10.7% per year, more than twice the rate of world seaborne trade
growth, driven by increases in containership power which grew at more than 16% CAGR over
these two decades. While the slope before 1980 appears similar to the slope after 1985, one can
observe the significant fleet restructuring ( particularly for tankers) during the economic recession
in the early 1980s. Choosing a period since 1970 ( inclusive of the 1980s shipping recession), the
rate of installed power growth for the world fleet ~ 5.1% CAGR; even so, power growth rates for
the liner fleet over this period were still greater than 9% CAGR.
Rephrasing, ocean shipping may have become more energy intensive, not more energy
conserving. This seemingly counter- intuitive observation is explainable in terms of globalization
and containerization of international trade. Globalization has resulted in longer shipping routes,
and containerization serves just- in- time ( or at least on- time) liner schedules; both of these drivers
motivated economic justification for larger and faster ships which require greater power to
perform their service. Increasingly over the past two decades, ships serving all routes became
faster and larger through intentional expansion and aging fleet transition from prime routes to
secondary markets.
Of course, trends in installed power serving North America may differ from this global
installed- power trend. Introduction of the fastest, largest ships first occurs on the most valuable
trade routes ( e. g., serving North America and Europe) where economics most justify the higher
performing freight services. Given this, recent power growth trends for North America could be
lower than the global average rate because recapitalization of ships on these mature
containerized routes is not so heterogeneous, while larger and faster ships sold on the current
second- hand market may have significantly more power than the ships they replace. We
observed this to be true. A simple exponential curve fit to installed power produced an initial
growth rate estimate of ~ 7% per year for North America, compared to ~ 11% globally.
2.3 Patterns of Change: First- order consideration at North American scale
This project identified heterogeneity in growth rates among several other dimensions.
Containership growth rates are significantly larger than growth in dry bulk and tanker ships, for
both seaborne trade volume and installed power. Energy use and emissions on routes to major
containerized ports, therefore grows faster than routes primarily serving bulk trades. Regionally,
growth in West Coast ports is generally stronger than North American average growth rates.
While results reveal heterogeneity in CMV growth rates, timing and budget limitations
prevented us from forecasting growth rates spatially by vessel- route combination. Maps
forecasting emissions applied North American average growth rates to our base- year inventory
patterns. By increasing emissions proportionally for all routes on all North American coastlines,
our spatially resolved forecasts necessarily underestimate growth on the West Coast where
emissions from containerized trade are growing faster than the national average and
overestimates emissions growth in regions where overall trade growth is slower, such as the Gulf
15
of Mexico served mostly by bulk ships. As such, this represents a first- order forecast appropriate
to consider the value of a SECA for North America but not explicit enough without additional
work to apply to other large- scale issues such as port development or regional shifts in traffic.
16
3.0 RESULTS
This section describes specific input parameters chosen for STEEM and presents 2002
baseline inventory results required under Task 1; we also summarize Task 2 comparisons and
validation using port- based and regional inventories. This section then presents results of BAU
forecast trends required in Task 3 using the adjusted power- based extrapolations discussed
previously, and a with- SECA scenario under Task 4 that assumes IMO- compliant reductions in
fuel sulfur to 1.5% by weight for all activity within the Exclusive Economic Zone ( 200 nautical
miles) of North American nations.
3.1 Baseline Emissions Estimates
Main engine power of individual ships was used to estimate ship energy, fuel use, or
emissions for each trip. We adopted the at- sea main engine load factors used by Corbett and
Koehler for the updated emissions inventory for international shipping ( 4). Based on engine
manufacturer data used in other global analyses, we assumed that 55% of passenger vessel total
main engine power is devoted to propulsion, and 25% of remaining power serves Auxiliary
Engine ( AE) power ( 4, 30). We used maneuvering load profile ( lower engine load factor and
slower ship speed) for the first and last 20 kilometers of each trip when a ship is entering or
leaving a port. If the trip was shorter than 20 kilometers, we assumed that ships were
maneuvering for the whole trip; although this assumption may underestimate emissions from
some short- sea routes. We assumed that main engines operate at 20% of the installed power
during maneuvering, the same number used by Entec UK Limited ( 17).
Since most of auxiliary engine data for ships are missing in the ship attributes data set,
average auxiliary power of each ship type was used to estimate the energy, fuel use, or emissions
from auxiliary engines. California Air Resources Board ( ARB) survey results indicate that " 29
percent of the auxiliary engines used marine distillate and 71 percent used HFO, except for
passenger vessels that use approximately 8 percent marine distillate and 92 percent HFO" ( 75).
This number was adopted to adjust the SO2 emissions factor for auxiliary engines. Table 1
summarizes the engine power and at- sea load profile used in this work. The average total
installed auxiliary engine power was adopted from ARB survey ( 75); as documented by ARB
and others, most vessels have multiple auxiliary engines.
Table 1. Summary of engine power and at- sea load profile
Vessel Type Average ME
Power ( kW)
At- sea ME
load (% MCR)
Average Total AE
Power ( kW)
At- Sea AE
Load
Bulk Carrier 7,954 75% 1,169 17%
Containership 30,885 80% 5,746 13%
General Cargo 9,331 80% 1,777 17%
Passenger/ Cruise 39,563 55% 39,563 25%
Refrigerated Cargo 9,567 80% 1,300 20%
Roll On- Roll Off 10,696 80% 2,156 15%
Tanker 9,409 75% 1,985 13%
Miscellaneous 6,252 70% 1,680 17%
We use emissions factors shown in Table 2. Consistent with previous studies and with
both the ICF report and ARB survey results, we assume all main engines use residual fuel - this
is standard practice especially in transit at sea. The emissions factors reported in the recent ARB
17
report “ Emissions Estimation Methodology for Ocean- Going Vessels” are nearly identical to
those in the ICF best practices paper, and indeed nearly identical to emission factors used in all
recent analyses in the U. S., Canada, and Europe ( 4, 17, 34, 75, 76). We use the composite EF for
our work because our data do not explicitly identify by voyage whether the main engine is slow
or medium speed or whether the auxiliary engine uses distillate or heavy fuel. This composite
may be recalculated for the Great Lakes if data for that region enables more specific analysis of
the vessel, engine, and fuel characteristics.
Table 2. Emission Factors
Main Engine Emission Factors - In- Transit Operations ( g/ kWh)
Engine Type Fuel Type NOx SOx CO2 HC PM* CO**
Slow Speed Heavy Fuel Oil 18.1 10.5 620 0.6 1.5 1.4
Medium Speed Heavy Fuel Oil 14 11.5 677 0.5 1.5 1.1
Composite EF Heavy Fuel Oil **** 17.9 10.6 622.9 0.6 1.5 1.4
Auxiliary Engine Emission Factors ( g/ kWh)
Engine Type Fuel Type NOx SOx CO2 HC PM CO***
Medium Speed Marine Distillate 13.9 4.3 MDO
1.1 MGO
690 0.4 0.3** 1.1
Heavy Fuel Oil 14.7 12.3 722 0.4 1.5* 1.1
Composite EF **** 14.5 9.1 713 0.4 1.2 1.1
* Emission Factors from ARB Staff
** Emission Factors from Environ Report
*** Port of Los Angeles
**** Composite used population weighting from ARB OGV Survey, 2005
Considering emissions factors used in previous studies, we used a composite SO2
emissions factor of 10.6 g/ kWh to estimate main engine SO2 emissions ( 4, 17). The SO2
emissions factors for auxiliary engines using marine distillate oil ( MDO) and heavy fuel oil are
4.3 g/ kWh and 12.3 g/ kWh respectively; for this study we do not assume oceangoing ships use
marine gas oil ( MGO). A composite SO2 emission factor was adopted for each type of ship,
weighted by the percent of marine distillate used by that type of vessel ( 75). Table 3 summarizes
the auxiliary engine SO2 emissions factors used for each type of ship in this work. The percent
in- use marine distillate of auxiliary engines was adopted from the ARB survey ( 75). For
estimating fuel consumption, 206 g/ kWh was used as Specific Fuel Oil Consumption ( SFOC) for
transport ships and 221 g/ kWh for miscellaneous ( non- transport) ships, including fishing and
factory vessels, research and supply ships, and tugboats, as adopted in other studies ( 4).
Table 3. Summary of auxiliary engine SO2 emissions factor
Vessel Type Percent In- Use Marine
Distillate
Composite Aux. EF
( g/ kWh)
Bulk Carrier 29% 9.98
Containership 29% 9.98
General Cargo 29% 9.98
Passenger/ Cruise 8% 11.66
Reefer 29% 9.98
RORO 29% 9.98
Tanker 29% 9.98
Miscellaneous 100% 4.3
18
We estimated that inter- port transport of North American commerce ( including global
voyage transits on route segments outside the project domain) consumed more than 44.7 million
tons of heavy fuel oil and emitted about 2.3 million tons of SO2 in 2002, about 16.5% of SO2
emissions from all sources in the U. S. in the same year ( 77). Given that in- port emissions are
about 2 to 6% of total emissions, as reported by Streets et al. and Entec UK Limited ( 8, 17), total
heavy fuel use and SO2 emissions from North American shipping are approximately 47 million
tons and 2.4 million tons, respectively. The North American shipping fuel use and SO2 emissions
are between 18- 20% of the world commercial fleet estimated by Corbett and Koehler and
between 28- 34% of the world cargo and passenger fleet estimated by Endresen et al. ( 4, 5).
We estimated that ships carrying U. S. foreign commerce consumed about 38 million tons
of fuel in 2002 ( again including global voyage transits on route segments outside the project
domain). This number agrees well with Energy Information Administration statistics that
estimate that ships consumed about 44 million tons of fuel in 2002. U. S. domestic waterborne
commerce, which we did not include in this work, may be partially responsible for the
difference. Moreover, it is likely that the actual distance ships travel often is longer than the
distance estimated by the STEEM because data for this work include North American voyages
only between prior and next ports and do not model multi- port logistics activity common to
commercial shipping ( especially containerships).
Containerships, bulk carriers, and tankers account for about 35%, 22%, and 17% of SO2
emissions from North American shipping, respectively. Other types of ships jointly account for
the remaining 26%. The top ten maritime countries collectively account for about 71% of the 2.3
million tons of SO2 emissions. Panama, the largest flag of convenience country, accounts for
23% of the SO2 emissions. Liberia, Bahamas, and the U. S. account for 13%, 8%, and 5% of the
emissions, respectively. The Norwegian International Register, Singapore, Greece, Cyprus,
Malta, and Hong Kong each account for between 3- 4% of the emissions. The other 111 countries
account for the remaining 29% of the emissions. The energy use profile is similar to the SO2
emissions profile.
3.2 Producing Spatially Resolved Emissions Inventories for Various Pollutants ( Task 1)
Based on relationships among trips, routes and segments of the network, we allocated
total emissions onto the waterway network. We buffered the network with the width of each
segment and calculated the area of the segments in ArcMap. We calculated average emissions
per square kilometer by dividing total emissions for each pollutant in each segment with its area.
We converted the buffered network to a raster file with a resolution of 4 kilometers by 4
kilometers, where each grid value is emissions from this 16 square kilometer area. We adjusted
emissions within a 20- kilometer radius circle of ports to match maneuvering load profiles. Table
4 summarizes interport inventory estimates for 2002 by coastal regions ( defined as the 200
nautical mile exclusive economic zone), by nation, and totals for areas outside coastal regions.
Figure 2 illustrates the spatial distribution of annual SO2 from North American shipping. Coastal
zones resemble the 200 nautical miles Exclusive Economic Zone ( EEZ) but national divisions
serve illustration purpose only. Monthly and annual pollutant inventories are posted at
http:// www. ocean. udel. edu/ cms/ jcorbett/ sea/ NorthAmericanSTEEM ( SOx as sulfur dioxide,
NOx as nitrogen dioxide, CO as carbon monoxide, CO2 as carbon dioxide, PM as PM2.5, and HC
as total hydrocarbons).
19
Table 4. Baseline 2002 inventory of emissions and fuel use in North American Domain1
Units: metric tonnes NOx as NO2 SO2 CO2 HC PM CO Fuel Use
United States EEZ2
West Coast 135,000 80,200 4,817,000 4,470 11,300 10,500 1,480,000
East Coast 255,000 151,000 9,095,000 8,440 21,300 19,900 2,800,000
Gulf Coast 174,000 103,000 6,201,000 5,750 14,500 13,600 1,910,000
Great Lakes 16,200 9,620 578,000 540 1,350 1,260 178,000
Alaska 63,300 37,600 2,260,000 2,100 5,300 4,940 697,000
Hawaii 20,500 12,200 732,400 680 1,720 1,600 226,000
Canada EEZ2,3
West Coast 21,900 13,000 781,000 720 1,830 1,700 241,000
East Coast 96,200 57,200 3,440,000 3,190 8,050 7,500 1,060,000
Great Lakes 10,100 5,980 359,000 330 840 800 111,000
Mexico EEZ2
West Coast 99,400 59,100 3,550,000 3,290 8,320 7,800 1,090,000
Gulf Coast 107,000 63,700 3,827,000 3,550 8,970 8,000 1,180,000
Total Coastal regions 998,000 593,000 35,640,000 33,100 83,500 77,900 10,980,000
Non- coastal regions4 1,740,000 1,040,000 62,200,000 57,700 146,000 136,000 19,170,000
Total in Domain 2,740,000 1,630,000 97,800,000 90,800 229,000 214,000 30,160,000
1. Values are rounded to three significant figures for presentation; sums may vary as a result of rounding.
2. National estimates of EEZ boundaries are approximate, using an ArcGIS buffer of 200 nautical miles
and informal divisions between nations.
3. Western Canada summaries include emissions in the Northwestern part of the domain; Eastern Canada
summaries include emissions in the Northeastern part of the domain.
4. Non- coastal regions are areas in the Domain not within the EEZ of Canada, United States or Mexico.
Figure 2. Illustration of spatial distribution of SO2 from North American shipping; shaded
areas represent approximate delineation of coastal exclusive economic zones ( EEZs).
20
3.3 Inventory Summary by Vessel Type
While the scope did not require a spatial or temporal representation of emissions by vessel type,
the STEEM input data did include vessel- type data. This enabled a post- hoc analysis to estimate
emissions contribution by vessel type, as shown in Table 5. This summary represents a
proportionally accurate distribution of global routes solved by STEEM, applied to the domain
inventory in Table 4. Further work with STEEM to produce maps by ship type would perhaps
refine these estimates within the study domain.
Table 5. Estimated Domain Emissions by Vessel Type
Ship Type NOx as NO2 SO2 CO2 HC PM CO Fuel Use
Bulk Carrier 610,000 363,000 21,756,000 18,000 45,300 42,300 5,968,000
Container 964,000 574,000 34,413,000 32,300 81,400 76,100 10,723,000
Fishing 1,000 1,000 51,000 20 60 50 7,000
General Cargo 228,000 136,000 8,152,000 7,800 19,800 18,500 2,601,000
Miscellaneous 45,000 27,000 1,605,000 1,600 4,100 3,800 536,000
Passenger 157,000 94,000 5,614,000 5,900 14,800 13,800 1,948,000
Reefer 60,000 36,000 2,150,000 2,300 5,700 5,400 756,000
RO- RO 213,000 127,000 7,607,000 7,100 17,900 16,800 2,362,000
Tanker 461,000 274,000 16,453,000 15,800 39,900 37,300 5,258,000
Total in Domain 2,740,000 1,630,000 97,800,000 91,000 229,000 214,000 30,160,000
These emissions represent both main and auxiliary engines, as discussed above. However, we
recognize that the distribution of auxiliary engines differs among vessel types in installed power,
fuel type, and emissions.
Table 6. Estimated Percent of Total Emissions from Auxiliary Engines ( AEs)
Ship Type AE NOx AE SOx AE CO2 AE HC AE PM AE CO AE Fuel Use
Bulk Carrier 2.60% 3.04% 3.70% 2.20% 2.60% 2.50% 3.22%
Container 2.40% 2.77% 3.30% 2.00% 2.40% 2.30% 2.93%
Fishing 5.00% 2.58% 7.00% 4.20% 5.00% 4.90% 6.13%
General
Cargo 3.20% 3.67% 4.40% 2.60% 3.10% 3.10% 3.89%
Miscellaneous 5.00% 2.58% 7.00% 4.20% 5.00% 4.90% 6.13%
Passenger 26.90% 33.33% 34.20% 23.30% 26.70% 26.30% 31.25%
Reefer 7.60% 8.76% 10.40% 6.40% 7.50% 7.40% 9.25%
Ro- Ro 3.00% 3.44% 4.10% 2.50% 2.90% 2.90% 3.64%
Tanker 2.90% 3.33% 4.00% 2.40% 2.80% 2.80% 3.53%
3.4 Comparison with Other Emissions Studies ( Task 2)
We compared emissions inventories that we produced using a top- down approach with
ICOADS as the spatial proxy with the inventories produced in this work using STEEM ( 35). In
Figure 2, U. S. Coasts are the areas within the 200 nautical mile Exclusive Economic Zone ( EEZ)
as defined by NOAA in its Office of Coast Survey ( 78); the Great Lakes include Lake Superior,
Lake Michigan, Lake Huron, Lake Erie, Lake Ontario, and connecting waters on both the U. S.
and Canadian sides. Figure 3 shows that the emissions calculated with these two approaches
agree very well for the US East Coast EEZ but differ to varying degrees on the other two coasts
and the Great Lakes ( both U. S. and Canadian side).
21
The amount of SO2 emissions within the Gulf Coast EEZ estimated by the network
approach is 109% higher than the amount estimated with ICOADS; the amounts by the network
approach for the West Coast and the Great Lakes are 32% and 89% lower than the ICOADS
approach. The discrepancies between the two inventories can be explained by geographic
sampling bias of ICOADS which significantly oversamples the Great Lakes and undersamples
the Gulf of Mexico ( 35).
We also compared our results with the inventories from other regional and port emissions
inventories studies ( 76, 79- 82). Figure 4 illustrates the domains of the ports and regions we
compared. The Great Lakes include the lakes and connecting waters within the Canadian
boundary ( 76). “ Western Canada” represents the coastal areas in British Columbia ( B. C.) outside
of the Greater Vancouver Regional District ( GVRD) and Fraser Valley Regional District
( FVRD), and a portion of Washington State, as defined in the Levelton report ( 82). The Port of
Los Angeles, Houston & Galveston area, and the Port of New York and New Jersey ( NYNJ) are
the areas defined by the Starcrest Consulting Group, LLC in its port- wide air emissions
inventory reports ( 79- 81).
Figure 5 shows that the regional/ port air emissions inventories produced with different
approaches look very different. The emissions inventory produced with the top- down approach
using ICOADS as a spatial proxy is significantly higher for the Great Lakes on the Canadian
side, but significantly lower for the “ Western Canada”, the Port of Los Angeles, and the Port of
New York and New Jersey. The conclusion can be drawn that ICOADS is spatially biased as
observed in other studies and small- scale emissions inventories produced with ICOADS as
spatial proxies may be greatly distorted ( 5, 35).
Figure 5 also shows that the amount of emissions estimated by STEEM are higher than
that of the regional/ port studies for the Port of Los Angeles and “ Western Canada”, but lower for
the Great Lakes on the Canadian side, the port of New York and New Jersey, and the Houston
and Galveston area.
We understand that: ( 1) the STEEM captures transit traffic, which might be ignored in
the port- wide studies ( Port of Los Angeles and Western Canada) that used arrivals and
departures of the specific ports ( e. g., the Port of Los Angeles study does not include shipping
activity to other San Pedro Bay ports); ( 2) port- wide studies used more complete arrivals and
departure data for the Great Lakes, the Port of New York and New Jersey, and Houston and
Galveston; ( 3) emissions from dockside hotelling are included in the port- wide studies for the
Port of New York and New Jersey, and the Houston and Galveston area but are not included in
the STEEM results ( the portion of hotelling emissions increases and might dominate the
emissions inventory when the domain becomes smaller around the terminals and when ships
spend less time in transiting); ( 4) the motivation behind the creation of the STEEM was to
improve the emissions inventories from inter- port movements; emissions around ports have to be
adjusted by either plugging in the inventories produced by port- wide studies or modifying the
model itself to include the dockside emissions; ( 5) comparisons showing both higher and lower
port and reigional estimates suggest there is no systemic error in the STEEM; and ( 6) our
assumption that ships generally maneuver within 20 km (~ 12.4 miles) of ports may be
conservative for many ports, since ARB reports recent Automatic Identification System ( AIS)
data suggests that ships may operate at sea- speeds until closer to port.
22
0
20,000
40,000
60,000
80,000
100,000
120,000
140,000
US Gulf
Coast
US East
Coast
US West
Coast
Great
Lakes
tons of SO2 emissions
STEEM ICOADS Top- down
Figure 3. Comparison of the inventories produced with Waterway Network- STEEM and
top- down approach using ICOADS.
Figure 4. Illustration of domains of regional/ port emission inventories studies
23
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
16,000
18,000
20,000
Great
Lakes
( Canada)
" Western
Canada"
POLA
NYNJ
Houston
&
Galveston
Metric tons of emissions
STEEM ICOADS Top- down Port/ Regional Inventory
Figure 5. Comparison of emissions inventories of different approaches; emissions for
Houston & Galveston are NOx, emissions for the other areas are SO2.
We also observe that the emissions from ships carrying foreign cargo within the 200
nautical miles coastal areas of the United States estimated by the STEEM are about five times of
the results estimated by Corbett and Fischbeck using cargo as a proxy ( 10). We understand that
the STEEM is superior to the method used by Corbett and Fischbeck and is more accurate,
consistent with the uncertainty discussion in the earlier paper and with the upward correction of
more accurate work for the Northwest, United States also published previously ( 9).
3.5 Forecasting principles
Forecasts can differ depending on their purposes and scales. Some forecasts look to
reveal where timely investment and action at a local scale or by a single firm can produce the
most benefit ( e. g., profit). Validity of insights is determined by whether recommended actions
produce expected outcomes for a given decision, not whether the forecast trend or future value is
realized. Other forests are intended to be conservative or aggressive; that is, they intend to be
biased to serve the decision makers’ value and tolerance for risk and surprise. This may describe
large scale forecasts such as emissions or trade trends. One challenging class of forecasts may
be considered “ difference” forecasts, where alternative scenarios illustrate how “ a path taken”
may differ from “ a path not taken” rather than to determine which is most probable. These kinds
of forecasts are common in policy domains, such as energy, environment, and economics ( e. g.,
IPPC scenarios). Certainly, freight forecasting presents one challenging example, especially at
the international or multinational scales, and especially when considering policy actions like a
SOx Emissions Control Area ( SECA) under IMO MARPOL Annex VI ( 36).
Admittedly, the quality of forecasts of maritime shipping and trade is limited ( 83), and
thus forecasting of environmental impact from shipping is constrained by the quality of shipping
and trade forecasts. Therefore, we employed a comparison of historic trends and forecast
24
indicators related to maritime trade and energy to provide reasonable insight into a range of
feasible forecasts. Individually, none of these forecasts can be considered more correct than
another, as they represent different assumptions about the relationship between transportation
energy, trade, and North American port activity. However, taken together, they reveal a bounded
range of trends with common insights useful in comparing sulfur controls with no action. We
look for converging growth trends that are representative at several scales ( port, region, coastal,
and national) and informed by historic data. These lead to a set of principles for describing how
freight transport emissions may change:
1. Define the forecast domain broadly through multiple perspectives on freight and economy.
2. Compare global, large regional forecasts with local efforts for converging insights, perhaps
allowing for probabilistic assessment.
3. Include the rear- view mirror in forecasting ( i. e., compare with persistence).
4. Consider first principles involving energy and environment: Some work- energy relationship
must hold if fuel price matters to freight.
5. Make extrapolation adjustments as simple as possible, but no simpler: Assumptions inter-relating
energy, economy, and technology should be checked for potential inconsistencies.
6. Look for surprise, avoid overconfidence: Recognize heterogeneity at all scales; use detailed
scenarios to help broaden or delineate the forecast range, but do not rely on them as likely.
3.6 Activity- based modeling of freight growth
Seaborne cargo activity has increased at significant rates over time. World seaborne
trade growth has increased monotonically except for a short period in the early 1980s ( 66- 69).
Containerized trade is growing faster than global rates. Figure 6 illustrates recent containerized
cargo trends and TEU throughput since 1980. U. S. Maritime Administration ( MARAD)
statistics include cargo on both government and non- government shipments by vessels into and
out of U. S. foreign trade zones, the 50 states, District of Columbia, and Puerto Rico, excluding
postal and military shipments; AAPA statistics describe total container throughput, including
empty container movements. Containerized cargo throughput ( including empty container
movements) grew at ~ 6.5% CAGR since 1985, with imported cargo grow since 1997 at more
than 10% CAGR and total cargo TEUs ( excluding empty container movements) growing at ~ 7%
CAGR since 1997. Given the high- value nature of containerized cargoes, it is not surprising that
these growth trends are most similar to growth in the value of cargo moved, reported by BTS.
Conceptually, growth in seaborne cargo movement should influence ( if not determine)
activity growth in the freight modes ( truck and rail) carrying imports and exports to or from U. S.
metropolitan regions and inland regions. For example, if growth in rail and truck modes is
primarily a result of increasing imports, observed in the U. S. to range between 4.6% and 4.8%
CAGR for all cargoes and between 6% and 9% for containerized ( intermodal) cargoes (~ 6.5%
CAGR for total container throughput including empty containers), then combining these modes
should reflect seaborne trade growth rates ( 71, 84). The multimodal transportation of empty
containers presents a unique challenge in understanding how international goods movement
affects landside freight modes ( 85). Moreover, trucking and rail movements include exported
and domestic freight movements, which are growing at much lower rates than containerized
imports, effectively dampening national growth rates in intermodal freight transportation
compared to port throughput. Considering these activities together helps provide an intuitively
consistent explanation reconciling steeper seaborne trade trends reported in major ports, and
obtained or derived from economic and trade analyses, with less- steep truck and rail freight
25
trends. In other words, we should expect growth rates in goods movement to be shared among
modes because freight transportation is an intermodal network of imports, exports, empty
repositioning, and domestic freight flows. 3
0
5
10
15
20
25
30
35
40
45
1980 1985 1990 1995 2000 2005
Million TEUs
TEU throughput Total Cargo TEUs Export TEUs Import TEUs
Figure 6. Container statistics from U. S. Maritime Administration and American
Association of Port Authorities ( 70, 71).
The U. S. Department of Transportation launched two of the first federal efforts to
consider together multimodal and intermodal freight effects of imported cargoes, generally
through its “ Assessment of the U. S. Marine Transportation System and spatially through the
Freight Analysis Framework ( FAF) ( 86, 87). This work produced a forecast of freight
transportation activity based on trade increases, primarily to identify infrastructure needs rather
than estimate energy and environmental impacts. According to the Freight Analysis Framework
( 87), 4 domestic freight volumes will grow by more than 65 percent from 1998 levels by the year
2020, increasing from 13.5 billion tons ( in 1998) to 22.5 billion tons ( in 2020). This represents a
~ 2.3% compound annual growth rate ( CAGR), similar to that obtained from VMT growth rates
( not adjusted for sales growth) in MOVES ( 88). In other reports, truck freight has doubled since
1980 ( an average annual increase of 3.7%), while domestic waterborne freight has declined by
nearly 30% ( an average annual decline of 1.8%) ( 89). 5 These rates represent the lowest growth
trends we could find in the literature for goods movement.
3 This background discussion does not necessarily imply a direct relationship between energy and emission growth
rates and seaborne trade growth rates; depending on efficiency gains and economies of scale ( e. g., shown for the rail
sector), the rate of change in energy and emissions for ships could be different. This background reinforces the
purpose of and need for the forecasts analysis presented in this report.
4 See Freight Analysis Framework documents at http:// ops. fhwa. dot. gov/ freight/ freight_ analysis/ faf/.
5 BTS Pocket Guide to Transportation 2003, http:// www. bts. gov/ publications/ pocket_ guide_ to_ transportation/ 2003/.
26
Currently, growth factors embedded in U. S. mobile source energy and emissions models
appear to capture better this economic- driven growth in freight transportation. Growth factors
for trucking ( single- unit and combination trucks) in the U. S. EPA’s mobile source models
include a combination of a population ( sales) and VMT growth factors, with adjustments for fuel
economy and other operational factors ( 88). U. S. EPA compared rail freight ton- miles with
railroad distillate fuel consumption data to indicate substantial improvements in rail freight
energy intensity, adjusting emissions based on regulatory requirements ( 90). And, in its 2003
rulemaking, U. S. EPA assumed that freight growth was linked to increased tonnage volume ( 23).
Historic and future growth rates for particular modes are consistent with coupled growth
in economic- energy- emissions trends. For example, U. S. EPA projects that truck population and
VMT will increase by 4.2% to 4.8% CAGR between 2002 and 2025 ( 88). For rail, U. S. EPA
showed that growth rates in cargo ton- miles transported nearly doubled in recent periods, from
~ 2.4% CAGR between 1980 and 1995 to ~ 4.8% CAGR between 1990 and 1995 ( illustrated in
Figure 1- 1 in U. S. EPA’s regulatory support document). In fact, updating observed growth rates
in cargo ton- miles moved by rail to include more recent years reveal a rail- cargo growth rate of
~ 3.6% CAGR from 1985 to 2004 ( 91, 92).
3.5.1 Growth Rates
Most forecasts essentially take historic trends for some recent period and extrapolate with
adjustment for expected change in trends ( e. g., response to economic and population drivers
affecting global trade or consumption). In coming decades, a number of events could modify an
unconstrained growth trend in energy use. In terms of technology, there could be further
improvements in thermal efficiency, fuel type and quality, and propulsion design. In terms of the
economy, we expect shipping cycles to continue to provide periods of slower or negative growth
in oceangoing goods movement ( 83). In terms of logistics operations, trends in containerization
economies of scale and vessel speed over the past three decades could change over the next three
decades if global inventory, energy, and labor costs change.
A simple exponential curve fit to installed power produced an initial growth rate estimate
of 7.1% per year for North America, before averaging with a linear extrapolation. While we
recognized the need for similar adjustment in our forecasts, we hesitated to arbitrarily insert
“ inflection points” in out- year forecasts corresponding to optimistic or pessimistic assumptions.
We acknowledge that an unconstrained exponential curve fit would likely overestimate future
emissions, particularly given expected shipping cycles; we also observed that a linear growth rate
did not match known or expected technology changes relative to cargo growth. A linear trend in
energy use would imply less power required to achieve the cargo throughput – where cargo
volumes are projected to see compounded growth. We don't believe that average technology in
the fleet will change that much from its current path over the next 35 years without strong policy
incentive or substantial changes in fleet energy pricing and supply. Overall, fleet propulsion
technologies will remain more similar than different to the current profile at least through 2040.
Moving more cargo will require more power, in a similar manner to the current fleet ( either
through larger ships, faster ships, more ships, or some combination). Moreover, we did not
identify physical capacity limits to ports or shipping routes ( that are not being addressed through
infrastructure investment) which would constrain trade growth.
Through discussions with ARB, we agreed that the unconstrained exponential trend and
the linear trend define bounding limits for expected change in ship activity. Averaging these
curves defines an arbitrary middle- growth trend, which implicitly describes a mix of positive and
27
negative drivers for ship energy requirements without articulating a detailed scenario of
conditional events. After adjustment, we estimate a growth trend for North America ( including
United States, Canada, and Mexico) of about 5.9%, compounded.
Studies for Southern California ( San Pedro Bay) ports supported this adjustment. These
studies agree that growth in cargo volumes equivalent to 6- 7% compounding annual growth rates
is expected ( 50- 53). 6 Some studies articulate different pathways of growth than simple
extrapolation; for example, the no- net- increase ( NNI) forecast produces nearly the same result
for 2020, but describes substantial increases in the near- term as a result of planned investment in
the ports ( 50). In Figure 7, we show bounding curves ( exponential and linear) and the average
growth curve for Southern California ports. We converted growth trends from the no- net-increase
study and from an unpublished trade- energy model ( by RTI under U. S. EPA direction)
to describe change in installed power and plotted them in Figure 7 with our extrapolation ( 50,
93). 7
These comparisons demonstrate that increasing cargo throughput is related to technology
innovation ( e. g., larger ship sizes, higher speeds, and containerization) that promotes economies
of scale with more powerful ships, more so than increased cargoes determine the number of
voyages. Independent derivations of growth trends all describe at least a doubling of commercial
marine energy use and emissions in California by 2020, corresponding to similar change in the
expected port cargo throughput.
Agreement between the draft trade- energy model by RTI and extrapolation of observed
data is even stronger for containerships. As shown in Figure 8, preliminary results from the draft
RTI trade- energy forecast are more aggressive than our power- based extrapolation. RTI’s trade-energy
model exception to calibrate on inbound containerized cargoes (“ heavy- leg” activity)
may explain this ( 93). Note excellent agreement in RTI draft model results with observed
power- trend history for containership calls to U. S. ports.
These sources of growth trends and forecasts are consistent with and validate our observed trends
in installed power and support our extrapolation of power- based trends to forecast emissions
under business- as- usual ( BAU) conditions. Using our adjusted extrapolation to forecast growth
at ~ 5.9%, we observe that power- based growth rates derived here are comparable to growth rates
for land- based freight modes, by about 1% to 2% ( 45- 48, 71, 84, 88, 91, 92, 94). 8 This
comparison is expected due to the fact that trucking and rail are also engaged in domestic and
intra- continental trade with Canada and Mexico that would not require commercial shipping.
Moreover, our forecast rates are generally lower than dollar- value growth in North American
seaborne trade, and a bit lower than growth in containerized cargo volume. Again, such
comparisons are expected given the importance of bulk cargoes ( liquid and dry) to North
American international trade. In addition, the lower growth in power- based rates compared to
cargo activity provide confirming evidence that economies of scale are improving the energy
intensity and emissions intensity of international shipping – but perhaps by not more than 1% to
2% overall yet. Additional analysis by vessel type could quantify these improvements in more
detail, perhaps discerning relative roles of speed, size, and operational factors ( e. g., average
6 Other studies interpret strong growth in cargo volume to produce a corresponding increase in port calls ( 51, 54).
7 While RTI work is in draft form, U. S. EPA and ARB coordinated discussions and comparisons between this
project and the RTI project. NNI shows only the Southern California ports of Los Angeles and Long Beach, while
the RTI work describes the “ South Pacific” ports, which are considered to be mainly LA and LB but could include
Oakland.
8 Multimodal comparisons are discussed in more detail in Technical Memorandum for Tasks 3- 4.
28
payload utilization rate). Lastly, we observe emissions and energy use by the fastest, most
powerful ships ( containerships) are increasing at the fastest rates, along with demand for
containerized trade.
0
200
400
600
800
1000
1200
1990 1995 2000 2005 2010 2015 2020 2025 2030
Forecasted Transport Energy ( kW)
Unconstrained 7- point Trend
Baseline NNI
This Work
Linear Trend
RTI US South Pacific
Figure 7. South Coast ( South Pacific) growth rates derived from historic data ( 1997- 2003),
showing upper- bound ( exponential), lower- bound ( linear), and average trends. Also shown
are trends from NNI Task Force and from unpublished draft RTI trade- energy model.
0
200
400
600
800
1000
1200
1400
1600
1990 1995 2000 2005 2010 2015 2020 2025 2030
Installed Power Estimate
Average
Observed Container Power
RTI Container
Figure 8. US container growth trends from data extrapolation ( 1997- 2003) and from
unpublished draft RTI trade- energy model.
29
Table 7 presents an overview of power- based growth rates for selected ports and North
American regions. Growth rates for North America, US, Mexico, and Canada use regression
statistics within each vessel type associating gross registered tonnage ( GRT) and rated power to
fill data gaps. General similarity is observed across all regions, with Canada installed- power data
presenting the highest rate of growth and with Mexico presenting the lowest rate of growth.
These growth rates represent an average of unconstrained exponential curve- fits with linear
extrapolation of the data, which implicitly describes an implicit mix of positive and negative
growth drivers. Given that such adjustments may not equally influence growth at different ports
or regions, it is possible that actual growth in emissions will be higher for some places ( and
perhaps lower for others), depending on events that modify unconstrained growth trends over the
next decades.
Table 7. Power- based growth rate summary for commercial ships 2002 - 2020 ( CAGR)
Ports, or Region Emissions Growth Rate
Los Angeles/ Long Beach 5.24%
Oakland/ San Francisco 5.68%
New York/ New Jersey 6.03%
California ( all ports) 5.53%
U. S. West Coast 5.93%
U. S. National 5.86%
Canada 6.57%
Mexico 5.06%
North America ( U. S., Canada and Mexico) 5.86%
1. Growth rates represent an average of exponential and linear fit extrapolations,
presented in terms of compound annual growth rate ( CAGR).
2. US data are from USACE and Lloyds Registry data, per this and other work by
Wang and Corbett.
3. Canada and Mexico data are from Lloyds Movement data ( LMIU)
3.5.2 Growth Patterns
We produced a set of baseline ( Tasks 1 and 2) emissions estimates and forecast estimates
( this work, Tasks 3 and 4) conforming to a consensus domain and resolution appropriate for most
of the atmospheric modeling that will use our North American ship emissions inventory. This
consensus resulted from several meetings with the SECA team. Annual emissions are resolved
into twelve monthly components, following time- resolved patterns in ship activity in North
America, as discussed in the report for Tasks 1 and 2. The North American inventory estimates
for each pollutant uses the following projection parameters from ESRI’s ArcGIS software:
Projection: Equidistant_ Cylindrical
Parameters:
False Easting: 0.0 – default ESRI parameter
False Northing: 0.0 – default ESRI parameter
Central Meridian: 180.0 degrees – UD defined
Standard Parallel_ 1: 0.0 – default ESRI parameter
Linear Unit: User Defined Unit ( 1000 m) – UD defined
Cell Units: kilograms per 16 square kilometers
We delivered inventory files using the following domain:
left - 1000 km, right 18000 km, top 8000 km, bottom 0 km.
30
A hypothetical SECA region conforming to the Exclusive Economic Zone ( EEZ) for
North America was defined for the with- SECA scenarios. Figure 9 shows the model domain and
also reproduces the SOx inventory illustration for the base- year 2002. The scale shown for
emissions is delineated using units common to forecast inventory illustrations discussed below.
Figure 9. Model domain showing hypothetical with- SECA region and baseline 2002 model results.
3.7 Future Emissions without SECA region ( Task 3)
Based on trend comparisons discussed above, we use the following ratios for SOx
forecasts: For 2010, we multiply the 2002 base year inventory by 1.61 times; for 2020, we
multiply the 2002 base year inventory by 2.79 times, corresponding to a growth rate of 5.9%
compounded annually.
For NOx emissions we make adjustment for the introduction of IMO- compliant engines
into the international cargo fleet. We use industry data to estimate ~ 11% percent average
reductions in NOx for new engines complying with MARPOL Annex VI ( 95). This estimated
reduction is similar but slightly lower than assigned in other analyses of IMO- compliant NOx
reductions ( 42, 44); further study into the NOx reduction from uncontrolled to IMO- compliant
engines is ongoing and may produce better per- engine reduction estimates. Introduction rates for
new engines into the fleet are based on fleet scrapping and new ship orders used in previous
work ( 3, 96). Following standard assumptions for the introduction of new engines in the fleet
used of 2% per year, we estimate that about 46% of the fleet in 2010 and about 78% of the fleet
in 2020 will be IMO- compliant. This accounts for fleet- weighted NOx reductions of 5% and
8.4% in 2010 and 2020, respectively, resulting in NOx multiplier ratios of 1.53 for 2010 and 2.55
for 2020.
Per project scope, we considered whether fuel- sulfur content may change in coming
years, e. g., would refining practices result in generally higher fuel- sulfur averages over time as
distillate fuels ( particularly diesel) removed more sulfur. We chose not to make any adjustments
to the average fuel- sulfur content in this work for two reasons. First, we observe very little
change in world- average fuel- sulfur content for residual fuels over the past decade; in fact, most
of the differences may be attributed to better statistical tracking on behalf of MARPOL Annex
VI, more so than real changes in the global average. Second, we recognize that variation in fuel-suflur
content regionally may be greater than the average change over time; we understand that
U. S. EPA is sponsoring study of this issue, and that results of that work are not yet available. If
31
such trends are proven, they can be implemented at the regional level using STEEM in future
work.
An illustration of 2020 emissions without applying any SECA reductions is presented in
Figure 10. Annual and monthly data files for 2010 and 2020 for all forecasted pollutants ( SOx as
SO2, NOx as NO2, CO2, PM, CO, and HC) are provided in both raster and ASCII formats at the
project website ( http:// coast. cms. udel. edu/ NorthAmericanSTEEM/).
Figure 10. Illustration of 2020 ship SOx emissions without SECA reductions.
3.8 Future Emissions with Potential SECA ( Task 4)
To produce with- SECA forecast scenarios, uncontrolled inventories for 2010 and 2020
are modified to depict a reduction in average fuel- sulfur content from 2.7% to 1.5%, a SOx
emissions reduction of about 44%. Only SOx emissions are assumed to change under this SECA
scenario; no additional reductions in primary PM, NOx or other pollutants are calculated. Within
GIS, we select the emissions within the hypothetical SECA region and multiply them by 66% ( 1
minus 44%). Similar to the forecast without SECA, this makes no assumptions for changes in
fuel quality or supply between now and 2020. Such changes could occur through regulatory
action in addition to an IMO- compliant SECA, or through a combination of fuel supply and price
effects not considered in this work. Such considerations could be included in updated forecasts,
based on insights from further ( ongoing) studies. Figure 11 illustrates annual SOx emissions in
2020 depicting compliance with the hypothetical SECA domain.
An illustration of 2020 emissions with SECA reductions is presented in Figure 11.
Annual and monthly files for 2010 and 2020 for SECA- compliant SOx emissions can be found
in both raster and ASCII formats at ( http:// coast. cms. udel. edu/ NorthAmericanSTEEM/). It is
worth noting that sulfur inventories represent stack emissions of gaseous sulfur dioxide ( SO2),
not aerosol sulfate. Ours is a stack emissions inventory, before total fate and transport impacts.
It is inappropriate to pre- process gaseous emissions from the stack within an inventory using
32
some set of assumptions to estimate total PM ( primary plus secondary). Atmospheric modeling
will convert the SO2 gas emissions to sulfate particles needed to estimate total PM health effects.
Figure 11. Illustration of 2020 ship SOx emissions with hypothetical SECA region.
In these forecasts, auxiliary use of distillate fuel was not taken into account when
adjusting for fuel sulfur content. Auxiliary engines only consume a small percentage of marine
fuels and only 29% of auxiliary engine fuels are marine distillate ( 8% of passenger vessel AE
fuels are marine distillate), which on average has 0.57% of sulfur by weight ( varying from 0.05
to 1.5) according to ARB Ocean Going Vessel Survey. We expect this will affect North
American estimates of fuel use and sulfur emissions by 1- 2% ( see baseline inventory estimates
for AEs, Table 6).
33
4.0 DISCUSSION
4.1 Comparison with global forecast trends
For validation, we considered whether analyses at a global scale might yield similar
results. We compared world fleet trends in installed power ( derived from average power by year
of build) with energy trends ( Eyring work and fuel sales), with trade- based historical data ( tons
and ton- miles). Activity- based energy results for similar base- years ( 2001 or 2002) are within
close agreement ( 72, 97, 98). 9 This allows us to index trends to nearly the same value and year,
to index trade- based trends similarly, and to compare these with trends in installed power, as
summarized in Figure 12. Three insights emerge from this global comparison.
1) Extrapolating past data ( with adjustments) produces a range of BAU trends that is
bounded and reveals convergence around a set of similar trends; in other words, while
the range of growth may vary within bounds of a factor of two, one cannot get “ any
forecast they want” out of the data. If we consider that global trade and technology
drivers mutually influence future trends, then we may interpret convergence within the
bounds as describing a likely forecast of global shipping activity.
2) World shipping activity and energy use are on track to double from 2002 by about
2030 (~ 2015 if one considers seaborne trade since 1985, ~ 2050 if one considers
Eyring’s BAU trend). 10 Growth rates are not likely to be reduced without significant
changes in freight transportation behavior and/ or changes in shipboard technology.
3) Confirming earlier discussion, trends in installed power are clearly coupled with trends
in trade and energy. This reinforces the analysis of installed power as a proxy for
forecasting growth, not only for use in baseline inventory estimates.
Coincidentally, averaging bounding extrapolations yields between 3.8% and 4.5% CAGR
growth in installed power, nearly the same 4.1% CAGR as observed for past world seaborne
trade ( 66, 67, 69). In other words, this explains and confirms the use of seaborne trade growth to
project ship fuel use and emissions, as other studies have done. Therefore, we consider this
BAU forecast to be informed by observed past trends and consistent with adjustments intended
to avoid overly aggressive growth estimates. Consistent with the market- forecast principles
reflected in the IMO study, and given the strong relationship observed between cargo moved
( work done) and maritime emissions ( fuel energy used), we estimate that global emissions from
CMVs are increasing at average growth rate of at least 4.1%. This suggests that the rate of
growth in emissions for North America is greater than the global average growth rate.
4.2 Uncertainty and Bounding
There are six types of uncertainty that affect these results. Three primary sources of
uncertainty involving parameters directly used in this study include a) uncertainty in the base-year
estimates; b) uncertainty in the trend used to produce the forecasted inventories; and c)
uncertainty in the patterns of future ship traffic. Additionally, uncertainty arises from factors not
addressed in this work to date – but that could improve future efforts using these methods.
Additional detail could be incorporated to describe better underlying drivers of change in freight
9 An exception is work by Endresen et al, that tends to adjust parameters to agree with international marine fuel sales
statistics; their results are within uncertainty ranges described in other work ( 5, 94, 99).
10 A review of forecast trends for global marine fuel use in unpublished draft results from RTI work suggests that the
trade- energy model developed in parallel with this project falls within these ranges.
34
activity and consumption, to include planned or proposed signals ( e. g., policy action) modifying
vessel activity and propulsion technology, to make alternate assumptions about fleet response in
terms of under- or over- compliance with standards or in terms of price- effects, and to better
depict spatially the asymmetric growth among vessel types and trade routes expected within the
shipping network.
0
0.5
1
1.5
2
2.5
1950 1960 1970 1980 1990 2000 2010 2020 2030
Seaborne Trade ( tons) Seaborne Trade ( ton- miles)
OECD HFO Int'l Sales Seaborne Trade ( trend since 1985)
World Marine Fuel ( Eyring, 2005) Installed Power- This work
Extrapolating trends since ~ 1980- 85
depending on data source
Figure 12. Global indices for seaborne trade, ship energy/ fuel demand, installed power.
35
5.0 SUMMARY AND CONCLUSIONS
5.1 Baseline Inventory
The 2002 inventory of emissions from North American shipping successfully applies
bottom- up estimation methods, extending best- practices for commercial marine inventories to
the largest spatially resolved scale so far, and the STEEM model is capable of conducting similar
analyses for other regions and even globally. STEEM achieves many of the goals of nonroad
marine modeling efforts, such as the U. S. EPA Mobile Vehicle Emissions Simulator
( MOVES). 11 STEEM exceeds MOVES current design in two important ways: 1) our approach
produces spatial and temporal assignment of emissions in GIS; and 2) our model considers
individual vessel movements, rather than binning vessels of similar type. ( Similar to binning by
MOVES, our model applies emissions factor and engine activity assumptions by vessel type, but
considers installed power, routing, and speed individually.)
Our results for U. S. EEZ regions in the North American interport shipping inventory can
be compared to US domestic freight overall, and compared to US domestic marine statistics ( 26,
27). For carbon dioxide, our results in U. S. EEZ regions are 32% of CO2 estimate by U. S. EPA
for all shipping ( coastal and inland ships and boats plus bunkers) and 85% of bunkers only; our
estimates represent 6% of CO2 for U. S. surface freight transportation. Our estimates of CO2
emissions and fuel use conform generally to the expected ratio implicit in the fuel- based CO2
emissions factors, around 3200 tons of CO2 per 1000 tons fuel, suggesting the bunker fuel
comparison with U. S. EPA is most appropriate for cargo ship activity addressed in this study.
The comparison of our work with international bunkers is very good agreement, given the
independent analysis and considering that we do not account for bunkers used in port or for fuel
used on voyages in addition to transits from prior port or to next port. For NOx, our estimates
are 70% of 2002 U. S. EPA NOx estimates from shipping, and represent approximately 12% of
NOx from all U. S. surface freight modes ( heavy- duty diesel truck, locomotive, and marine
including bunkers). For SO2, our estimates in U. S. EEZ regions are 2.5 times greater than
estimated by U. S. EPA for shipping, and 1.2 times greater than SO2 from U. S. surface freight
transportation. For PM2.5, our inventory estimates in U. S. EEZ regions are 1.4 times greater than
current estimates for US shipping, and 34% of U. S. surface freight transportation.
It is important to recognize that at least parts of our inventory may represent shipping not
included in these national inventories, and that our inventory does not include some marine
activity included in these comparison statistics. For example, we do not include inland river
navigation12, and our data does include Canadian and Mexican vessel activity that may transit
within U. S. coastal regions. In this regard, the emissions estimated in this work both augment
and complement current national inventories. Therefore, further work would be required to
evaluate the degree that our inventory may increase existing estimates; therefore, the percentages
resulting from this comparison represent a first- order comparison.
11 See http:// www. epa. gov/ otaq/ ngm. htm for MOVES information.
12 Inland river navigation refers to voyages entirely within inland river regions, typically not navigable by deep- draft
or oceangoing vessels. River transits by deepwater vessels in bays and deepwater river channels are included in this
study ( e. g., in San Francisco Bay to Benicia or Redwood City, or in the Columbia River to the Port of Portland).
36
5.2 Forecast Trends
Important conclusions from this comparison and validation of independent forecast
approaches include the following two points. First, these forecasts are not fundamentally more
or less “ correct” than comparison forecasts, as they all extrapolate observed trends with
adjustments for factors expected to influence future ocean freight activity and ship technologies.
In this regard, insights that result from our analysis of independent forecast models reveal a
range of future scenarios within which our emissions forecasts fall. Second, all models agree
that ship emissions are increasing along with growth in trade, and that these growth trends are
non- linear. Using 2002 as a base year, these models agree under BAU scenarios that energy
used by ships in global trade will double by or before 2020; some scenarios predict doubling
before 2015. Insights support the significant attention that international, federal, state and other
agencies are devoting to understanding the impacts and mitigation options for ocean