													

				

MODELING SULFUR OXIDES (SOx) EMISSIONS TRANSPORT

FROM SHIPS AT SEA

Prepared by

Prakash Karamchandani 

Christian Seigneur

Shu-Yun Chen

Atmospheric & Environmental Research, Inc.

2682 Bishop Drive, Suite 120

San Ramon, CA 94583

Prepared for

U.S. Environmental Protection Agency

Office of Transportation and Air Quality

1200 Pennsylvania Avenue, NW

Washington, DC 20460

Document CP212-06-03a

May 2006

TABLE OF CONTENTS

1. 	Introduction	1-1

2.	Modeling Approach	2-1

	2.1	Air Quality Model (CALPUFF)	2-1

	2.2	Meteorological Model (CALMET)	2-3

	2.3	Approach	2-3

		2.3.1	Receptors	2-4

		2.3.2	Sources	2-4

		2.3.3	Modeling domains	2-5

3.	Model Inputs	3-1

	3.1	Meteorology	3-1

		3.1.1	Measurements	3-2

		3.1.2	MM5 outputs	3-7

		3.1.3	CALMET winds	3-7

		3.1.4	CALMET mixing heights	3-17

	3.2	Emissions	3-17

		3.2.1	Emission rates	3-26

		3.2.2	Stack parameters	3-27

4.	Results		4-1

	4.1	Results for the Southern Pacific U.S. Coastline	4-2

	4.2	Results for the Northern Pacific U.S. Coastline	4-16

	4.3	Results for the Gulf of Mexico Coastline	4-32

	4.4	Results for the Atlantic Ocean Coastline	4-43

5.	Summary and Conclusions	5-1

6.	References	6-1

Appendices

LIST OF TABLES

Table 3-1	Stack characteristics.	3-28

Table 5-1	Percentage of SO2 concentrations below the design value as a
function of the distance from the coastline.	5-2

Table 5-2	Percentage of sulfate concentrations below the design value as
a function of the distance from the coastline.	5-2

LIST OF FIGURES

Figure 2-1	Modeling domain for the Southern Pacific Ocean U.S.
coastline.	2-6

Figure 2-2	Modeling domain for the Northern Pacific Ocean U.S.
coastline.	2-7

Figure 2-3	Modeling domain for the Gulf of Mexico coastline.	2-9

Figure 2-4	Modeling domain for the Atlantic Ocean coastline.	2-10

Figure 3-1	Land-based surface stations and over-water stations for the
Southern Pacific Ocean U.S. coastline	3-3

Figure 3-2	Land-based surface stations and over-water stations for the
Northern Pacific Ocean U.S. coastline	3-4

Figure 3-3	Land-based surface stations and over-water stations for the
Gulf of Mexico coastline	3-5

Figure 3-4	Land-based surface stations and over-water stations for the
Atlantic Ocean coastline	3-6

Figure 3-5	MM5 modeling domain	3-8

Figure 3-6a	Wind roses based on CALMET outputs for the southern Pacific
Ocean during winter and spring 2002	3-9

Figure 3-6b	Wind roses based on CALMET outputs for the southern Pacific
Ocean during summer and fall 2002	3-10

Figure 3-7a	Wind roses based on CALMET outputs for the northern Pacific
Ocean during winter and spring 2002	3-11

Figure 3-7b	Wind roses based on CALMET outputs for the northern Pacific
Ocean during summer and fall 2002	3-12

Figure 3-8a	Wind roses based on CALMET outputs for the Gulf of Mexico
during winter and spring 2002	3-13

Figure 3-8b	Wind roses based on CALMET outputs for the Gulf of Mexico
during summer and fall 2002	3-14

Figure 3-9a	Wind roses based on CALMET outputs for the Atlantic Ocean
during winter and spring 2002	3-15

Figure 3-9b	Wind roses based on CALMET outputs for the Atlantic Ocean
during summer and fall 2002	3-16

Figure 3-10a	Mixing heights for the southern Pacific Ocean during winter
and spring 2002	3-18

Figure 3-10b	Mixing heights for the southern Pacific Ocean during summer
and fall 2002	3-19

Figure 3-11a	Mixing heights for the northern Pacific Ocean during winter
and spring 2002	3-20

Figure 3-11b	Mixing heights for the northern Pacific Ocean during summer
and fall 2002	3-21

Figure 3-12a	Mixing heights for the Gulf of Mexico during winter and
spring 2002	3-22

Figure 3-12b	Mixing heights for the Gulf of Mexico during summer and
fall 2002	3-23

Figure 3-13a	Mixing heights for the Atlantic Ocean during winter and
spring 2002	3-24

Figure 3-13b	Mixing heights for the Atlantic Ocean during summer and
fall 2002	3-25

Figure 4-1	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 125 km from the Southern Pacific U.S.
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel 	4-3

Figure 4-2	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 125 km from the Southern Pacific U.S.
coastline 	4-4

Figure 4-3	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Southern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel 	4-5

Figure 4-4	Cumulative frequency distribution of design ratios of sulfate
concentrations from ships at 125 km from the Southern Pacific U.S.
coastline 	4-6

Figure 4-5	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 250 km from the Southern Pacific U.S.
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel 	4-8

Figure 4-6	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 250 km from the Southern Pacific U.S.
coastline 	4-9

Figure 4-7	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Southern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel 	4-10

Figure 4-8	Cumulative frequency distribution of design ratios of sulfate
concentrations from ships at 250 km from the Southern Pacific U.S.
coastline 	4-11

Figure 4-9	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 375 km from the Southern Pacific U.S.
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel 	4-12

Figure 4-10	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 375 km from the Southern Pacific U.S.
coastline 	4-13

Figure 4-11	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 375 km from the Southern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel 	4-14

Figure 4-12	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 375 km from the Southern Pacific
U.S. coastline 	4-15

Figure 4-13	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 500 km from the Southern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel 	4-17

Figure 4-14	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 500 km from the Southern Pacific
U.S. coastline 	4-18

Figure 4-15	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 125 km from the Northern Pacific U.S.
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel 	4-19

Figure 4-16	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 125 km from the Northern Pacific U.S.
coastline 	4-20

Figure 4-17	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel 	4-22

Figure 4-18	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 125 km from the Northern Pacific
U.S. coastline 	4-23

Figure 4-19	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 250 km from the Northern Pacific U.S.
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel 	4-24

Figure 4-20	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 250 km from the Northern Pacific U.S.
coastline 	4-25

Figure 4-21	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel 	4-26

Figure 4-22	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 250 km from the Northern Pacific
U.S. coastline 	4-27

Figure 4-23	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 375 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel 	4-28

Figure 4-24	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 375 km from the Northern Pacific
U.S. coastline 	4-29

Figure 4-25	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 500 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel 	4-30

Figure 4-26	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 500 km from the Northern Pacific
U.S. coastline 	4-31

Figure 4-27	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 125 km from the Gulf of Mexico
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel 	4-33

Figure 4-28	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 125 km from the Gulf of Mexico coastline 
4-34

Figure 4-29	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel 	4-35

Figure 4-30	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 125 km from the Gulf of Mexico
coastline 	4-36

Figure 4-31	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 250 km from the Gulf of Mexico
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel 	4-37

Figure 4-32	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 250 km from the Gulf of Mexico coastline 
4-38

Figure 4-33	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel 	4-39

Figure 4-34	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 250 km from the Gulf of Mexico
coastline 	4-40

Figure 4-35	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 375 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel 	4-41

Figure 4-36	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 375 km from the Gulf of Mexico
coastline 	4-42

Figure 4-37	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 500 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel 	4-44

Figure 4-38	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 500 km from the Gulf of Mexico
coastline 	4-45

Figure 4-39	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 125 km from the Atlantic Ocean
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel 	4-46

Figure 4-40	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 125 km from the Atlantic Ocean coastline 
4-47

Figure 4-41	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel 	4-48

Figure 4-42	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 125 km from the Atlantic Ocean
coastline 	4-49

Figure 4-43	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 250 km from the Atlantic Ocean
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel 	4-50

Figure 4-44	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 250 km from the Atlantic Ocean coastline 
4-51

Figure 4-45	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel 	4-52

Figure 4-46	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 250 km from the Atlantic Ocean
coastline 	4-53

Figure 4-47	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 375 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel 	4-55

Figure 4-48	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 375 km from the Atlantic Ocean
coastline 	4-56

Figure 4-49	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 500 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel 	4-57

Figure 4-50	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 375 km from the Atlantic Ocean
coastline 	4-58

EXECUTIVE SUMMARY

A screening study was conducted to determine the air quality impacts
(annual average ground-level concentrations of SO2 and sulfate) on land
due to SOx emissions from ships burning high-sulfur fuel at sea at
various distances from the coastline.  The CALPUFF dispersion model was
used for this screening study.  Meteorological inputs were prepared with
the CALMET model using the outputs of a prognostic meteorological model,
MM5, in combination with surface measurements over water and on land. 
The meteorology represents the year 2002 because it was the most recent
year for which an MM5 simulation covering the contiguous United States
was available.  CALPUFF tends to overestimate the conversion of SO2 to
sulfate in the gas phase and the results presented here are likely to
provide conservative estimates of the impacts of emissions from ships at
sea on inland air quality (because of the simplified treatment of
aqueous-phase chemistry in CALPUFF, this overestimation may not hold for
cases where the interactions of the ship plumes with fog dominate
sulfate formation). 

Four domains were studied: the southern Pacific coastline, the northern
Pacific coastline, the Gulf of Mexico coastline and the Atlantic
coastline.  The results were compared with those calculated for ships
burning low-sulfur fuel at the coastline to determine upper bounds for
Sulfur Emission Control Areas (SECAs), i.e., off-shore distances at
which the switch to high-sulfur fuel would not impair air quality.  For
each offshore distance investigated, the percentage of receptors for
which the air quality impacts of ships at sea were lower than the
impacts of ships at the coastline was calculated.

Emission rates were estimated to be representative of ocean-going ships
along U.S. coastlines.  The sulfur content of the fuel was assumed to be
15,000 ppm within the SECA (i.e., here at the coastline) and 27,000 ppm
outside the SECA (i.e., at the four off-shore distances considered here,
125, 250, 375 and 500 km).  The gas-phase SO2 and particulate-phase
sulfate emissions per ship were estimated to be 100,320 g/h and 3.040
g/h, respectively, within the SECA and 180,640 g/h and 5,600 g/h,
respectively, outside the SECA.  Based on an analysis of ship traffic
off the Pacific coastline, a distance of 25 km between ships was used
for all coastlines.

The results are summarized in Tables E-1 and E-2 for concentration
ratios of SO2 and sulfate, i.e., the ratio of the concentration
calculated for ships at sea to the concentration calculated for ships at
the coastline (the design value).

The results for SO2 were different from those for sulfate, primarily due
to differences in the behavior of these two species downwind of a
source.  For all the coastlines studied, the majority of the SO2
concentration ratios were less than one at shorter off-shore distances
than for sulfate.  Thus, sulfate concentration ratios were the limiting
factor for defining the upper bounds of the SECA for each coastline. 

The results showed some differences in results among the various
coastlines studied. These differences are due to differences in the wind
fields bringing the offshore ship emissions and their secondary products
to land as well as differences in precipitation, which removes
pollutants from the atmosphere.

The results from the two Pacific Ocean coastline simulations were
qualitatively similar.  For both Pacific Ocean coastlines, over 90% of
the receptors showed SO2 concentration ratios less than one for ships at
250 km from the coastline.  For sulfate, only about 49% and 56% of the
receptors had concentrations less than one for ships at 500 km from the
southern Pacific Ocean and northern Pacific Ocean coastlines,
respectively.

For the other two coastlines (Atlantic Ocean and Gulf of Mexico), the
SO2 results were qualitatively similar to those for the Pacific Ocean
coastlines, i.e., over 90% of the receptors showed SO2 concentration
ratios less than one for ships at 250 km from the coastline.  However,
there were some large differences for sulfate.  For the Gulf of Mexico
coastline, over 70% of the receptors showed sulfate concentration ratios
less than one for ships at 250 km from the coastline.  For the Atlantic
Ocean coastline, nearly 60% of the receptors showed sulfate
concentration ratios less than one for ships at 250 km from the
coastline.

Table E-1.	Percentage of SO2 concentrations below the design value as a
function of the distance from the coastline.

Distance from coastline	125 km	250 km	375 km	500 km

Southern Pacific	40.7%	90.7%	100%	100%

Northern Pacific	46.6%	97.9%	100%	100%

Gulf of Mexicoa	84.4%	98.1%	100%	100%

Atlantic	86.6%	100%	100%	100%

aNote that Florida values correspond to a shorter ship-coastline
distance and the values presented in the table should be seen as lower
limits.

Table E-2.	Percentage of sulfate concentrations below the design value
as a function of the distance from the coastline.

Distance from coastline	125 km	250 km	375 km	500 km

Southern Pacific	4.4%	24.9%	41.9%	48.7%

Northern Pacific	0.01%	3.6%	20.3%	55.7%

Gulf of Mexicoa	40.4%	72.0%	80.5%	84.0%

Atlantic	1.2%	57.9%	92.5%	100%

aNote that Florida values correspond to a shorter ship-coastline
distance and the values presented in the table should be seen as a lower
limits.



These results suggest that an off-shore distance of 500 km should be
sufficient when conducting refined modeling of the potential impacts of
ship emissions on air quality inland, if a criterion of about 50% of
inland receptors having sulfate concentrations below the design value is
acceptable to define the SECA.

INTRODUCTION

This document describes a screening study to model on-shore SO2 and
sulfate concentrations due to emissions of SOx from ships at sea.  The
objective of this screening study is to obtain quantitative information
on the shortest distance at which ships burning high sulfur fuel (fuel
content of 27,000 ppm) will have air quality impacts at land receptors
that are less than those anticipated from emissions from ships burning
low sulfur fuel (fuel content of 15,000 ppm) within coastal waters. 
This distance can subsequently be used as the basis for defining the
modeling domain for a more refined Eulerian modeling study using the
U.S. EPA Community Multiscale Air Quality model (CMAQ).  The results of
the CMAQ modeling will yield the information required to define the
outer boundary of a Sulfur Emission Control Area (SECA) for various U.S.
coastlines.  Because of differences in meteorology and other factors
governing the transport and transformation of ship emissions among the
various coastlines, each coastline is modeled separately in the
screening study described here.

A review of available models and data was conducted prior to defining
our modeling approach (Seigneur et al., 2005a; see Appendix A).  The
modeling approach was then formally documented in an analysis plan that
was reviewed by EPA (Seigneur et al., 2005b; see Appendix B).

This report is organized as follows. Section 2 describes the modeling
approach, including brief descriptions of the air quality model
(CALPUFF) used for the screening study, and the meteorological
preprocessor for CALPUFF, referred to as CALMET.  CALPUFF is recommended
by EPA for regulatory applications to assess the long-range transport of
pollutants.  While CALPUFF has some limitations, as discussed in Section
2, it is suitable for a screening study since it will tend to
overestimate the oxidation of SO2 to sulfate in the gas phase
(Karamchandani et al., 2006) and may thus provide a conservative bound
for the distance of interest for defining the SECA (because of the
simplistic treatment of aqueous-phase chemistry in CALPUFF, one cannot
assess whether sulfate concentrations would be overestimated if fog
processes dominate sulfate formation). Section 3 describes the
development of meteorological, emissions and geophysical data inputs for
the CALPUFF simulations. Section 4 presents the results for the various
U.S. coastlines that were simulated, and Section 5 provides a summary of
the study and presents some conclusions.



MODELING APPROACH

2.1	Air Quality Model (CALPUFF)

We used the EPA-recommended long-range transport model, CALPUFF, for
this screening study.  CALPUFF is a multi-layer, multi-species
non-steady-state puff dispersion model that can simulate the effects of
time- and space-varying meteorological conditions on pollutant
transport, transformation, and removal. It can accommodate arbitrarily
varying point source, area source, volume source, and line source
emissions. It is intended for use on scales from tens of meters to
hundreds of kilometers from a source.

Detailed descriptions of the formulation and features of CALPUFF are
provided in the CALPUFF documentation (Scire et al., 2000a).  Here, we
briefly summarize some of the features of CALPUFF that are relevant to
our study and discuss the limitations of CALPUFF in its treatment of
atmospheric chemistry.  CALPUFF includes algorithms for near-source
effects such as building downwash, transitional plume rise, partial
plume penetration, sub-grid scale terrain interactions as well as longer
range effects such as pollutant removal due to wet and dry deposition,
simplified chemical transformations, vertical wind shear, over-water
transport and coastal interaction effects.  Because the latter features
were relevant to simulating the transport and chemistry of SOx emissions
from ships, they were activated for our study.

CALPUFF offers several options to simulate the formation of secondary
sulfate and nitrate particles from the oxidation of the emitted primary
gaseous pollutants, SO2 and NOx respectively.  Since the oxidation of
SO2 to sulfate was of interest for this study, we selected the more
advanced chemistry module available in CALPUFF, which is based on the
RIVAD/ARM3 chemical mechanism (Morris et al., 1988).  This option treats
the NO and NO2 conversion processes in addition to the NO2 to inorganic
nitrate and SO2 to sulfate conversions.  The scheme assumes low
background VOC concentrations and is not suitable for urban regions. 
The NO-NO2-O3 chemical system is first solved to get pseudo-steady-state
concentrations of NO, NO2, and O3.  During the day, this system consists
of the NO2 photolysis reaction to yield NO and O3 and the NO-O3
titration reaction to yield NO2.  During the night, only the NO-O3
titration reaction is considered.

In the implementation of the RIVAD/ARM3 scheme in CALPUFF, the
background O3 concentration is used as the initial O3 concentration at
each puff chemistry time step (i.e., the plume O3 concentration does not
evolve as a function of the downwind distance but instead it is
replenished at each time step).  This may lead to errors if the sources
that are being simulated are large NOx emitters.  For such sources, the
high NO concentrations in the plume deplete the O3 concentrations near
the source and, as a result, OH concentrations are very low and the
gas-phase rates of NO2 and SO2 oxidation to HNO3 and H2SO4,
respectively, are negligible (Karamchandani et al., 1998; Karamchandani
and Seigneur, 1999).  In CALPUFF, the lack of depletion of O3 in the
plume leads to an overestimate of the steady-state daytime concentration
of the hydroxyl radical, OH, which is calculated from the final O3
concentration after the solution of the NO-NO2-O3 system and is,
therefore, also overestimated in the near field.  Because the OH
concentrations are overestimated, CALPUFF overestimates the rates of
formation of HNO3 and H2SO4 in the near field.

CALPUFF uses dry deposition velocities to calculate the dry deposition
of gaseous and particulate pollutants to the surface.  These dry
deposition velocities can either be user-specified or calculated
internally in CALPUFF using a resistance-based model.  For this study,
we selected the latter option to calculate dry deposition velocities. 
For gaseous pollutants, the resistances that are considered are the
atmospheric resistance, the deposition layer resistance, and the canopy
resistance.  For particles, a gravitational settling term is included
and the canopy resistance is assumed to be negligible.  The various
resistances and particle settling rates are calculated as functions of
atmospheric variables (e.g., stability and wind speed), surface
characteristics (e.g., surface roughness, vegetation type, physiological
state), and the properties of the depositing material (gas diffusivity,
solubility, and reactivity; particle size, shape, and density).

CALPUFF uses the scavenging coefficient approach to parameterize wet
deposition of gases and particles.  The scavenging coefficient depends
on pollutant characteristics (e.g., solubility and reactivity), as well
as the precipitation rate and type of precipitation.  The model provides
default values for the scavenging coefficient for various species and
two types of precipitation (liquid and frozen).

2.2	Meteorological Model (CALMET)

The recommended meteorological inputs for applying CALPUFF are the
time-dependent outputs of CALMET, a meteorological model that contains a
diagnostic wind field module and overwater and overland boundary layer
modules (Scire et al., 2000b).  The outputs of CALMET are hourly gridded
fields of micro-meteorological parameters and three-dimensional wind and
temperature fields.  The wind field module in CALMET combines an
objective analysis procedure using wind observations with parameterized
treatments of slope flows, valley flows, terrain kinematic effects,
terrain blocking effects, and sea/lake breeze circulations. The boundary
layer modules of CALMET produce gridded fields of micrometeorological
parameters, such as friction velocity, convective velocity scale, and
Monin-Obukhov lengths, as well as mixing heights and PGT stability
classes.

Inputs to CALMET include surface and upper air meteorological data. 
Optionally, CALMET can also use the outputs of prognostic meteorological
models, such as MM5 and CSUMM, to supplement observations and create the
meteorological fields required by CALPUFF.  A processor (CALMM5) is
available to convert MM5 data to the format required for CALMET.  For
this study, we used the U.S. EPA’s MM5 simulation outputs for 2002. 
The MM5 domain contains the entire contiguous United States and portions
of Canada and Mexico and extends out to the Pacific Ocean in the west,
the Gulf of Mexico to the south and the Atlantic Ocean in the east. 
Thus, MM5 results for all the coastlines relevant to our study were
available from the EPA.  Section 3 provides additional details on the
preparation of the meteorological data inputs for CALPUFF for this
study.

2.3	Approach

For each coastline, a number of annual CALPUFF simulations were
conducted.  We used 2002 as our reference year because it corresponds to
the most recent year for which an MM5 simulation covering all coastlines
was available.  The first simulation for each coastline was to establish
the target values of annual-average SO2 and sulfate concentrations at an
array of inland receptors (the placement of the receptors is described
in Section 2.3.1 below).  These target values correspond to emissions
from ships at dockside, i.e., those ships that are within the SECA and,
therefore, will likely have to burn low sulfur fuel (15,000 ppm fuel
content).  Then, we conducted annual CALPUFF simulations for ships
located at various distances from the coastline.  For these simulations,
the ship emissions used were those based on ships burning high sulfur
fuel (i.e., 27,000 ppm fuel content).  The comparisons of annual-average
concentrations of SO2 and sulfate from these simulations with the target
values determined previously provide a basis for defining the SECA
modeling domain for the CMAQ simulations to be conducted by EPA.

In the following sections, we provide additional details on the
placement of receptors and sources for the simulations as well as the
modeling domain for each coastline.

	2.3.1	Receptors

Ground-level receptors were located on land along the coastline and at
various distances from the coastline.  The first line of receptors was
located along the coastline, with a distance of 10 km between adjacent
receptors.  This distance provides a finer spatial resolution than that
of the ship emissions (see Section 3).  Nine additional lines of
receptors were then located inland parallel to the coastline receptors. 
The distances between adjacent lines of receptors were variable, with
higher resolutions near the coastline and coarser resolutions further
inland.  The first line of inland receptors was located at 10 km from
the coastline receptors, while the last line of receptors was located at
240 km (about 150 miles) from the coastline.  Depending on the coastline
being simulated, the total number of receptors varied from about 1100 to
2800.

	2.3.2	Sources

Ship emissions were represented by a set of stationary point sources. 
Each point source represents one ship.  The use of stationary sources to
represent moving ships is an appropriate approximation for this
screening modeling study, because using stationary sources will
overestimate the downwind air quality impacts (emissions will be
concentrated in specific locations rather than continuously distributed
along the shipping lane, thereby leading to greater ambient air
concentrations).

For each simulation, the sources were located at a selected distance
from shore (along the coastline for the target value simulations, and at
125 km, 250 km, 375 km, and 500 km from the coastline for the SECA
boundary simulations).  The spacing between adjacent ships for a given
simulation was determined from ship traffic and estimated shipping lane
density (see Section 3).  We maintained the same number of ships for the
at-sea emissions scenarios as the number of ships for the dockside
emissions scenario.  This number varied from about 40 to 100 depending
on the coastline being simulated.

	2.3.3	Modeling domains

The following U.S. coastlines were simulated in this screening study:

Southern Pacific Ocean coastline

Northern Pacific Ocean coastline

Gulf of Mexico coastline

Atlantic Ocean coastline

All modeling domains were selected to allow an off-shore distance of at
least 500 km from the coastline to include all ship-at-sea scenarios
(see above) and an inland distance of at least 240 km from the coastline
to provide sufficient spatial coverage for calculating air quality
impacts.

The modeling domain for the Southern Pacific U.S. coastline extends from
about 30 degrees North to 38 degrees North and includes Southern
California, the Central California coastline, and the southern portion
of Northern California.  Figure 2-1 shows this domain as well as the
locations of coastline ships for the target value simulation and the
locations of the receptors at the coastline and inland where SO2 and
sulfate concentrations were calculated.

Figure 2-2 shows the domain for the Northern Pacific U.S. coastline,
which extends from about 35 degrees North to about 52 degrees North and
includes Northern California, 

Figure 2-1.	Modeling domain for the Southern Pacific Ocean U.S.
coastline.

Figure 2-2.	Modeling domain for the Northern Pacific Ocean U.S.
coastline.

Oregon, and Washington in the U.S., and the southern half of the
British Columbia coastline in Canada.

In the east-west direction, the modeling domains for both the Southern
Pacific and Northern Pacific coastlines begin at the western boundary of
the modeling domain for the MM5 simulations that provided the hourly 3-D
meteorological inputs for our study (see Section 3).  The eastern
boundaries of the modeling domains for the Southern and Northern Pacific
coastline extend to about 115 degrees West and 112 degrees West,
respectively.

The modeling domain for the Gulf of Mexico coastline is shown in Figure
2-3.  The east-west extent of the modeling domain is from Western Texas
(about 105 degrees West) in the west to Florida and the Atlantic Ocean
(about 75 degrees West).  The southern boundary of the modeling domain
coincides with the southern boundary of the MM5 domain while the
northern boundary is at about 34 degrees North.

The southern boundary of the Atlantic Ocean modeling domain, shown in
Figure 2-4, also coincides with the southern boundary of the MM5 domain.
 The northern boundary is at about 52 degrees North, just a few degrees
lower than the northern boundary of the MM5 domain.  The western
boundary of the Atlantic Ocean domain is at about 85 degrees West, while
the eastern boundary coincides with the eastern boundary of the MM5
domain.

Figure 2-3.	Modeling domain for the Gulf of Mexico coastline.

Figure 2-4.	Modeling domain for the Atlantic Ocean coastline.



MODEL INPUTS

The following inputs were required for the CALPUFF simulations

Meteorology

Emissions

Land use and terrain elevation data

Coastline data

Terrain elevation data at 1 degree DEM (Digital Elevation Model)
resolution were downloaded from the U.S. Geological Survey (USGS).  Land
use data were also obtained from the USGS at a 1:250,000 scale.  These
data were processed by the CALPUFF/CALMET preprocessors: TERREL,
CTGCOMP, CTGPROC, and MAKEGEO.

 Coastline data were obtained using the ZXPLOT package from the Center
for the Analysis and Prediction of Storms at the University of Oklahoma.

The preparation of meteorological and emission inputs for the CALPUFF
simulations is described below.

3.1	Meteorology

As described in Section 2.2, CALMET is the companion meteorological
model that is used to prepare the meteorological fields used by CALPUFF.
 CALMET is a diagnostic meteorological model that can use standard
surface and upper air meteorological data, and also has an over-water
option that allows the use of special over-water measurements for grid
cells that are over the ocean.  In addition, CALMET can use 3-D gridded
meteorological fields from prognostic models, such as MM5, to either
supplement observations or to provide an initial guess field for the
diagnostic procedure.

For this study, we used a combination of land-based surface
measurements, over-water measurements, and MM5 outputs to create the
CALPUFF meteorological fields.  The MM5 fields provide the vertical
structure with sufficient temporal (hourly) and spatial (36 km)
resolution to supplement the surface measurements.  This approach
provides consistency with the subsequent grid-based modeling that will
be conducted by OAQPS to define SECAs using CMAQ, because CMAQ will be
driven with the MM5 meteorology.  It also addresses a weakness in the
official release of CALMET (Scire et al., 2005) in its calculation of
mixing heights over-water surfaces.  The mixing height algorithm in
CALMET underestimates over-water mixing heights, especially during light
wind conditions over warm water, since it only calculates
mechanically-derived mixing over water surfaces.  This weakness has been
corrected in a new version of CALMET described by Scire et al. (2005). 
However, this new version was not available to us at the time we
performed the CALMET simulations for this study.  Based on our
discussions with the CALMET developers (Scire, 2005), we used the MM5
mixing heights directly in CALMET. 

	3.1.1	Measurements

The National Climatic Data Center (NCDC) Integrated Surface Hourly
Observations database provided the land-based hourly surface
measurements.  These include wind speed, wind direction, temperature,
and dew point temperature.  Over-water measurements were available for
all the coastlines from the National Data Buoy Center (NDBC)
(http://www.ndbc.noaa.gov).  These measurements are taken from buoys. 
The buoys are at varying distances from the coast.  Those near the coast
are frequently near harbors or bays.  Most of the buoys are owned and
operated by NDBC but there are also several other agencies that submit
their data to the NDBC database.  The over-water measurement coverage is
sparse, as shown in Figures 3-1 through 3-4, which show the locations of
the land-based surface and over-water measurements for each of the
coastlines that were simulated in this study.  We used 2002 observations
for consistency with the MM5 outputs (see below). 



Figure 3-1.	Land-based surface stations and over-water stations for the
Southern Pacific Ocean U.S. coastline.

Figure 3-2.	Land-based surface stations and over-water stations for the
Northern Pacific Ocean U.S. coastline.

Figure 3-3.	Land-based surface stations and over-water stations for the
Gulf of Mexico coastline.

Figure 3-4.	Land-based surface stations and over-water stations for the
Atlantic Ocean coastline.

	3.1.2	MM5 outputs

We used the outputs of the 2002 MM5 simulations sponsored by the U.S.
EPA to supplement the meteorological measurements.  These outputs were
provided to us by ENVIRON Corporation.  The MM5 modeling domain, shown
in Figure 3-5, covers the entire contiguous United States and extends
significantly over the oceans.  The horizontal spatial resolution for
the MM5 outputs is 36 km.

An interface program (CALMM5) converts the MM5 data into a form
compatible with CALMET.  A beta version (not yet officially approved by
the EPA) of CALMM5 processes MM5 Version 3 output data directly.  This
processor is available from the CALPUFF-CALMET Download BETA-Test page.

	3.1.3	CALMET Winds

As mentioned above, the wind fields were calculated with CALMET using
the outputs of MM5 in combination with available data.  To illustrate
the variability of wind speed and direction with location and season, we
present wind roses over the ocean based on the calculated CALMET wind
fields for the southern Pacific coastline, northern Pacific coastline,
Gulf of Mexico coastline and Atlantic coastline in Figures 3-6, 3-7, 3-8
and 3-9, respectively.

In the southern Pacific Ocean, winds are mostly from the southwest
except during winter in the northern part of the domain where the wind
direction is more variable.  In the northern Pacific, the prevailing
winds are from the west in the southern part of the domain (i.e., off
the coast of California and Oregon) during spring, summer and fall. 
Wind direction is variable during winter in the southern part of the
domain and for all seasons in the northern part of the domain.

In the Gulf of Mexico, the wind direction varies with season and
location.  During winter, in the western part of the domain, the
prevailing winds are from the southwest and the northeast; they are
mostly from the west and north in the central part of the domain and
with more variable direction near the Florida coast.  During spring and 



Figure 3-5.	MM5 modeling domain.

Figure 3-6a.	Wind roses based on CALMET outputs for the southern Pacific
Ocean during winter 2002 (top) and spring 2002 (bottom).

Figure 3-6b.	Wind roses based on CALMET outputs for the southern Pacific
Ocean during summer 2002 (top) and fall 2002 (bottom).

Figure 3-7a.	Wind roses based on CALMET outputs for the northern Pacific
Ocean during winter 2002 (top) and spring 2002 (bottom).

Figure 3-7b.	Wind roses based on CALMET outputs for the northern Pacific
Ocean during summer 2002 (top) and fall 2002 (bottom).

Figure 3-8a.	Wind roses based on CALMET outputs for the Gulf of Mexico
during winter 2002 (top) and spring 2002 (bottom).

Figure 3-8b.	Wind roses based on CALMET outputs for the Gulf of Mexico
during summer 2002 (top) and fall 2002 (bottom).

Figure 3-9a.	Wind roses based on CALMET outputs for the Atlantic Ocean
during winter 2002 (left) and spring 2002 (right).

Figure 3-9b.	Wind roses based on CALMET outputs for the Atlantic Ocean
during summer 2002 (left) and fall 2002 (right).

summer, the winds in the western and central parts of the domain are
mostly from the north, but they are variable in direction near the
Florida coast (the prevailing wind direction varies from
north-north-east in the western part of the domain to north-north-west
in the eastern part of the domain).  During fall, the winds are more
variable with a tendency to be from the west to north-east in the
western part of the domain and from the north to north-west in the
central and eastern parts of the domain.

In the Atlantic Ocean, winds are mostly from the south to south-west in
the southern part of the domain.  They are more variable in the northern
part of the domain with a prevailing northern trend that evolves from a
northeastern direction during winter to a northwestern direction during
summer.

	3.1.4	CALMET Mixing heights

As mentioned above, the CALMET mixing heights were obtained from the MM5
outputs.  They vary spatially and temporally.  We illustrate such
variability in Figures 3-10 through 3-13 where seasonally-averaged
mixing heights are depicted for the southern Pacific coast, northern
Pacific coast, gulf of Mexico coast and Atlantic coast, respectively.

Mixing heights are lowest in winter (December – February) and highest
in summer (June – August).  They are lower over water than over land;
they also tend to be greater over the Gulf of Mexico than over the
Pacific and Atlantic oceans.  Ship emissions were predominantly released
after plume rise within the mixing layer.

3.2	Emissions

As described in Section 2.3.2, ship emissions were represented by a set
of stationary point sources.  The point source emissions information
required for the CALPUFF simulations include stack locations, stack
characteristics such as stack heights and stack flow rates, and emission
rates of SO2, sulfate, NO and NO2.

Figure 3-10a.	Mixing heights for the southern Pacific Ocean during
winter 2002 (top) and spring 2002 (bottom).

Figure 3-10b.	Mixing heights for the southern Pacific Ocean during
summer 2002 (top) and fall 2002 (bottom).

Figure 3-11a.	Mixing heights for the northern Pacific Ocean during
winter 2002 (top) and spring 2002 (bottom).

Figure 3-11b.	Mixing heights for the northern Pacific Ocean during
summer 2002 (top) and fall 2002 (bottom).

Figure 3-12a.	Mixing heights for the Gulf of Mexico during winter 2002
(top) and spring 2002 (bottom).

Figure 3-12b.	Mixing heights for the Gulf of Mexico during summer 2002
(top) and fall 2002 (bottom).

Figure 3-13a.	Mixing heights for the Atlantic Ocean during winter 2002
(left) and spring 2002 (right).

Figure 3-13b.	Mixing heights for the Atlantic Ocean during summer 2002
(left) and fall 2002 (right).



	3.2.1	Emission rates

Emission factors are needed to estimate the emissions of SOx (gas-phase
SO2 and particulate-phase sulfate) associated with various ship
activities.  Based on the review of available emission factors of
Seigneur et al. (2005), the most recent EPA emission factors were
selected (EPA, 2002).  Those emission factors pertain to ships with
engines with displacement exceeding 30 liters (so-called Category 3
engines).  Emission factors are reported for three different engine
types (slow speed, medium speed and steam boiler) for transit modes and
hoteling modes.  For this study of ships at sea, we are interested in
medium speeds for transit modes.

The SO2 emission factor per unit of work is reported to be 9.56 g/hp-h
for a 3% sulfur fuel (i.e., 30,000 ppm) for a ship at slow or medium
speed in transit mode.  This is equivalent to 12.8 g/kW-h.  For a ship
within the SECA, we assumed a fuel sulfur content of 15,000 ppm,
resulting in an emission factor of 6.4 g/kW-h.  For ships at sea outside
of the SECA, a fuel sulfur content of 27,000 ppm was assumed. 
Therefore, the emission factor for such ships was estimated to be 11.52
g/kW-h.

EPA assumes that 2% of sulfur is emitted as primary sulfate PM from
Category 3 marine diesel engines.  Therefore, we treated 2% of total
sulfur emissions as sulfate emissions and the SO2 emission factor was
adjusted down accordingly to maintain the sulfur mass balance. (Note
that for the same amount of S, the sulfate emission factor is 1.5 the
SO2 emission factor to account for the different molecular weights of
SO2 and sulfate.)

Therefore, within the SECA, the gas-phase SO2 and particulate-phase
sulfate emission factors are 6.27 g/kW-h and 0.19 g/kW-h, respectively. 
Outside of the SECA, the gas-phase SO2 and particulate-phase sulfate
emission factors are 11.29 g/kW-h and 0.35 g/kW-h, respectively.

The sulfate emission rates calculated above are consistent with
available data on the sulfate fraction of particulate matter (PM)
emitted from ship diesel engines.  Fleischer et al. (1998) report that
20 to 30% of PM emissions from ship diesel engines are sulfate (for a 3%
sulfur fuel content).  The EPA (2002) emission factor for PM is 1.3
g/hp-h, i.e., 1.74 g/kW-h.  These values lead to an emission factor for
sulfate in the range of 0.31 to 0.47 g/kW-h for a sulfur fuel content of
27,000 ppm.  The emission factor of 0.35 g/kW-h calculated above falls
within this range.

Based on data from Corbett and Koehler (2003), the power of a typical
ship was estimated to be 16,000 kW (Corbett, 2005).  It should be noted
that there is a wide range of power among various ships, with the
largest container ships having power exceeding 65,000 kW.

The gas-phase SO2 and particulate-phase sulfate emissions per ship are
then calculated to be 100,320 g/h and 3,040 g/h, respectively, within
the SECA and 180,640 g/h and 5,600 g/h, respectively, outside the SECA

A similar approach was used to calculate the NO and NO2 emission rates. 
The NOx emission factor per unit of work is reported to be 12.38 g/hp-h
(as NO2) for a ship at slow or medium speed in transit mode (EPA, 2002).
 This is equivalent to 16.6 g/kW-h.  For a typical ship with a power of
16,000 kW, the resulting NOx emission rate is 266,000 g/h (as NO2). 
Assuming that 5% of the NOx emissions are released as NO2 on a molar
basis, the NO and NO2 emission rates were calculated to be 164,800 g/h
and 13,300 g/h, respectively.  These emission rates were used for ships
within and outside the SECA, i.e., it was assumed that the switch to
lower sulfur content fuel within the SECA did not affect the NOx
emission rates.

	3.2.2	Stack parameters

These parameters include the locations of the sources and their stack
characteristics.  As discussed in Section 2.3.2, the ships were placed
along the coastline for the target value calculations and at various
distances from the coastline for the SECA boundary estimation.  In this
section, we discuss the spatial density of the ships, i.e., the spacing
between each ship.  This was determined based on analysis of ship
activity data, as described below.

We used the average number, N, of ships in transit along the coast per
year and average cruising speed, V (km/h), to calculate the average
distance, D (km), between two ships along a shipping lane.

D = V * (24 h/day * 365 days/yr) / N

The annual number of ships transiting along the southern California
coast was estimated to be 13,000 (ICOADS, 2002).  This number includes
all ships transiting to and from ports located on the southern Pacific
coast as well as ships transiting southward/northward from/to ports
located on the northern Pacific coast.  It is likely to be an
overestimate of the number of ships transiting along the coast because a
fraction of those ships will be transiting along shipping lanes that
extend from the ports westward into the Pacific Ocean.  The cruising
speed varies according to ship type.  It is about 24 knots for container
ships and about 16 knots for tankers.  Here, the average ship cruising
speed was estimated to be about 20 knots, i.e., 36 km/h (ICOADS, 2002). 
Thus, the average distance estimated for the southern Pacific coast was
calculated as follows.

D = 36 * 24 * 365 / 13,000 = 24.3 km

Based on this analysis, we used a distance of 25 km between ships to
calculate ship emissions.  The same distance was used for the other
coastlines.

The other stack parameters required include stack characteristics such
as stack height, stack diameter, stack exhaust velocity, and stack exit
temperature.  These parameters were obtained for typical container and
tanker ship type categories from an ARB report (ARB, 2000).  For this
study, we used the average values for these two categories (see Table
3-1).

Table 3-1.  Stack characteristics.

Stack height	35.3 m

Stack diameter	1.9 m

Exhaust velocity	24.6 m/s

Exhaust temperature	537 K





RESULTS

The initial CALPUFF baseline (i.e., ships along the coastline) and SECA
boundary simulations for the Southern Pacific, Northern Pacific, and
Gulf of Mexico coastlines were conducted using the latest EPA-approved
version of CALPUFF.  However, after discussions with the CALPUFF
developers (Scire, 2006), it was decided that the final simulations
would be conducted using the latest BETA-Test version of the model. 
This version addresses problems reported to the model developers by
CALPUFF users.  The results presented here are all based on simulations
conducted with the beta version of CALPUFF.

Because the objective of the study is to identify upper limits for the
off-shore distances at which sea-going ships may switch from cleaner
fuel to high-sulfur content fuel, the results are presented in terms of
the ratios of the ground-level SO2 and sulfate concentrations at
land-based receptors calculated from the off-shore source simulations to
those calculated from the coastline source simulations.  This allows us
to determine the percentage of receptors at which the emissions from the
sea-going ships will lead to air quality impacts that are less than or
equal to the target values, i.e., the ground-level SO2 and sulfate
concentrations calculated from the baseline simulation.

Before presenting the results, it is useful to discuss the expected
differences between SO2 and sulfate in terms of the evolution of their
downwind concentrations, and how these differences affect the results
obtained here.  SO2 and sulfate concentrations will display different
behaviors downwind of the ships.  SO2 concentrations will decrease
continuously with distance from the source (due to dilution, removal,
and conversion to sulfate), whereas sulfate concentrations will first
decrease (dilution and removal of primary, i.e., directly emitted
sulfate), then increase (formation of secondary sulfate from the
oxidation of SO2) before finally decreasing (dilution and removal
exceeding formation).

This behavior of sulfate introduces an additional complication: the
sulfate target values at receptors near the coastline will be determined
by the directly emitted sulfate, while the target values at larger
distances inland will be determined by some combination of primary and
secondary sulfate, with the secondary sulfate component increasing and
the primary sulfate component decreasing.  Even further inland, both
components will decrease as the rate of dilution and removal exceeds the
formation of sulfate.

These differences between the behavior of SO2 and sulfate suggest that
the SO2 concentrations will become smaller than the design values at a
smaller distance than the sulfate concentrations will.  The SO2
concentrations due to the higher SOx emissions from the ships at sea
burning higher sulfur fuel will be offset by the dilution and conversion
of SO2 much sooner than the sulfate concentrations since the latter will
initially experience an increase from the SO2 conversion.

In the discussion of the results that follows, we will refer to
concentrations calculated from the emissions of coastline sources (i.e.,
ships within the SECA burning low-sulfur fuel) as the “target”
values, and the concentrations due to emissions from ships at sea (i.e.,
ships outside the SECA burning high-sulfur fuel) as the “design”
values.  The ratios of the “design” concentrations to the
“target” concentrations will be referred to as the “design
ratios”.

 

Results for the Southern Pacific U.S. Coastline

Figure 4-1 shows the spatial patterns of the design ratios of the
ground-level annual average SO2 concentrations for ships at 125 km from
the coastline.  While there are large areas where the ratios are less
than one, particularly near the southern part of the domain, the ratios
are larger than one for the majority of the receptors.  This is depicted
in Figure 4-2, which shows the cumulative frequency distribution of the
design ratios. The design ratios are less than one at about 41% of the
receptors.

The spatial distribution of the sulfate design ratios for ships at 125
km from the coastline is shown in Figure 4-3.  In contrast to the SO2
results, the sulfate ratios are less than one over a very small portion
of the domain near the southern boundary.  From Figure 4-4, we see that
the percentage of receptors for which the sulfate design ratios are less
than one is only about 4%.  These differences between the SO2 and
sulfate results are consistent with our expectations as discussed
earlier.



Figure 4-1.	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 125 km from the Southern Pacific U.S.
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel. The red dots represent the
locations of the sea-going ships.

Figure 4-2.	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 125 km from the Southern Pacific U.S.
coastline.

 

Figure 4-3.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Southern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.

Figure 4-4.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 125 km from the Southern Pacific
U.S. coastline.

The SO2 results for ships at 250 km from the coastline are shown in
Figures 4-5 and 4-6.  From Figure 4-5, we see that, except for a small
region in the Central Valley of California (Kings county, most of Fresno
county, and portions of Tulare and Kern counties) and isolated locations
along the coast in Santa Barbara county, most of the receptors have
ratios less than one.  Figure 4-6 shows that the percentage of receptors
that have ratios less than one for ships at 250 km from the coastline is
nearly 91%.

For sulfate, even when the ships are at a distance of 250 km, we see
from Figure 4-7 that the sulfate design ratios are less than one only
near the southern portion of the modeling domain, in Orange and San
Diego counties, southern Imperial county, western Riverside county, and
a small region of southern Los Angeles county.  In the rest of the
domain, the design ratios are larger than one, suggesting that increases
in downwind sulfate concentrations from the conversion of SO2 to sulfate
are still the determining factors for ship emissions at 250 km.  Figure
4-8 shows that the percentage of receptors for which the sulfate design
ratios is less than one for ships at 250 km from the coastline is only
about 25%.

Figures 4-9 and 4-10 show the SO2 results for ships at 375 km from the
coastline. As seen in Figure 4-9, except for one location along the
coastline in Santa Barbara county, all the receptors show design ratios
less than one.  The cumulative frequency distribution, shown in Figure
4-10, confirms that SO2 air quality impacts from the ships burning
high-sulfur fuel are less than those from coastline ships burning
low-sulfur fuel at over 99.99% of the land-based receptors.

Figures 4-11 and 4-12 show that the sulfate results for ships at 375 km
from the coastline still show larger air quality impacts than the
coastline ships for a large majority of the receptors.  From Figure
4-11, a clear north-south gradient is evident.  In the region north of
Los Angeles county, and including portions of northern Los Angeles
county, the sulfate ratios are larger than one.  In the region south,
all the sulfate ratios are less than one.  Figure 4-12 shows that about
42% of the receptors have sulfate ratios less than one.

Figure 4-5.	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 250 km from the Southern Pacific U.S.
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel. The red dots represent the
locations of the sea-going ships.

Figure 4-6.	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 250 km from the Southern Pacific U.S.
coastline.

Figure 4-7.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Southern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.

Figure 4-8.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 250 km from the Southern Pacific
U.S. coastline.

Figure 4-9.	Ratios of annual-average SO2 concentrations due to sea-going
ships burning high-sulfur fuel at 375 km from the Southern Pacific U.S.
coastline to the concentrations (target values) due to dockside ships at
the coastline burning low-sulfur fuel. The red dots represent the
locations of the sea-going ships.

Figure 4-10.	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 375 km from the Southern Pacific U.S.
coastline.

Figure 4-11.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 375 km from the Southern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.

Figure 4-12.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 375 km from the Southern Pacific
U.S. coastline.

The sulfate results for ships at 500 km from the coastline are shown in
Figures 4-13 and 4-14 (the corresponding SO2 results are not shown here
since the 375 km results presented earlier show that a distance of 375
km is more than adequate for setting the upper limit of the SECA for SO2
impacts).  The north-south gradient is still evident, as shown in Figure
4-13, but the boundary between the two regions of ratios less than one
in the south to ratios larger than one in the north has shifted to the
north (to Ventura county in the west and to Kern and Tulare counties in
the east).  The two regions are approximately equal in area, as
confirmed by the cumulative frequency distribution in Figure 4-14.

In these analyses, the Santa Barbara area tends to show higher
concentrations of SO2 and sulfate and, in some cases high
concentration/target value ratios.  One reason for such high
concentrations is that the Santa Barbara area extends westward into the
Pacific Ocean and, as a result, receptors near the coast have more ship
emission sources in their close vicinity than receptors located in other
areas along the coast.  All sources will not impact the Santa Barbara
receptors simultaneously because such impacts will depend on the wind
flow (see Figure 3-6).  Nevertheless, the probability of impact from
ship emission sources should be higher for the Santa Barbara area than
for other areas along the southern Pacific coast because of the design
of the source/receptor locations in this screening study.  This
characteristic of the source/receptor relationship should be kept in
mind when interpreting the simulation results.  Note that the air
quality modeling to be conducted later with the 3-D CMAQ model will
locate ship emissions along shipping lanes and, therefore, will provide
a more realistic set of source/receptor relationships.

Results for the Northern Pacific U.S. Coastline

Figures 4-15 and 4-16 show the SO2 results for ships at 125 km from the
Northern Pacific coastline.  We see from Figure 4-15 that the regions
with ratios less than one are approximately equal in area to the regions
with ratios greater than one.  As shown in Figure 4-16, the ratios are
less than one at about 47% of the receptors.  The higher ratios
typically occur inland in areas of high elevation (e.g., the Cascade
mountain range).  The

Figure 4-13.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 500 km from the Southern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.



Figure 4-14.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 500 km from the Southern Pacific
U.S. coastline.

Figure 4-15.	Ratios of annual-average SO2 concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.

Figure 4-16.	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 125 km from the Northern Pacific U.S.
coastline.

corresponding sulfate results are shown in Figures 4-17 and 4-18.  The
sulfate ratios are larger than one over the entire domain except for a
few isolated locations.  Over a large portion of the domain, the ratios
range from 1.4 to 1.8.  From Figure 4-18, we see that over 99.99% of the
receptors have ratios larger than one.  The results from the San
Francisco Bay Area are similar to those that were obtained for the
southern Pacific domain, which suggests that most of the ships impacting
this area are within the modeling domains.

At 250 km from the coastline, the SO2 ratios are less than one at nearly
100% of the receptors, as shown in Figures 4-19 and 4-20.  However,
sulfate ratios are still larger than one at a majority (nearly 96%) of
the receptors, as shown in Figures 4-21 and 4-22.

The SO2 ratios for ships at 375 km and 500 km from the North Pacific
U.S. coastline are less than one at all the receptors and are not shown
here.  Figure 4-23 shows the spatial distribution of the sulfate ratios
for ships at 375 km from the coastline, while Figure 4-24 shows the
cumulative frequency distribution of the ratios.  From Figure 4-24, we
see that only about 20% of the receptors show ratios less than one. 
However, over a very large part of the domain, the ratios larger than
one are usually in the range of 1 to 1.4, as shown in Figure 4-23.  The
largest ratios, in the range of 1.4 to 1.8, are concentrated in the
western parts of southern Oregon and northern California, near the
boundary between the two states.  This area is in the center of the
domain and is, therefore, exposed to the ship emissions located west and
southwest from its coastline (see wind roses in Figure 3-7).  

The sulfate results for ships at 500 km from the North Pacific U.S.
coastline are shown in Figures 4-25 and 4-26.  At a majority (56%) of
the receptors, the ratios are less than one for ships at this distance. 
The largest ratios are again near the boundary region between California
and Oregon.  The results for the San Francisco Bay Area are
significantly lower than those obtained for the southern Pacific domain
because, at that distance, most of the ships that impact this area are
located southwest of this area (see windroses in Figures 3-6 and 3-7).

Figure 4-17.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.

Figure 4-18.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 125 km from the Northern Pacific
U.S. coastline.

Figure 4-19.	Ratios of annual-average SO2 concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.

Figure 4-20.	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 250 km from the Northern Pacific U.S.
coastline.

Figure 4-21.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.

Figure 4-22.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 250 km from the Northern Pacific
U.S. coastline.

Figure 4-23.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 375 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.

Figure 4-24.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 375 km from the Northern Pacific
U.S. coastline.

Figure 4-25.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 500 km from the Northern
Pacific U.S. coastline to the concentrations (target values) due to
dockside ships at the coastline burning low-sulfur fuel. The red dots
represent the locations of the sea-going ships.

Figure 4-26.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 500 km from the Northern Pacific
U.S. coastline.

4.3	Results for the Gulf of Mexico Coastline

The results for the Gulf of Mexico coastline are quite different from
the two coastlines on the West Coast.  The relative air quality impacts
at land-based receptors from ships at sea are generally lower for the
Gulf of Mexico than for the Pacific Ocean even for ships located at 125
km from the coastline.  The SO2 results for ships at 125 km from the
Gulf of Mexico coastline are shown in Figures 4-27 and 4-28.  From
Figure 4-27, we see that the SO2 ratios are less than one over most of
the receptor network, except in southern Florida.  The ratios are larger
than one at only about 16% of the receptors, as shown in Figure 4-28. 
The larger values simulated in Florida result in part from the design of
the “shipping lane” that is located 125 km south of the coastline
but, in the case of Florida, closer from a coastline located directly
east from the ships.  Therefore, the fraction of receptors that have
ratios below one should be seen as a lower limit.

The sulfate results for ships at 125 km from the Gulf of Mexico
coastline, shown in Figures 4-29 and 4-30, are also different from the
sulfate results for the Pacific Ocean coastlines.  Nearly 40% of the
receptors have sulfate ratios less than one.  Ratios larger than one are
seen in Florida, Georgia, Alabama, and portions of Mississippi and
Louisiana, as well as at the tip of southern Texas near the border with
Mexico.

The SO2 results for ships at 250 km from the Gulf of Mexico coastline
are shown in Figures 4-31 and 4-32.  We see from Figure 4-31 that,
except for a small region in southern Florida, the ratios at all the
receptors are less than one.  The percentage of receptors with ratios
less than one is over 98%, as shown in Figure 4-32.  Figures 4-33 and
4-34 show the 250 km results for sulfate.  We see from Figure 4-33 that
the region with sulfate ratios larger than one is confined to most of
Florida and southern Georgia.  Figure 4-34 shows that only 28% of the
receptors have sulfate ratios larger than one.

We only show the sulfate results for the 375 km and 500 km distances
from the Gulf of Mexico coastline, since all receptors satisfy the
criterion of SO2 ratios less than one at these distances.  Figure 4-35
shows the spatial pattern of sulfate ratios for ships at 375 km from the
coastline.  Ratios larger than one are only seen in Florida and small
areas of southern Georgia.  From Figure 4-36, we see that over 80% of
the receptors show sulfate ratios less than one.

Figure 4-27.	Ratios of annual-average SO2 concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-28.	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 125 km from the Gulf of Mexico coastline.

Figure 4-29.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-30.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 125 km from the Gulf of Mexico
coastline.

 

Figure 4-31.	Ratios of annual-average SO2 concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-32.	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 250 km from the Gulf of Mexico coastline.

Figure 4-33.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-34.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 250 km from the Gulf of Mexico
coastline.

Figure 4-35.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 375 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-36.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 375 km from the Gulf of Mexico
coastline.

The sulfate results for ships at 500 km from the Gulf of Mexico
coastline are qualitatively similar to the 375 km distance results, as
shown in Figures 4-37 and 4-38.  For the 500 km scenario, sulfate ratios
are larger than one only in Florida, as shown in Figure 4-37.  Figure
4-38 shows that the percentage of receptors with sulfate ratios less
than one increases only marginally (by about 3.5%) when the ships are
placed at 500 km instead of 375 km.

4.4	Results for the Atlantic Ocean Coastline

Figure 4-39 shows the spatial distribution of annual-average SO2 ratios
for ships at 125 km from the Atlantic Ocean coastline, while Figure 4-40
shows the cumulative frequency distribution of the ratios.  As in the
case of the Gulf of Mexico coastline, the SO2 ratios are less than for a
majority of the receptors at the 125 km distance.  From Figure 4-39, we
see that the ratios are larger than one only in southern Georgia, most
of North Carolina, and portions of Connecticut and Massachusetts.  The
percentage of receptors with SO2 ratios less than one is nearly 87%, as
shown in Figure 4-40. In contrast, the 125 km sulfate results for the
Atlantic Ocean show that the ratios are larger than one for almost the
entire domain, as shown in Figures 4-41 and 4-42.

The SO2 results for ships at 250 km from the Atlantic Ocean coastline
are shown in Figures 4-43 and 4-44.  At this distance, the SO2 ratios
are less than one throughout the domain.  Figures 4-45 and 4-46 show the
corresponding results for sulfate.  We see from Figure 4-45 that sulfate
ratios are less than one in the southeastern U.S. (Florida, Georgia, and
South Carolina) and some of the New England states, such as Vermont, New
Hampshire and Maine.  The ratios are larger than one in most of North
Carolina, eastern Virginia, eastern Pennsylvania, New Jersey, southern
New York, Connecticut, Rhode Island, and southern Massachusetts.  Figure
4-46 shows that the sulfate ratios are less than one at nearly 58% of
the receptors.

Figure 4-37.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 500 km from the Gulf of
Mexico coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-38.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 500 km from the Gulf of Mexico
coastline.

Figure 4-39.	Ratios of annual-average SO2 concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-40.	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 125 km from the Atlantic Ocean coastline.

 

Figure 4-41.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 125 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-42.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 125 km from the Atlantic Ocean
coastline.

Figure 4-43.	Ratios of annual-average SO2 concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-44.	Cumulative frequency distribution of design ratios of SO2
concentrations from ships at 250 km from the Atlantic Ocean coastline.

Figure 4-45.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 250 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-46.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 250 km from the Atlantic Ocean
coastline.

Figure 4-47 shows the spatial distribution of sulfate ratios for ships
at 375 km from the Atlantic Ocean coastline.  We see that the sulfate
ratios are less than one almost everywhere, except in portions of
eastern North Carolina and southern Massachusetts.  As shown in Figure
4-48, the sulfate ratios are less than one at over 92% of the receptors.

For ships at 500 km from the Atlantic Ocean coastline, the sulfate
ratios are less than one everywhere as shown in Figures 4-49 and 4-50.

Figure 4-47.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 375 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-48.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 375 km from the Atlantic Ocean
coastline.

Figure 4-49.	Ratios of annual-average sulfate concentrations due to
sea-going ships burning high-sulfur fuel at 500 km from the Atlantic
Ocean coastline to the concentrations (target values) due to dockside
ships at the coastline burning low-sulfur fuel. The red dots represent
the locations of the sea-going ships.

Figure 4-50.	Cumulative frequency distribution of design ratios of
sulfate concentrations from ships at 500 km from the Atlantic Ocean
coastline.



SUMMARY AND CONCLUSIONS

A screening study with the CALPUFF dispersion model was conducted to
determine the air quality impacts (annual average ground-level
concentrations of SO2 and sulfate) at an array of land-based receptors
due to SOx emissions from ships burning high-sulfur fuel at sea at
various distances from the coastline.  CALPUFF tends to overestimate the
conversion of SO2 to sulfate in the gas phase (Karamchandani et al.,
2006) and the results presented here are likely to provide conservative
estimates of the impacts of emissions from ships at sea on inland air
quality.  (Because of the simplified treatment of aqueous-phase
chemistry in CALPUFF, this assessment may be altered if the interactions
of the ship plumes with fog dominate sulfate formation.)  The results
were compared with those calculated for ships burning low-sulfur fuel at
the coastline to determine upper bounds for Sulfur Emission Control
Areas (SECAs), i.e., off-shore distances at which the switch to
high-sulfur fuel would not impair air quality.  For each offshore
distance investigated, the percentage of receptors for which the air
quality impacts of ships at sea were lower than the impacts of ships at
the coastline was calculated.

The U.S. coastlines considered in this study include the Pacific Ocean
coastline, the Gulf of Mexico coastline, and the Atlantic Ocean
coastline.  The northern and southern parts of the Pacific Ocean
coastline were studied separately.  The results are summarized in Tables
5-1 and 5-2 for concentration ratios of SO2 and sulfate, i.e., the ratio
of the concentration calculated for ships at sea to the concentration
calculated for ships at the coastline.

The results for SO2 were different from those for sulfate, primarily due
to differences in the behavior of these two species downwind of a
source.  For all the coastlines studied, the majority of the SO2
concentration ratios were less than one at shorter off-shore distances
than for sulfate.  Thus, sulfate concentration ratios were the limiting
factor for defining the upper bounds of the SECA for each coastline. 

Table 5-1. 	Percentage of SO2 concentrations below the design value as
a function of the distance from the coastline.

Distance from coastline	125 km	250 km	375 km	500 km

Southern Pacific	40.7%	90.7%	100%	100%

Northern Pacific	46.6%	97.9%	100%	100%

Gulf of Mexicoa	84.4%	98.1%	100%	100%

Atlantic	86.6%	100%	100%	100%

aNote that Florida values correspond to a shorter ship-coastline
distance and the values presented in the table should be seen as lower
limits.

Table 5-2.	Percentage of sulfate concentrations below the design value
as a function of the distance from the coastline.

Distance from coastline	125 km	250 km	375 km	500 km

Southern Pacific	4.4%	24.9%	41.9%	48.7%

Northern Pacific	0.01%	3.6%	20.3%	55.7%

Gulf of Mexicoa	40.4%	72.0%	80.5%	84.0%

Atlantic	1.2%	57.9%	92.5%	100%

aNote that Florida values correspond to a shorter ship-coastline
distance and the values presented in the table should be seen as lower
limits.

The results showed some differences in results among the various
coastlines studied. These differences are due to differences in the wind
fields bringing the offshore ship emissions and their secondary products
to land as well as differences in precipitation, which removes
pollutants from the atmosphere.

The results from the two Pacific Ocean coastline simulations were
qualitatively similar.  For both Pacific Ocean coastlines, over 90% of
the receptors showed SO2 concentration ratios less than one for ships at
250 km from the coastline.  For sulfate, only about 49% and 56% of the
receptors had concentrations less than one for ships at 500 km from the
southern Pacific Ocean and northern Pacific Ocean coastlines,
respectively.

For the other two coastlines (Atlantic Ocean and Gulf of Mexico), the
SO2 results were qualitatively similar to those for the Pacific Ocean
coastlines, i.e., over 90% of the receptors showed SO2 concentration
ratios less than one for ships at 250 km from the coastline.  However,
there were some large differences for sulfate.  For the Gulf of Mexico
coastline, over 70% of the receptors showed sulfate concentration ratios
less than one for ships at 250 km from the coastline.  For the Atlantic
Ocean coastline, nearly 60% of the receptors showed sulfate
concentration ratios less than one for ships at 250 km from the
coastline.

These results suggest that an off-shore distance of 500 km should be
sufficient when conducting refined modeling of the potential impacts of
ship emissions on air quality inland, if a criterion of about 50% of
inland receptors having sulfate concentrations below the design value is
acceptable to define the SECA.



REFERENCES

ARB, 2000.  Air Quality Impacts from NOx Emissions of Two Potential
Marine Vessel Control Strategies in the South Coast Air Basin,
California Air Resources Board, Sacramento, CA.

Corbett, J.J. and H.W. Koehler, 2003.  Updated emissions from ocean
shipping, J. Geophys. Res., 108, doi:10.1029/2003JD003751.

Corbett, J.J., 2005.  Private communication to Christian Seigneur, AER,
July 2005.

EPA, 2002.  Commercial Marine Emission Inventory, Final Report from
PECHAN, prepared by ENVIRON International Corporation, U.S.
Environmental Protection Agency, Office of Transportation and Air
Quality, Ann Arbor, MI.

Fleischer, F., E.J. Ulrich, R. Krapp and W. Grundmann, 1998.  Comments
on particulate emissions from diesel engines when burning heavy fuels,
Proc. Of the 22nd CIMAC Internat. Congress on Combustion Engines, Vol.
6, Copenhagen, Denmark, May 18-21.

ICOADS, 2002.  International Comprehensive Ocean Atmospheric Data Set,
as transmitted from ERG by Office of Transportation and Air Quality,
U.S. Environmental Protection Agency, Washington, D.C.

Karamchandani, P., A. Koo, and C. Seigneur, 1998. A reduced gas-phase
kinetic mechanism for atmospheric plume chemistry, Environ. Sci.
Technol., 32, 1709–1720.

Karamchandani, P., and C. Seigneur, 1999. Simulation of sulfate and
nitrate chemistry in power plant plumes, J. Air Waste Manage. Assoc.,
49, PM-175–181.

Karamchandani, P., S.-Y. Chen, N. Kumar and M. Gupta, 2006. A
comparative evaluation of two reactive puff models using power plant
plumes measurements, AWMA Guideline on Air Quality Models Conference,
Denver CO, 26-28 April.

Morris, R.E., R.C. Kessler, S.G. Douglas, K.R. Styles and G.E. Moore,
1988.  Rocky Mountain Acid Deposition Model Assessment: Acid Rain
Mountain Mesoscale Model (ARM3), report prepared for the U.S. EPA,
Research Triangle Park, NC.

Scire, J.S., D.G. Strimaitis and R.J. Yamartino, 2000a. A User’s Guide
for the CALPUFF Dispersion Model (Version 5), Earth Tech, Inc. Report,
Concord, MA, January 2000.

Scire, J.S., F.R. Robe, M.E. Fernau and R.J. Yamartino, 2000b. A
User’s Guide for the CALMET Dispersion Model (Version 5), Earth Tech,
Inc. Report, Concord, MA, January 2000.

Scire, J.S., 2005.  Communication via e-mail of Christian Seigneur, AER,
with Joe Scire, EarthTech, 18-19 March 2005.

Scire, J.S., D.G. Strimaitis and F.R. Robe, 2005.  Evaluation of
enhancements to the CALPUFF model for offshore and coastal applications,
Proceedings of the 10th International Conference on Harmonisation with
Atmospheric Dispersion Modelling for Regulatory Purposes, Crete, Greece,
17-20 October 2005.

Scire, J.S., 2006.  Private communication to Prakash Karamchandani, AER,
February 2006.

Seigneur, C., K. Lohman and P. Karamchandani, 2005a.  Review of
Technical Information relevant to Sulfur Oxides (SOx) Emissions
Transport for Ships at Sea, Final Report to Office of Transportation and
Air Quality, U.S. Environmental Protection Agency, Washington, D.C.

Seigneur, C., P. Karamchandani and K. Lohman, 2005b.  Analysis Plan -
Modeling Sulfur Oxides (SOx) Emissions Transport for Ships at Sea, Final
Report to Office of Transportation and Air Quality, U.S. Environmental
Protection Agency, Washington, D.C.



APPENDIX A

REVIEW OF TECHNICAL INFORMATION RELEVANT TO

SULFUR OXIDES (SOx) EMISSIONS TRANSPORT

FOR SHIPS AT SEA

Prepared by

Christian Seigneur

Kristen Lohman

Prakash Karamchandani

Atmospheric & Environmental Research, Inc.

2682 Bishop Drive, Suite 120

San Ramon, CA 94583

Prepared for

U.S. Environmental Protection Agency

Office of Transportation and Air Quality

1200 Pennsylvania Avenue, NW

Washington, DC 20460

Document CP212-05-01b

June 2005

INTRODUCTION

Marine shipping represents a major and growing source of uncontrolled
air pollution in coastal regions and inland areas downwind of coastal
regions in many parts of the world, particularly North America and
Europe.  This can be attributed to both growth in global trade and port
activity, as well as controls on land-based emissions.  In 1973, an
international conference of the International Maritime Organization
(IMO) adopted the International Convention for the Prevention of Marine
Pollution from Ships (MARPOL) designed to prevent pollution from ships. 
In 1997, the IMO agreed to MARPOL Annex VI, a global treaty to reduce
air emissions from ships.  This treaty went into effect on May 19, 2005.
 The treaty sets limits on emissions of sulfur oxides (SOx) and nitrogen
oxides (NOx) and prohibits the international emissions of
ozone-depleting substances, such as chlorofluorocarbons.  One key
element of Annex VI is the establishment of “SOx Emission Control
Areas” (SECAs) near coastal regions where controls on SOx emissions
from ships are more stringent (1.5% fuel content or 15,000 ppm) than in
the open seas (4.5% or 45,000 ppm).

Countries wanting to obtain SECA designation for their coastal areas
must submit a formal application to the IMO.  The U.S. Environmental
Protection Agency (EPA) is currently in the process of exploring the
feasibility of a SECA for U.S. coastal areas and plans to work with
affected states to obtain the necessary data.  As part of the
application process, emissions inventories will be developed and air
quality modeling analyses will be conducted.

EPA will conduct the air quality modeling analyses for the SECA
application in-house using three-dimensional (3-D) grid-based models
such as CMAQ and/or CAMx.  One of the issues of interest for this
modeling exercise is the determination of the “scales of interest”,
i.e., a delineation of the extent and scope of the SECA for each coastal
region that will be considered in the analysis.  This determination will
be performed using a “screening-level” modeling analysis, in which a
methodology will be developed and applied to estimate SOx emissions
transport from ships at sea to areas off the U.S. coasts (Pacific,
Atlantic, and the Gulf of Mexico) using realistic ship emissions and
meteorology. SOx emissions transport from ships on the Great Lakes will
be addressed separately under the U.S.-Canada binational program.

This document describes the first component of the SOx emissions
transport methodology, which is a literature review of available tools
and data to quantify the transport and residence times of SOx over
water.  The fate and transport of pollutants over water has long been of
interest because of the potential impacts of off-shore platforms on air
quality over land and the potential impacts of ship emissions on global
climate change and air quality over land.  Consequently, there is a
significant body of information available on the atmospheric transport,
dispersion and chemistry of pollutants emitted over water.

The Minerals Management Service (MMS) has conducted several studies to
investigate the meteorology and the fate and transport of oil platform
emissions in the Gulf of Mexico (e.g., Yocke et al., 1998).  Those
studies are not directly applicable to the present study because the
emission source is different; nevertheless, some valuable data and
useful experience were obtained in the MMS studies that are relevant to
the present study.  The relevant aspects are discussed in this report. 
The U.S. Navy investigated the potential of ship emissions reaching
shore and that report provides useful information regarding the
different meteorological regimes along the U.S. coastline (Eddington and
Rosenthal, 2003).  The California Air Resources Board (2000) also
conducted a modeling study of the transport and dispersion of NOx
emissions from ships in the southern California region.  There have also
been several academic investigations on the fate and transport of
pollutants emitted from ships.  The most recent and relevant one (Song
et al., 2003) pertains to the simulation of sulfur chemistry in a ship
plume released in the marine boundary layer.  The authors used a simple
box model to simulate the plume.  They concluded that, in the presence
of non-precipitating clouds, non-sea salt sulfate could attain about 2
µg/m3 after a few hours of plume travel time.  The SO2/sulfate
chemistry was found to be linear (i.e., a change in SO2 emissions would
lead to a proportional change in sulfate concentrations) except near the
ship where SO2 concentrations exceeded the hydrogen peroxide (H2O2)
concentrations.

In this report, we first examine the fate and transport models available
for simulating the transport of pollutants over water.  Then, available
data sources are discussed for some of the most important input data
beginning with meteorology, then emission factors, and finally ship
activity data.

AIR QUALITY MODELS

Air quality models can be grouped in two major categories: grid-based
Eulerian models and Lagrangian plume (or puff) models.  Eulerian models
are well suited to address air quality for urban and regional pollutants
that are emitted from a large variety of sources.  However, their
spatial resolution is limited by the grid size and they are not well
suited for addressing air quality impacts associated with individual
sources or groups of sources.  Plume or puff models are better suited
for such air quality impacts since their formulation takes into account
the dispersion of the emitted material from the source to the downwind
distances of interest.

For this screening study of the potential impacts of SOx emissions from
ships at sea, we are considering Lagrangian plume and puff models since
they are the most suitable.  We are considering three models: OCD,
CALPUFF and SCICHEM.  We briefly describe these models below and discuss
their advantages and shortcomings before making our recommendations for
the air quality model to be used for this study.

OCD

OCD was developed under funding from the Minerals Management Service
(MMS) to simulate plume dispersion and transport from offshore sources
to receptor areas on land or water.

OCD is a steady-state Gaussian model that uses hourly inputs.  The
steady-state assumption implies that the wind direction and speed are
constant for an air parcel after it leaves the source regardless of the
time needed for an air parcel to travel between the source and the
receptor point. Its formulation includes enhancements that take into
account differences between overwater and overland dispersion
characteristics, the sea-land interface and off-shore platform
aerodynamic effects.

OCD requires both overwater and overland meteorological data (i.e.,
including wind speed and direction, water surface temperature, overwater
air temperature, mixing height and relative humidity).  Missing
overwater meteorological data such as turbulence intensities are
parameterized using bulk aerodynamic wind and temperature profile
relationships.

The effect of the source on plume dispersion (stack-tip downwash and
building downwash) can be taken into account.  Corrections are made for
the presence of complex terrain.  The evolution of the thermal boundary
layer near the coast is simulated.  Transitional plume rise and the
partial penetration of elevated temperature inversions are simulated.

OCD can simulate the chemical decay of pollutants using first-order
transformation rates that are user-specified.  However, the formation of
secondary pollutants from an emitted primary pollutant (e.g., formation
of sulfate from emitted SO2) cannot be simulated; this is a major
deficiency for this study since it addresses the possible impacts of
sulfate concentrations on air quality.  Removal processes (e.g., dry
deposition) are simulated using a first-order decay.

OCD is listed by EPA as a guideline model but only for primary
pollutants (Federal Register, 2003).  Therefore, it is not recommended
by EPA for secondary air pollutants such as sulfate formed from SO2
oxidation in the atmosphere.

CALPUFF

CALPUFF was originally developed under funding from the California Air
Resources Board (ARB) along with its associated meteorological model,
CALMET (Scire et al., 2000a, 2000b).

CALPUFF is a non-steady-state puff dispersion model that can simulate
the effects of time- and space-varying meteorological conditions on
pollutant transport, transformation, and removal.  It can accommodate
arbitrarily varying point, area, volume, and line source emissions.

The recommended meteorological inputs for applying CALPUFF are the
time-dependent outputs of CALMET, a meteorological model that contains a
diagnostic wind field module and overwater and overland boundary layer
modules.  Optionally, CALMET can use the outputs of prognostic
meteorological models, such as MM5 and CSUMM, to create the
meteorological fields required by CALPUFF.

CALPUFF includes algorithms for near-source effects such as building
downwash, transitional plume rise, partial plume penetration, sub-grid
scale terrain interactions as well as longer range effects such as
pollutant removal due to wet and dry deposition, simplified chemical
transformations, vertical wind shear, overwater transport and coastal
interaction effects.  

CALPUFF offers several options to simulate the formation of secondary
sulfate and nitrate particles from the oxidation of the emitted primary
gaseous pollutants, SO2 and NOx respectively.  The oxidation of SO2 to
sulfate is of interest for this study.  The more advanced chemistry
module available in CALPUFF uses the RIVAD/ARM3 chemical mechanism
(Morris et al., 1988).  This simple mechanism treats the conversion of
NO to NO2 accompanied by its further transformation to total nitrate and
conversion of SO2 to sulfate.  It is assumed that background
concentrations of reactive hydrocarbons (VOC) are low and, therefore,
this mechanism is not considered suitable for urban regions.  It may be
suitable for oversea situations where VOC concentrations are not too
high.  However, in areas such as southern California, where urban
coastal pollution may be transported over the ocean via the land-sea
breeze, the assumption of low background VOC concentrations may
sometimes be invalid.

In the RIVAD/ARM3 chemical mechanism, the NO-NO2-O3 chemical system is
first solved to generate pseudo-steady-state concentrations of NO, NO2,
and O3.  During the day, this system consists of the NO2
photodissociation to yield NO and O3 and the NO-O3 titration reaction to
yield NO2.  During the night, only the NO-O3 titration reaction is
considered.  The steady-state daytime concentration of the hydroxyl
radical (OH) is calculated from the O3 concentration after the solution
of the NO-NO2-O3 system.  Gupta et al. (2001) have noted that the O3
concentrations are incorrectly treated in CALPUFF, resulting in the
overestimation of OH concentrations, and thus overestimations in the
gas-phase oxidation rates of SO2 to sulfate and NOx to nitrate.  The
RIVAD/ARM3 mechanism does not explicitly calculate the aqueous-phase
oxidation of SO2 to sulfate.  Instead, a constant heterogeneous SO2
oxidation rate (0.2% per hour) is added to the gas phase conversion
rate.  The partitioning of semi-volatile chemical species (ammonium
nitrate) between the gas phase and the particulate phase is simulated
with a simple thermodynamic model.

CALMET is the companion meteorological model that is used with CALPUFF. 
A weakness of CALMET has recently been identified (Wheeler, 2005). 
CALMET does not correctly handle cases of unstable convective
atmospheric conditions over water (when water temperature is warm and
air temperature is cold, for example) because it assumes near-neutral
conditions over water.  Consequently, the mixing height is calculated
based on a neutral mixing relationship and, under conditions of light
wind speeds when the mechanical mixing heights are small, CALMET
underpredicts the actual mixing height.  This weakness can be an issue
in areas such as the Gulf of Mexico where warm water temperatures are
possible.  EarthTech, the developer of CALMET is addressing this problem
by adding a convective mixing height calculation in CALMET.  This new
version of the model is currently being tested but it is not yet
publicly available.  Based on our discussion with EarthTech (Scire,
2005), we will circumvent this potential problem by inputting measured
or modeled mixing heights directly into CALMET (meteorological data are
discussed in Section 3).

CALPUFF is listed by EPA as a preferred air quality model for assessing
the long-range transport of air pollutants and on a case-by-case basis
for certain near-field applications involving complex meteorological
conditions (Federal Register, 2003).  CALPUFF is also recommended by the
Federal Land Managers’ Air Quality Values Workgroup (FLAG) for
assessing the effects of distant plumes on atmospheric visibility.

SCICHEM

SCICHEM is an extension of the Second-order Closure Integrated PUFF
model (SCIPUFF) that includes atmospheric chemical transformations.  It
has been developed under funding from EPRI and the Defense Threat
Reduction Agency (DTRA) ((Sykes et al., 1988, 1993; Sykes and Henn,
1995; Karamchandani et al., 2000; EPRI, 2000). 

SCICHEM is a non-steady-state multi-species model that incorporates a
comprehensive treatment for gas- and aqueous-phase chemistry, and PM
formation.

SCIPUFF represents a plume with a multitude of three-dimensional puffs
that are advected and dispersed by the local micrometeorological
conditions.  Each puff has a Gaussian representation of the
concentrations of individual species.  SCIPUFF simulates the plume
transport and dispersion using a second-order closure approach to solve
the turbulent diffusion equations, which provide a direct connection
between measurable velocity statistics and predicted dispersion rates.  

SCIPUFF can assimilate observational data ranging from a single wind
measurement to multiple profiles.  Alternatively, three-dimensional
gridded wind and temperature fields generated by a prognostic model or
other analyses can be used as input to the model.  SCIPUFF can simulate
the effect of wind shear since individual puffs evolve according to
their respective locations in an inhomogeneous velocity field.  As puffs
grow larger, they may encompass a volume that cannot be considered
homogenous in terms of the meteorological variables.  A puff splitting
algorithm accounts for such conditions by splitting puffs that have
become too large into a number of smaller puffs.  Conversely, individual
puffs that are affected by the same (or very similar) micrometeorology
may also merge to produce a larger single puff.  Also, the effects of
buoyancy on plume rise and initial dispersion are simulated by solving
the conservation equations for mass, heat, and momentum.  

For PM related regulatory applications, it is important that the
underlying model should account for processes responsible for
phase-dependent chemical transformations and PM characterization.  We
provide a brief description of chemical and PM components of SCICHEM.  

In SCICHEM, the gas-phase chemical reactions within the puffs are
simulated using a general framework that allows any chemical kinetic
mechanism (e.g., CBM-IV, SAPRC) to be treated.  Therefore, SCICHEM can
simulate atmospheric conditions ranging from the clean atmosphere to
polluted areas. To minimize the need for computational resources needed
to treat the typical chemical mechanisms, the gas-phase puff chemistry
can optionally be simulated using a three-staged chemical kinetic
mechanism where the number of reactions treated increases as the puff
mixes with background air (Karamchandani et al., 1998).  This multistage
approach offers reasonable accuracy (within ±10%) with increased
computational speed.

Chemical species concentrations in the puffs are treated as
perturbations from the background concentrations. This approach allows
the treatment of overlapping puffs and, therefore, provides great
flexibility for simulating processes such as calm conditions, wind shear
and overlapping plumes for different sources.  Optionally, SCICHEM can
explicitly simulate the effect of turbulence on chemical kinetics.

SCICHEM includes aqueous-phase chemistry.  It is simulated using the
RADM chemical mechanism.  When the aqueous-phase chemistry option is
selected, the wet deposition of pollutants is computed from the cloud
water concentrations of pollutants and the precipitation rate. 
Otherwise, scavenging coefficients are used to calculate wet deposition.
 The partitioning of semi-volatile chemical species between the gas
phase and the particulate phase is simulated with the thermodynamic
model ISORROPIA.

EPA has added SCIPUFF to the list of alternate models (Appendix B of the
EPA Guideline on Air Quality Models, Federal Register, 2003) for the
simulation of the long-range transport and dispersion of air pollutants.

Recommendations

Table 2-1 presents a summary of the advantages and shortcomings of the
three models reviewed here.  The major shortcomings of OCD are its use
of the steady-state assumption and its lack of treatment of chemical
transformations.  The steady-state assumption implies that the wind
direction and wind speed are assumed to be constant for a puff released
from the source, whereas the other two models allow for changes in wind
direction and wind speed.  Chemical transformations in OCD are limited
to a simple decay of the emitted pollutants and do not allow the
treatment of secondary pollutant formation (such as the formation of
sulfate from SO2).  The major shortcoming of CALPUFF is its simplified
chemistry that tends to overestimate sulfate formation in the gas phase
and uses a simple parameterization for the cloud/fog aqueous phase. 
CALPUFF offers the major advantage of being widely used and being an EPA
preferred guideline model for the long-range transport of SOx.  SCICHEM
offers a more comprehensive formulation than CALPUFF.  However, SCICHEM
is not yet an EPA preferred guideline model.  It is still considered a
research-grade model and its computational requirements are
significantly greater than those of the other two models.

On the basis of this review, we recommend that CALPUFF be used to
simulate the transport, transformation and deposition of SOx emissions
from ships, with the caveat that one must bear its limitations in mind.

	

Table 2-1.	Comparison of the advantages and shortcomings of three
plume/puff models for emissions from off-shore sources.

Characteristics	OCD	CALPUFF	SCICHEM

Steady-state vs. transient	Steady-state	Transient	Transient

Spatial resolution	Gaussian plume	Puffs	Puffs

Wind-shear	No	Yes	Yes

Plume overlaps	Yes	Yes	Yes

Near-source effects	Yes	Yes	Yes

Chemical transformations	First-order decay	Simplified chemistry
Comprehensive chemistry

Dry deposition	First-order decay	Yes	Yes

Wet deposition	No	Yes	Yes

Source types	Point, line and area	Point, line, area and volume	Point

Regulatory status	EPA preferred guideline model for primary pollutants
released over water	EPA preferred guideline model for long-range
transport and visibility impacts of air pollutants	EPA alternate
guideline model

Computational requirements	Low	Moderate	High



METEOROLOGICAL DATA

Meteorological data are necessary to run an air dispersion model.  For
the CALPUFF model, the meteorological input data must first be formatted
by the CALMET pre-processor.  CALPUFF requires standard surface and
upper air meteorological data.  CALMET also has an overwater option that
allows the use of special overwater measurements for grid cells that are
over the ocean.  The data required for the overwater option are: air-sea
temperature difference, air temperature, relative humidity, wind speed
and wind direction.  Two optional measurements, overwater mixing height
and overwater temperature gradients, may be supplied if available.  If
the optional parameters are not supplied, CALMET uses default values.

Land-based Measurements

Land-based meteorological measurements are required for both surface and
upper air observations above land portions of the domain.  The data
required are standard format data from the National Climatic Data Center
(NCDC) (Scire et al., 2000).  The upper air data required are standard
NCDC format TD6201 radiosonde data including pressure, elevation,
temperature, wind direction, and wind speed for each sounding level. 
The surface observations that are needed are provided in the NCDC
Integrated Surface Hourly Observations.  These include wind speed, wind
direction, temperature, and dew point temperature.  

Fixed Overwater Measurements

The required parameters are all available for the Pacific and Atlantic
oceans near the U.S. coastline and for the Gulf of Mexico and Great
Lakes from the National Data Buoy Center (NDBC) (NDBC, 2005).  The
measurements are taken from buoys.  The buoys are at varying distances
from the coast.  Those near the coast are frequently near harbors or
bays.  Though the coverage is not uniform, the full length of the
continental U.S. coastline is covered by those data.  Most of the buoys
are owned and operated by NDBC but there are also several other agencies
that submit their data to the NDBC database.  Figures 3-1 through 3-6
show the locations of the NDBC buoys as well as those that are run by
other agencies and are included in the NDBC database.

The Minerals Management Service (MMS) has performed modeling for the
Breton National Wilderness Area which is on the Coast of the Gulf of
Mexico.  The MMS has provided us additional overwater data for the Gulf
of Mexico including both surface and upper air data.  The availability
of upper air data will allow a more thorough modeling of the unique
conditions above the Gulf of Mexico.  These data are available for the
years 1999-2001.



Figure 3-1.	NDBC buoys along the Washington, Oregon, and northern
California coastline.

Figure 3-2.	NDBC buoys along the southern California coastline

Figure 3-3.	NDBC buoys along the western Gulf of Mexico

Figure 3-4.	NDBC buoys along the eastern Gulf of Mexico

Figure 3-5.	NDBC buoys along the southeastern U.S. coastline

Figure 3-6.	NDBC buoys along the northeastern U.S. coastline.

Data from Ships

The CALMET preprocessor also allows for the location of meteorological
observations to vary so that measurements made from ships can be used
for modeling as well.  NCDC as well as other agencies have shipboard
measurements.  

One available database is the International Comprehensive
Ocean-Atmosphere Data Set (ICOADS) from NOAA (NOAA, 2005).  It provides
data on the location of the ship, as well as air temperature, sea
temperature, wind speed and direction, pressure, and dew point.  These
data are available from 1950 through 2002.

Model Outputs

The outputs of meteorological models can be used, particularly in cases
where there are insufficient meteorological observations.  This may be
the case for upper air data in the Atlantic and Pacific Oceans. 
Examples of model outputs that could be used as surrogates for upper air
data include those from the 2001 or 2002 MM5 simulations sponsored by
EPA, those from the NCEP/NCAR reanalysis project and those from the
Advanced Climate Modeling and Environmental Simulations (ACMES)
database.

Recommendations

There appears to be sufficient meteorological data to model the
transport of SOx emissions from ships.  The availability of upper air
data for the Gulf of Mexico will be particularly valuable.  In the
absence of upper air data over water for the other areas, either some
default assumptions will need to be made regarding atmospheric stability
or the outputs of archived meteorological simulations will be used.  We
will discuss our proposed technical approach in the Analysis Plan that
will be prepared in Task 2.

4.	EMISSION FACTORS

Emission factors are needed to estimate the emissions of SOx associated
with various ship activities.  We reviewed available emission factors
and provide our recommendations below.

EPA (2000) Emission Factors

Emission factors for air pollutant emissions from ships are provided in
the report titled “Analysis of Commercial Marine Vessels Emissions and
Fuel Consumption Data” (EPA, 2000).  Emission factors are provided for
several air pollutants including SO2 and PM.  Those emission factors are
provided for different oceangoing ship categories that include bulk
carriers and tankers, general cargo ships, container/RoRo/auto
carriers/refrigerated ships, and passenger ships.  The emission factors
are a function of the operating mode of the engine; four modes were
considered: normal cruise, slow cruise, maneuvering and docking
(hoteling).

For SO2, the emission factor is a function of the fuel consumption rate
and sulfur content of the fuel.  The fuel consumption rate is provided
per unit of work (i.e., g/kW-h) as a function of the fractional load. 
The engine work (kW-h) is a function of the ship type (see above) and
ship deadweight.  The fractional load is the ratio of the actual engine
output and rated engine output; it is a function of the engine mode and
ship type.

If one assumes a fuel sulfur content of 3%, the SO2 emission factor per
unit of work is 16 g/kW-h for a cruising ship and 20 to 25 g/kW-h for a
maneuvering ship.  The SO2 emission factor per unit of fuel is 71
kg/metric ton.

EC Emission Factors

A recent report from the European Commission (EC) provides emission
factors for air pollutants from ships (EC, 2002).  Emission factors are
reported for pollutants including SO2 and PM.  The emission factors are
provided either by engine type and fuel type (15 combinations) or by
ship type (16 oceangoing ship types); different factors are provided for
at sea and in port activities (emission factors for PM are only provided
for in port activities).

The emission factors are reported in g/kW-h.  Thus, the engine
horsepower must be estimated as a function of the ship type and
activity.

The SO2 emission factor per unit of work is in the range of 10 to 13
g/kW-h for a ship at sea and 11 to 13 g/kW-h for a ship at port.  The
SO2 emission factor per unit of fuel is in the range of 46 to 54
kg/metric ton.  These emission factors are slightly lower (by 20 to 35%)
than those reported in the EPA report cited above.  

EPA (2002) Emission Factors

A recent EPA report (2002) presents a review of emission factors
available from several sources.  The emission factors reviewed were for
ships with engines with displacement exceeding 30 liters (so-called
Category 3 engines).

Emission factors are reported for three different engine types (slow
speed, medium speed and steam boiler) for transit modes and hoteling
modes.

For slow and medium speed engines, the SO2 emission factor per unit of
work is about 13 g/kW-h for a ship in transit mode, and 1.4 g/kW-h for a
ship in hoteling mode.  For steam boilers, the SO2 emission factor per
unit of work is 20 g/kW-h for a ship in both transit and hoteling modes.
The SO2 emission factor per unit of fuel is assumed to be 60 kg/metric
ton in transit and 7 kg/metric ton when hoteling (steam boilers were
assumed to use the same fuel while hoteling as in transit, i.e., 30
kg/metric ton).  These emission factors appear to be consistent with
those from the EC and lower than those from the EPA 2000 report.

Recommendations

This brief review of available emission factors for SO2 emissions from
ships show that there is some general consistency among the different
sources of information.  The differences among the various references
are well within the uncertainty ranges that one would expect for
emission factors of air pollutants.  We propose to use the most recent
EPA emission factors (EPA, 2002) for this study because they represent
the most recent source of information.  These emission factors combined
with ship type and ship activity data will provide emission rates of
SO2.

It should be noted that there are no emission factors for sulfate.  SO2
emission factors are estimated as a function of the sulfur content of
the fuel and the implicit assumption is that all sulfur is emitted as
gaseous SO2.  There is evidence that particulate sulfate emissions are
associated with diesel engines.  For example, sulfate may account for up
to 12% of PM emissions from cars and trucks (Shi et al., 2000).  PM
emission factors are available for ship emissions.  By using the EPA
(2002) PM emission factor for diesel engines of a ship in transit and
assuming that PM is 12% sulfate, we obtain an emission factor of 0.2
g/kW-h.  This value corresponds to 1.6% of the SO2 emission factor. 
Data from the Navy Pilot Emission Control Program (NPECP) on PM
emissions from marine diesel engines confirm these results, although the
sulfate fraction of PM depends on the fuel type and the engine regime,
ranging from 2 to 19% of PM.  Furthermore, EPA assumes that 2% of sulfur
is emitted as primary sulfate PM from Category 3 marine diesel engines
(i.e., those engines with displacement > 30 liters per cylinder). 
Because the rate of oxidation of SO2 to sulfate is slow in the absence
of fog or clouds (on the order of 1% per hour), emissions of sulfate
from ships may contribute significantly under such conditions to the
sulfate concentrations over land that are due to ship emissions. 
Therefore, we will treat 2% of total sulfur emissions as sulfate
emissions and the SO2 emission factor will be adjusted down accordingly
to maintain the sulfur mass balance.

5.	SHIP ACTIVITY DATA

Ship activity data must be determined so that the emission factors can
be applied to provide air pollutant emissions from ships.  The activity
data are typically calculated based on four types of information: port
locations, vessel descriptions, trip records, and shipping lane
definitions.  Port locations are available from the United States Army
Corps of Engineers (USACE) (USACE, 2005).  For efficiency, these data
should be aggregated so that all nearby ports are treated as one.   

Once port data have been aggregated, trip data need to be added.  Trip
data are necessary to track how many of each type of ship move between
each port.  The emission factors will be applied according to ship type,
therefore, it is important to characterize the ship types per shipping
lane per year.  The USACE provides data on entrances and clearances
(USACE, 2005) for vessels traveling under foreign flags.  This database
lists each entry and departure of a vessel bearing a foreign flag. 
Through these databases, a ship can be traced through its travels
through U.S. ports.  Information on domestic ship traffic is also
compiled by the USACE.

The entrances and clearances databases list a ship code that can then be
matched up to another USACE database.  This database provides data on
each foreign ship that has registered at a U.S. port providing
information on type, size, and power.

Once all of these data have been gathered and processed, they can be
combined to provide a list of potential trips (e.g., Portland to San
Francisco) by type of ship.  The final step is to provide a geographic
location for the ship emissions.  Since CALPUFF allows the modeling of
line sources, we need to determine the geographic definitions of the
shipping lanes that will be input into the model.  These data are
available from the USACE in the form of the Waterways Network (USACE,
2005).  It provides information on the latitude and longitude of each
node in the U.S. waterways.

Alternatively, the ICOADS database that provides meteorological
measurements from ships can be used to determine shipping lanes (NOAA,
2005).  ICOADS provides time- and space-resolved meteorological data. 
Because each record provides both the ship code and a latitude and
longitude, ships can be traced along their actual route.  In some places
this approach may vary significantly from the theoretical ship lanes
available from the USACE Waterways Network.  

Recommendations

Information on ship activity data is not currently available in a format
ready to use for an air quality modeling study.  For the Pacific coast,
a moderate amount of work would be required to complete the processing
of the available data into a format suitable for air quality modeling. 
For the other areas, a large amount of work would be required based on
the data that we identified.  One may consider using hypothetical ship
emissions for this air quality modeling study; however, those
hypothetical emissions should be representative of actual ship emissions
in order to lead to realistic air quality predictions.

6.	REFERENCES

California Air Resources Board. 2000.  Air Quality Impacts from NOx
Emissions of Two  Potential Marine Vessel Control Strategies in the
South Coast Air Basin. Final  Report, prepared by the California Air
Resources Board and the South Coast Air  Quality Management District in
Consultation with the Deep Sea Vessel/Shipping  Channel Technical
Working Group, Sacramento, CA.

EC, 2002.  Quantification of emissions from ships associated with ship
movements between ports in the European Community, Final Report prepared
by Entec UK Limited.

Eddington, L. and J. Rosenthal, 2003.  The Frequency of Offshore
Emissions Reaching the Continental United States Coast Based on Hourly
Surface Winds from a 10 Year Mesoscale Model Simulation, Naval Air
Systems Weapons Division, Point Mugu, CA.

EPA, 2000.  Analysis of Commercial Marine Vessels Emissions and Fuel
Consumption Data, EPA-420-R-00-002, U.S. Environmental Protection
Agency, Office of Transportation and Air Quality, Washington, DC

EPA, 2002.  Commercial Marine Emission Inventory, Final Report from
PECHAN, prepared by ENVIRON International Corporation, U.S.
Environmental Protection Agency, Office of Transportation and Air
Quality, Ann Arbor, MI.

.

EPRI, 2000.  SCICHEM Version 1.2: Technical Documentation, EPRI Report
1000713, EPRI, Palo Alto, CA.

Federal Register, 2000. Environmental Protection Agency, 40 CFR Part 51,
Requirements for Preparation, Adoption, and Submittal of State
Implementation Plans (Guideline on Air Quality Models); Proposed Rule,
Vol. 65, No. 78, Friday, April 21, 2000, pp. 21506–21542.

Federal Register, 2003.  40 CFR Part 51, Revision to the Guideline on
air quality Models: Adoption of a Preferred Long Range Transport Model
and Other Revisions; Final Rule, 18440-18482, April 15, 2003.

Gupta, M., N. Kumar, P. Karamchandani, and S.-Y. Wu, 2001. 
Intercomparison of SCICHEM and CLAPUFF models using Cumberland plume
data, Air & Waste Management Association Conference on Guidelines on Air
Quality Models: A New Beginning, April 4-6, Newport, RI.

Hanna, S., L. Schulmann, R. Paine, and J. Pleim, 1984.  The Offshore and
Coastal Dispersion (OCD) Model User’s Guide Revised, MMS 84-0069,
Minerals Management Service.

Hanna, S., L. Schulman, R. Paine, J. Pleim and M. Baer, 1985. 
Developoment and evaluation of the offshore and coastal dispersion (OCD)
model, J. Air Pollut. Control Assoc., 35, 1039-1047.

Karamchandani, P.; A. Koo; C. Seigneur, 1998. Environ. Sci. Technol.,
32, 1709–1720.

Karamchandani, P.; L. Santos; I. Sykes; Y. Zhang; C. Tonne; C. Seigneur,
2000. Environ. Sci. Technol., 34, 870–880.

Morris, R.E., R.C. Kessler, S.G. Douglas, K.R. Styles and G.E. Moore,
1988.  Rocky Mountain Acid Deposition Model Assessment: Acid Rain
Mountain Mesoscale Model (ARM3), Report prepared for the U.S. EPA,
Research Triangle Park, NC.

National Ocean and Atmospheric Administration (NOAA), 2005.    HYPERLINK
"http://www.cdc.noaa.gov/coads/"  http://www.cdc.noaa.gov/coads/  
“ICOADS” Last accessed April 20, 2005.

National Data Buoy Center (NDBC), 2005.    HYPERLINK
"http://www.ndbc.noaa.gov/index.shtml" 
http://www.ndbc.noaa.gov/index.shtml  “National Data Buoy Center” 
Last accessed April 20, 2005.

Scire, J.S., D.G. Strimaitis and R.J. Yamartino, 2000. A User’s Guide
for the CALPUFF Dispersion Model (Version 5), Earth Tech, Inc. Report,
Concord, MA, January 2000.

Scire, J.S., F.R. Robe, M.E. Fernau and R.J. Yamartino, 2000. A User’s
Guide for the CALMET Dispersion Model (Version 5), Earth Tech, Inc.
Report, Concord, MA, January 2000.

Scire, J.S., 2005.  Communication via e-mail of Christian Seigneur, AER,
with Joe Scire, EarthTech, 18-19 March 2005.

Shi, J.P., D. Mark and R.M. Harrison, 2000.  Characterization of
particles from a current technology heavy-duty diesel engine, Environ.
Sci. Technol., 34, 748-755.

Song, C.H., G. Chen and D. Davis, 2003.  Chemical evolution and
dispersion of ship plumes in the remote marine boundary layer:
investigation of sulfur chemistry, Atmos. Environ., 37, 2663-2679.

Sykes, R. I., W. S. Lewellen, S. F. Parker and D. S. Henn, 1988.  A
Hierarchy of Dynamic Plume Models Incorporating Uncertainty, Volume 4:
Second-order Closure Integrated Puff, EPRI, EPRI EA-6095 Volume 4,
Project 1616-28.

Sykes, R. I.; S. F. Parker; D. S. Henn; W. S. Lewellen, 1993 J. Appl.
Met., 32, 929–947.

Sykes, R. I.; D. S. Henn, 1995. J. Appl. Met., 34, 2715–2723.

United States Army Corps of Engineers, 2005.    HYPERLINK
"http://www.iwr.usace.army.mil/ndc/data/data1.htm" 
http://www.iwr.usace.army.mil/ndc/data/data1.htm   “Navigation Data
Center – U.S. Waterway Data”  Last accessed April 20, 2005.

Wheeler, N., 2005.  Private communication from Neil Wheeler, Sonoma
Technology, Inc. to Christian Seigneur, AER, 14 March.

Yocke, M.A. et al., 1998.  Meteorology of the northeastern Gulf of
Mexico. ENVIRON International Corp. U.S. DOI. OCS Study: final report,
data from 1995 to 1997. 2000. 154 p. Available from GOM (with 3 CD's).
MMS 2000-075.

APPENDIX B

ANALYSIS PLAN

MODELING SULFUR OXIDES (SOx) EMISSIONS TRANSPORT

FOR SHIPS AT SEA

Prepared by

Christian Seigneur

Prakash Karamchandani

Kristen Lohman

Atmospheric & Environmental Research, Inc.

2682 Bishop Drive, Suite 120

San Ramon, CA 94583

Prepared for

U.S. Environmental Protection Agency

Office of Transportation and Air Quality

1200 Pennsylvania Avenue, NW

Washington, DC 20460

Document CP212-05-02a

July 2005

INTRODUCTION

This document describes the analysis plan for modeling the SO2 and
sulfate concentrations due to emissions of SOx from ships at sea.  The
results of this screening modeling study will provide quantitative
information on the shortest distance at which ships burning higher
sulfur fuel (here, 27,000 ppm) will have air quality impacts at land
receptors that are less than those anticipated from emissions from ships
burning low sulfur fuel (here, 15,000 ppm) within coastal waters.  This
resulting distance can subsequently be used as the basis for defining
the modeling domain for sources to be included in a subsequent modeling
study using an Eulerian model (CMAQ).  The results of the CMAQ modeling
will yield information to define the outer boundary of a Sulfur Emission
Control Area (SECA).  We focus here on the southern Pacific coast.  The
methodology presented here is consistent with an approach developed by
the Office of Transportation and Air Quality (OTAQ) of the U.S.
Environmental Protection Agency (EPA) which included input from EPA
regional modelers, and staff from the U.S. Navy. 

We first describe the overall modeling approach including the fate and
transport model, CALPUFF, that will be used to simulate the transport,
transformation and removal of pollutants over water and land.  Then, we
describe the selection of the model input data including meteorological
data, SOx emissions and ship activity data.

2.	AIR QUALITY MODELING APPROACH

Air Quality Model

For this screening study of the potential impacts of SOx emissions from
ships at sea, we will use the CALPUFF model (Scire et al., 2000a,
2000b).  CALPUFF is a non-steady-state puff dispersion model that can
simulate the effects of time- and space-varying meteorological
conditions on pollutant transport, transformation, and removal.  The
rationale for selecting CALPUFF was described in the Task 1 report
(Seigneur et al., 2005). 

The recommended meteorological inputs for applying CALPUFF are the
time-dependent outputs of CALMET, a meteorological model that contains a
diagnostic wind field module and overwater and overland boundary layer
modules.  Optionally, CALMET can use the outputs of prognostic
meteorological models, such as MM5 and CSUMM, to create the
meteorological fields required by CALPUFF.  The preparation of the
meteorological data inputs for CALPUFF for this study is described in
Section 3.

CALPUFF includes algorithms for near-source effects such as building
downwash, transitional plume rise, partial plume penetration, sub-grid
scale terrain interactions as well as longer range effects such as
pollutant removal due to wet and dry deposition, simplified chemical
transformations, vertical wind shear, overwater transport and coastal
interaction effects.  Because the latter features are relevant to
simulating the transport and chemistry of SOx emissions from ships, they
will all be activated for our study.

CALPUFF offers several options to simulate the formation of secondary
sulfate and nitrate particles from the oxidation of the emitted primary
gaseous pollutants, SO2 and NOx respectively.  Since the oxidation of
SO2 to sulfate is of interest for this study, we will select the more
advanced chemistry module available in CALPUFF which is based on the
RIVAD/ARM3 chemical mechanism (Morris et al., 1988).  The limitations of
this chemistry module were discussed in the Task 1 report (Seigneur et
al., 2005).

Modeling Domain

The modeling domain for the southern Pacific coast will extend from
about 32 degrees North to 36 degrees North and will, therefore, cover
southern California.  (Northern California will be grouped with Oregon
and Washington, i.e., from 36 degrees North to 50 degrees North, to
constitute the modeling domain for the northern Pacific coast.)  The
modeling domain will extend 240 km (150 miles) inland to allow enough
distance to assess the potential air quality impacts of emissions from
ships at sea.  It will extend over water at a distance from the coast
that corresponds to air quality impacts below the target concentration
at all receptors.

Physiographic data (coastline and terrain elevation) will be obtained
from the U.S. Geological Survey.

Receptors

Receptors will be located on land as follows.  A line of receptors will
be located at the coastline, 10 km apart.  Such a distance provides a
finer spatial resolution than that of the ship emissions along the coast
(see Section 5).  Inland receptors will then be located eastward at 10,
10, 20, 20, 30, 30, 40, 40 and 40 km apart from each other, i.e., up to
240 km (150 miles) from the coastline; there will, therefore, be 10
lines of receptors from the coast (included) up to 240 km inland.  All
receptors will be located at ground level.  The total number of
receptors is, therefore, estimated to be on the order of 500.

Sources

Ship emissions will be represented by a set of stationary point sources.
 Each point source will represent a ship.  They will be located at a
selected distance from shore (see below) and apart at a distance to be
defined based on ship traffic (see Section 5).  The use of stationary
sources to represent moving ships is an appropriate approximation for
this screening modeling study, because using stationary sources will
overestimate the downwind air quality impacts (emissions will be
concentrated in specific locations rather than continuously distributed
along the shipping lane, thereby leading to greater ambient air
concentrations).

We considered but rejected an alternative approach.  The approach would
treat each ship as an individual source and simulate its impact on air
quality inland.  Target concentrations would be calculated from
individual ships at the coast (dockside mode) with the highest
concentration obtained at each receptor being selected as the target
concentration for that receptor.  Then, the impacts of individual ships
would be evaluated against those target concentrations.  This
alternative approach offers the advantage of providing more detailed
information regarding the impacts from ships since it addresses
individual ships rather than a shipping lane; thus, different SECA
distances could be identified in different parts of the domain.  Such an
approach requires many more model simulations than the approach proposed
here, however, and, therefore, could not be considered for this
screening study. Also, comparing with the highest concentration obtained
for that receptor does not account for variability of concentrations at
receptors, and may result in an overestimation of the boundary distance.
 Nevertheless, we point out below how the variability of the SECA
distance within the study domain will be addressed.

Modeling Approach

Our modeling approach will consist of two phases.  In the first phase,
we will calculate, at each inland receptor, the target values for the
SO2 and sulfate concentrations that correspond to emission from ships at
dockside; i.e., those ships that are within the SECA and therefore must
burn low sulfur fuel; i.e., 15,000 ppm).  These will be annual average
concentrations. (It is not necessary to calculate the light extinction
coefficient because it will be proportional here to the sulfate
concentration.)  In the second phase, we will calculate the annual
average values of the SO2 and sulfate concentrations corresponding to
emissions from ships burning high sulfur fuel (i.e., 27,000 ppm) at
various distances from the coast and will compare those values to the
target values obtained in the first modeling phase.

All simulations will be conducted for one year and we will calculate and
use annual average values in our analysis.  We propose to use 2002 as
our reference year because it corresponds to the year that will be used
for grid-based air quality modeling by the EPA Office of Air Quality
Planning and Standards (OAQPS).

For the first phase, we will locate the ships at the coastline (dockside
mode).  They will be distributed spatially according to their estimated
density in a shipping lane (see Section 5).  The SECA SOx emission rates
will be used (see Section 4).  We will calculate the annual SO2 and,
sulfate concentrations at each receptor.  These values will be defined
as the target values that will be used as benchmarks for the Phase 2
modeling.

For the second phase, we will locate the ships at various distances from
the coastline.  For a given modeling scenario, all ships will be at the
same distance from the coastline; they will be distributed spatially
according to their estimated density in a shipping lane (see Section 5),
and for all modeling scenarios the number of dockside ships will equal
the number of off-shore ships.  The SOx emission rates outside of the
SECA will be used (see Section 4).  The objective is to determine a set
of distances at which those ship emissions will lead to air quality
impacts that are less than or equal to the target values calculated in
Phase 1 for the following percents of onshore receptors: 50, 60, 70, 80,
and 90.  To that end, we will conduct CALPUFF annual simulations for
various distances from the coastline.  We will start with a 100 km
distance, and receptor percentage of 50.  If the modeling results show
air quality impacts lower than the target values, at 50 percent or more
of the onshore receptors we will then use a shorter distance (50 km). 
Conversely, if the modeling results show air quality impacts greater
than the target values at 50 percent or more of the onshore receptors,
we will use a greater distance (200 km).  This process will be repeated
until we identify the distance of interest (i.e., the distance where air
quality impacts are commensurate with the target values).  For example,
if the modeling results conducted for a distance of 50 km show air
quality impacts lower than the target values, for at least 50 percent of
the onshore receptors, we will next use a shorter distance (20 or 30
km).  If those modeling results show air quality impacts greater than
the target values for at least 50 percent of the onshore receptors, we
will then use a greater distance (70 or 80 km).  We will stop when we
have identified a distance that leads to air quality impacts
commensurate with the target values.  This process will be repeated for
the other percentages of onshore receptors (60, 70, 80 and 90).  For all
percentages, a tolerance of plus/minus 2 percent will be used.  We
propose to use a resolution of 10 km (i.e., we will not refine those
distances within less than 10 km increments).

The criterion of percentages of onshore receptors is used as an initial
investigation.  As we approach the distance of interest, some receptors
will show values greater than the target values whereas other receptors
may show values lower than the target values.  The distribution of these
receptors is significant.  For example, by definition, fewer receptors
have concentrations in excess of target concentrations at the 60% level
than 50%.  But if the receptors in excess of the target concentrations
at both the 50 and 60% levels are located say, within 10 km of the
coastline, then even at the greater distances comparable levels of
population may still be exposed to concentrations greater than target
levels.  In this example, the distribution may indicate that a greater
distance should be considered.  Therefore, evaluation of these various
distances will be conducted by the modeling review team as part of the
Task 3 analysis.   

Another reason for using the criterion of percentages of onshore
receptors is that SO2 and sulfate concentrations will display different
behaviors downwind of the ships.  SO2 concentrations will decrease
continuously with distance from the source (due to dilution, removal,
and conversion to sulfate), whereas sulfate concentrations will first
decrease (dilution and removal of primary, i.e., directly emitted
sulfate), then increase (formation of secondary sulfate from the
oxidation of SO2) before finally decreasing (dilution and removal
exceeding formation).

This behavior of sulfate introduces an additional complication:  the
sulfate target values at receptors near the coastline will be determined
by the directly emitted sulfate, while the target values at larger
distances inland will be determined by some combination of primary and
secondary sulfate, with the secondary sulfate component increasing and
the primary sulfate component decreasing.  Even further inland, both
components will decrease as the rate of dilution and removal exceeds the
formation of sulfate.

To understand how this complex behavior of sulfate may impact the
analysis, let us consider the extreme case of no primary sulfate, i.e.,
all the SOx is emitted as SO2.  In this case, the target sulfate values
next to the coastline will be negligible because there will be minimal
time for conversion of SO2 to sulfate.  However, there will be some
plume travel time for emissions from ships at sea that will allow some
conversion of SO2 to sulfate.  Consequently, it may be impossible in
this extreme case to meet target values at the coastline receptors
unless a very large SECA is defined.

Therefore, it is possible that all sulfate concentrations may not fall
below the target values as we approach the distance of interest for the
SECA.  Accordingly, we will need to report the results in terms of the
fraction (or percentage) of receptors that exceed the target values for
each pollutant.

We will report the results for each distance in terms of maximum
concentration, average concentration and fraction of receptors above the
target value for SO2 and for sulfate (all values will be for receptors
over land).  If significant differences appear for different areas of
the study domain (e.g., one area shows impacts above target
concentrations for at least 50 percent of the onshore receptors for a
shorter distance than another area), we will identify those differences
and discuss whether they suggest the need for some variability for the
SECA distance within the study domain.

3.	METEOROLOGICAL DATA

CALMET is the companion meteorological model that is used to prepare the
meteorological fields used by CALPUFF.

A weakness of CALMET has recently been identified (Wheeler, 2005). 
CALMET does not correctly handle cases of unstable convective
atmospheric conditions over water (when water temperature is warm and
air temperature is cold, for example) because it assumes near-neutral
conditions over water.  Consequently, the mixing height is calculated
based on a neutral mixing relationship and, under conditions of light
wind speeds when the mechanical mixing heights are small, CALMET
underpredicts the actual mixing height.  This weakness can be an issue
in areas where warm water temperatures are possible, such as the Gulf of
Mexico, the southern Pacific coast and the southern Atlantic coast. 
Therefore, we address this potential issue here as it is important for
this area as well as for subsequent modeling areas.  Based on our
discussion with the CALMET developer, EarthTech (Scire, 2005), we will
circumvent this potential problem by inputting measured or modeled
mixing heights directly into CALMET.  For the southern Pacific coast, no
upper air measurements are available over water and we will, therefore,
use modeled mixing heights, as described below.

Meteorological data are necessary to run an air dispersion model.  For
the CALPUFF model, the meteorological input data must first be formatted
by the CALMET pre-processor.  CALPUFF requires standard surface and
upper air meteorological data.  CALMET also has an overwater option that
allows the use of special overwater measurements for grid cells that are
over the ocean.  The data required for the overwater option are: air-sea
temperature difference, air temperature, relative humidity, wind speed
and wind direction.  Two optional measurements, overwater mixing height
and overwater temperature gradients, may be supplied if available.  If
the optional parameters are not supplied, CALMET uses default values. 
We propose to supply temperature gradients obtained from the outputs of
a prognostic meteorological model (see below).

Land-based Measurements

Land-based meteorological measurements are required for both surface and
upper air observations above land portions of the domain.  The data
required are standard format data from the National Climatic Data Center
(NCDC) (Scire et al., 2000b).

The upper air data required are standard NCDC format TD6201 radiosonde
data including pressure, elevation, temperature, wind direction, and
wind speed for each sounding level.  There are four upper air stations
that are located within the modeling domain:

San Nicolas Island (33.25 degrees North, -199.45 degrees West)

Miramar (32.87 degrees North, -117.15 degrees West)

Point Mugu (34.10 degrees North, -119.12 degrees West)

Vandenberg (34.67 degrees North, -120.58 degrees West)

The surface observations that are needed are provided in the NCDC
Integrated Surface Hourly Observations.  These include wind speed, wind
direction, temperature, and dew point temperature.  There are many
surface stations within the modeling domain (255 for the state of
California).

Overwater Measurements

The required CALMET parameters are all available for the Pacific Ocean
near the U.S. coastline from the National Data Buoy Center (NDBC) (NDBC,
2005).  The measurements are taken from buoys.  The buoys are at varying
distances from the coast.  Those near the coast are frequently near
harbors or bays.  Most of the buoys are owned and operated by NDBC but
there are also several other agencies that submit their data to the NDBC
database.  Though the coverage is not uniform, there is a fairly
comprehensive coverage for the southern Pacific coast.    Figure 3-1
shows the locations of the NDBC buoys as well as those that are run by
other agencies and are included in the NDBC database.

Model Outputs

The outputs of meteorological models can be used, particularly in cases
where there are insufficient meteorological observations.  This is the
case for upper air data over water in the Pacific Ocean.  Examples of
model outputs that could be used as surrogates for upper air data
include those from the 2001 or 2002 MM5 simulations sponsored by EPA,
those from the NCEP/NCAR reanalysis project and those from the Advanced
Climate Modeling and Environmental Simulations (ACMES) database.

CALMET can take as input the output of MM5.  It can also combine MM5
output with observations.  An interface program (CALMM5) converts the
MM5 data into a form compatible with CALMET.  A new version of this
processor has been added to the CALPUFF-CALMET Download BETA-Test page
recently (May 25, 2005).  This beta version (not yet officially approved
by the EPA) of CALMM5 processes MM5 Version 3 output data directly. 
Using the output of another meteorological model (e.g., ACMES) would
require the development of a new CALMET pre-processor that would be
outside the scope of this project.  Therefore, we will use the MM5
output for this application.  Another advantage of using the MM5 outputs
is that it will provide consistency with the subsequent grid-based
modeling that will be conducted by OAQPS using the Community Multiscale
Air Quality model (CMAQ), because CMAQ will be driven with the MM5
meteorology.

The MM5 modeling domain covers the entire contiguous United States and
extends significantly over the oceans.   For the southern Pacific coast
domain, it extends at least 400 to 900 km westward from the coast. 
Therefore, it will cover the CALPUFF modeling domain needed to address
the SECA.



Figure 3-1.	NDBC buoys along the southern California coastline

Summary

We will use a combination of MM5 model output, surface observations over
water from the NDBC database, surface observations over land from the
NCDC database and upper air observations over land from four stations
from the NCDC database.  These data will be processed by CALMET to
prepare a three-dimensional meteorological data set for CALPUFF.

 4.	SOx EMISSIONS

Emission factors are needed to estimate the emissions of SOx (gas-phase
SO2 and particulate-phase sulfate) associated with various ship
activities.  Based on the review of available emission factors of
Seigneur et al. (2005), the most recent EPA emission factors were
selected (EPA, 2002).  Those emission factors pertain to ships with
engines with displacement exceeding 30 liters (so-called Category 3
engines).

Emission factors are reported for three different engine types (slow
speed, medium speed and steam boiler) for transit modes and hoteling
modes.  For this study of ships at sea, we are interested in medium
speeds for transit modes.

The SO2 emission factor per unit of work is reported to be 9.56 g/hp-h
for a 3% sulfur fuel (i.e., 30,000 ppm) for a ship at slow or medium
speed in transit mode.  This is equivalent to 12.8 g/kW-h.

For a ship within the SECA, a fuel sulfur content of 15,000 ppm will be
assumed.  Therefore, the emission factor will be 6.4 g/kW-h.

For a ship at sea outside of the SECA, a fuel sulfur content of  27,000
ppm will be assumed.  Therefore, the emission factor will be 11.52
g/kW-h.

EPA assumes that 2% of sulfur is emitted as primary sulfate PM from
Category 3 marine diesel engines.  Therefore, we treat 2% of total
sulfur emissions as sulfate emissions and the SO2 emission factor is
adjusted down accordingly to maintain the sulfur mass balance. (Note
that for the same amount of S, the sulfate emission factor is 1.5 the
SO2 emission factor to account for the different molecular weights.)

Therefore, within the SECA, the gas-phase SO2 and particulate-phase
sulfate emission factors will be 6.27 g/kW-h and 0.19 g/kW-h,
respectively.  Outside of the SECA, the gas-phase SO2 and
particulate-phase sulfate emission factors will be 11.29 g/kW-h and 0.35
g/kW-h, respectively.

The sulfate emission rates calculated above are consistent with
available data on the sulfate fraction of particulate matter (PM)
emitted from ship diesel engines.  Fleischer et al. report that 20 to
30% of PM emissions from ship diesel engines are sulfate (for a 3%
sulfur fuel content).  The EPA (2002) emission factor for PM is 1.3
g/hp-h, i.e., 1.74 g/kW-h.  These values lead to an emission factor for
sulfate in the range of 0.31 to 0.47 g/kW-h for a sulfur fuel content of
27,000 ppm.  The emission factor of 0.35 g/kW-h calculated above falls
within this range.

Based on data from Corbett and Koehler (2003), the power of a typical
ship was estimated to be 16,000 kW (Corbett, 2005).  It should be noted
that there is a wide range of power among various ships, with the
largest container ships having power exceeding 65,000 kW.

The gas-phase SO2 and particulate-phase sulfate emissions per ship are
then calculated to be 100,320 g/h and 3,040 g/h, respectively, within
the SECA and 180,640 g/h and 5,600 g/h, respectively, outside the SECA.

5.	SHIP ACTIVITY DATA

Ship activity data must be estimated so that the density of ships within
the modeling domain can be calculated.  Knowing the average number, N,
of ships in transit along the southern Pacific coast per year and
assuming an average cruising speed, V (km/h), we can calculate the
average distance, D (km), between two ships along a shipping lane.

D = V * (24 h/day * 365 days/yr) / N

The annual number of ships transiting along the southern California
coast was estimated to be 13,000 (ICOADS, 2002).  This number includes
all ships transiting to and from ports located on the southern Pacific
coast as well as ships transiting southward/northward from/to ports
located on the northern Pacific coast.  It is likely to be an
overestimate of the number of ships transiting along the coast because a
fraction of those ships will be transiting along shipping lanes that
extend from the ports westward into the Pacific Ocean.  The cruising
speed varies according to ship type.  It is about 24 knots for container
ships and about 16 knots for tankers.  Here, the average ship cruising
speed was estimated to be about 20 knots, i.e., 36 km/h (ICOADS, 2002). 
Thus, the average distance is estimated for the southern Pacific coast
as follows.

D = 36 * 24 * 365 / 13,000 = 24.3 km

Based on this analysis, we propose to use a distance of 25 km between
ships to calculate ship emissions. 

6.	REFERENCES

Corbett, J.J. and H.W. Koehler, 2003.  Updated emissions from ocean
shipping, J. Geophys. Res., 108, doi:10.1029/2003JD003751.

Corbett, J.J., 2005.  Private communication to Christian Seigneur, AER,
July 2005.

EPA, 2002.  Commercial Marine Emission Inventory, Final Report from
PECHAN, prepared by ENVIRON International Corporation, U.S.
Environmental Protection Agency, Office of Transportation and Air
Quality, Ann Arbor, MI.

Fleischer, F., E.J. Ulrich, R. Krapp and W. Grundmann.  Comments on
particulate emissions from diesel engines when burning heavy fuels.

.

ICOADS, 2002.  International Comprehensive Ocean Atmospheric Data Set,
as transmitted from ERG by Office of Transportation and Air Quality,
U.S. Environmental Protection Agency, Washington, D.C.

Morris, R.E., R.C. Kessler, S.G. Douglas, K.R. Styles and G.E. Moore,
1988.  Rocky Mountain Acid Deposition Model Assessment: Acid Rain
Mountain Mesoscale Model (ARM3), Report prepared for the U.S. EPA,
Research Triangle Park, NC.

Scire, J.S., D.G. Strimaitis and R.J. Yamartino, 2000a. A User’s Guide
for the CALPUFF Dispersion Model (Version 5), Earth Tech, Inc. Report,
Concord, MA, January 2000.

Scire, J.S., F.R. Robe, M.E. Fernau and R.J. Yamartino, 2000b. A
User’s Guide for the CALMET Dispersion Model (Version 5), Earth Tech,
Inc. Report, Concord, MA, January 2000.

Scire, J.S., 2005.  Communication via e-mail of Christian Seigneur, AER,
with Joe Scire, EarthTech, 18-19 March 2005.

Seigneur, C. K. Lohman and P. Karamchandani, 2005.  Review of Technical
Information relevant to Sulfur Oxides (SOx) Emissions Transport for
Ships at Sea, Final Report to Office of Transportation and Air Quality,
U.S. Environmental Protection Agency, Washington, D.C.

Wheeler, N., 2005.  Private communication from Neil Wheeler, Sonoma
Technology, Inc. to Christian Seigneur, AER, 14 March.

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	 
PAGE  viii 

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	E- 
PAGE  4 

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	1- 
PAGE  2 

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	2- 
PAGE  10 

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	3- 
PAGE  28 

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	4- 
PAGE  58 

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	5- 
PAGE  3 

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	6- 
PAGE  2 

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	A- 
PAGE  21 

Modeling Sulfur Oxides (SOx) Emissions Transport From Ships At Sea	B- 
PAGE  14 

