Chapter 3. Atmospheric iron deposition in the Northwestern Pacific Ocean and its
3.2.1 Numerical models
CMAQ considers the chemistry, transport, and deposition processes associated
with trace gases and aerosol particles. In the model, transport through atmospheric
circulation is driven by meteorological fields simulated using the Fifth-Generation Penn
State/National Center for Atmospheric Research NCAR Mesoscale Model (MM5)
[Grell et al., 1994], with the final analysis data provided by National Centers for
Environmental Prediction (NCEP) as the initial and boundary conditions. The physical
schemes used in the MM5 include a simple ice moisture scheme, Kain-Fritsch cumulus
scheme, Medium Range Forecast planetary boundary layer scheme, and Noah land
surface multilayer soil temperature model. The simulation covered continental East Asia
and the NWPO and NSCS (Figure 3.1), with 85 and 67 grids of 81 km by 81 km
resolution in the latitudinal and longitudinal directions, respectively. A total of 31 layers
was applied in the vertical direction using the sigma coordinate [Arakawa and Suarez,
1983], with 9 grids below the 0.8 sigma level (roughly the height of the planetary
boundary layer) and 10 grids above the 0.2 sigma level (roughly the height of the
tropopause). We employed the inventory of Zhang et al. [2009] for the emission rates of
natural and anthropogenic gases. This inventory includes the emission rates of aerosol
particles such as elementary carbon and organic carbon; however, it lacks emission data
related directly to dust and fly ash, which are the main focus of this study. The methods
used to estimate the emissions of dust and fly ash are discussed in section 3.2.2.
The ability to describe particle size distribution with a high degree of precision is
crucial to the calculation of wet and dry deposition rates. We substituted the aerosol
scheme in CMAQ with the accurate and computationally efficient Statistical-Numerical
Aerosol Parameterization (SNAP) scheme developed by Chen et al. [2013]. The SNAP
scheme uses several log-normal modes to represent particle size distribution and takes
three moments of size distribution (i.e. number, surface area and mass concentrations)
as prognostic variables for each mode. In this study, mineral dust and fly ash particles
were assumed to exist in two modes -- fine and coarse. They were assumed to mix
externally (i.e., exist in the same air parcel but not the same particle) with other aerosol
species.
In calculating the rates of dry deposition of aerosol particles, we took into account
mechanisms associated with aerodynamic resistance, quasi-laminar resistance and
gravitational sedimentation [Wesely, 1989]. Gravitational sedimentation has been
parameterized as a function of the particle size spectrum and ambient conditions in
SNAP. In CMAQ, the calculations used to determine the rates of wet deposition are
performed in two steps for each of the two types of precipitation specified in the
meteorological model: subgrid convective precipitation and grid point (nonconvective)
precipitation. The cumulus parameterization scheme used to calculate sub-grid
precipitation indicates rain intensity only at the surface, with no information with regard
to rain in the air or the size of raindrops. Thus, wet deposition rates can only be
calculated to a rough approximation using an empirical formula [Pruppacher and Klett,
1997; Wang et al., 2000], which is directly proportional to precipitation rate and particle
mass concentration. A lack of information on airborne precipitation particles prevent us
from calculating in-cloud scavenging, so we only calculated below-cloud scavenging
due to convective precipitation. However, the explicit microphysical scheme used to
calculate grid point precipitation provides information on airborne precipitation
particles. This made it possible to treat wet scavenging in greater detail using the SNAP
scheme with consideration of wet scavenging due to Brownian diffusion, phoretic forces,
and gravitational impaction. We also keep track of aerosols collected by raindrops such
that in-cloud scavenging can be calculated.
3.2.2 Dispersion of Iron
Our model deals with two types of airborne iron: iron from mineral dust deflation
and from the anthropogenic burning of coal. Estimating iron dispersion requires
information related to the overall mass flux and particle chemical composition.
3.2.2.1 Iron in airborne dust
To model the dispersion of iron in mineral dust, this study adopted the dust
deflation module developed by Wang et al. [2000], which provided accurate results
when incorporated into the Taiwan Air Quality Model (TAQM) by Chen et al. [2004]
and into CMAQ by Chen et al. [2013]. Calculating the size distribution of dust particles
in TAQM involves applying 12 size bins. However, Chen et al. [2013] demonstrated
that the SNAP scheme, which applies two lognormal modes with three moments each,
can provide essentially the same results as the binned approach, with significantly less
computation time. Thus, we adopted the SNAP scheme in the CMAQ to describe the
processes associated with airborne mineral dust.
In accordance with the approach of Wang et al. [2000], the intensity of dust
deflation was described as a function of relative humidity (a surrogate for soil moisture),
friction velocity, and the characteristics of the land. The types of land that provide dust
most readily are deserts and semiarid areas, as well as nondesert areas such as cropland
and grassland, in accordance with the MODIS classification [Hall et al., 2002]. The size
distribution of deflated dust varies with land type according to the specification outlined
by Wang et al. [2000] (Table 3.1). It should be noted that the simulation of Luo et al.
[2008] included dust from desert and desertified areas; however, they disregarded dust
emissions from nondesert areas. This study accounted for the effects of snow cover on
the dispersion of dust by linearly reducing dust emission according to the MODIS snow
cover fraction.
Iron content in mineral dust was estimated according to the mass fraction of iron in
various source materials (Figure 3.2). The iron content in the dust originating in the
deserts of Asia is relatively constant compared to other species shown in Figure 3.2,
with a mass fraction ranging from 1.2 to 5.9%. This study applied a median value of
4.00%, in accordance with that used by [Zhang et al., 2003]. The values for agricultural
and other types of soil are in a similar range, with a mean value of 4.05%. Figure 2 also
listed the aluminum (Al) content which we use to help identify the lithogenic iron
sources as suggested by Ho et al. [2010].
3.2.2.2 Iron in fly ash
Most previous studies dealing with atmospheric iron deposition into the oceans
have failed to consider the contribution of fly ash produced by coal burning. Luo et al.
[2008] and Ito and Feng [2010] simulated iron emissions associated with coal burning
using a global model that assumed fixed fractions of fine-mode and coarse-mode
particulate matter (PM) in the emissions. Their PM emissions data were obtained from
Bond et al. [1998] and Streets et al. [2001], while the proportion of fine-mode particles
in the total PM was obtained from the U.S. Environmental Protection Agency [Streets et
al., 2001]. However, North America data are not necessarily applicable to the East Asian
region. Zhang et al. [2009] implemented a series of improved methodologies to gain a
better understanding of pollutant emissions from China and other Asian countries. They
provided important anthropogenic data for SO2, NOx, BC, CO, and PM, but excluded
emissions from biomass burning.
Luo et al. [2008] have estimated Fe emissions as a fraction of PM emissions.
Because the fraction is not well known for East Asia, we use an alternative approach
that estimates Fe emission from SO2emissions and the chemical analysis of coal and
coal fly ash from Chinese data (Appendix 3.A). We then use the reported SO2emissions
in China, which generally have uncertainties within 14% [Streets et al., 2003; Zhao et
al., 2011], to establish the relationship between S and Fe emissions associated with coal
burning and steel smelting (Figure 3.3). This approach assumes that other anthropogenic
sources of aerosol Fe are negligible.
In accordance with the mass balance principle, the sulfur contained in coal (Sc) is
converted to gaseous SO2(Sg), sulfur in fly ash (Sf), and sulfur in bottom ash (Sb) via
the process of burning. Similarly, iron in coal ( ୡ) is converted to iron in fly ash ( ),
and iron in bottom ash ( ୠ). These mass balances can be written as follows:
ୡ ൌ ୠ (3.1)
ୡ ൌ ୠ (3.2)
The ratio of iron in fly ash to SO2generated from coal burning, which we call the
composition factor, can then be derived as follows:
ୣ
ୗౝ ؠ Ƚ ൌ ౙ ଵ
ూౙሺଵାሻି
ూିౘ
ూౘ (3.3)
where q is the ratio of bottom ash to fly ash, under the assumption that the iron fraction
in fly ash and in bottom ash is the same. Information on the q value is difficult to obtain,
particularly with regard to the East Asian region. Only one report was available, which
is for the power plants in Taiwan [Yen, 2011], with a q value of 0.25, which is the same
as that reported in Kim et al. [2005] for North America. By adopting this q value and
data shown in Figure 3.3 and Tables 3.A1-3.A3, we estimated the composition factor α
to be about 1.5 (for details see Appendix 3.A).
However, the above derivation fails to consider the recovery of sulfur or fly ash,
which is widely applied in power plant operations and a number of industries to reduce
air pollution. Equation (3.3) therefore must be modified by adding the following
coefficient:
Ⱦ ൌሺଵିሻ
ሺଵିೞሻ (3.4)
where ݎ and ݎ௦ are the recovery ratios of fly ash and SO2, respectively. We call Ethe “recovery factor.” The net emission flux of iron (in the form of fly ash) can then
be written as follows:
ୣ ൌ Ƚ ή Ⱦ ή ୗ (3.5)
where ୗ is the emission flux of sulfur (in the form of SO2). The parameters D and E
have very little data to constrain them, since they depend on the deployment of
desulphurization technologies in coal-burning power plants. Values used for E in the
model are 0.033, 0.2, and 1.0 for power plants, industry, and residential sources,
respectively (for details see Appendix 3.A).
An important factor that was not considered in the estimates of Fe emissions by
Luo et al. [2008] and Ito and Feng [2010] is the Fe enrichment during steel
manufacturing. This issue is particularly important in East Asia, because iron and steel
industries are among the most energy-intensive industries worldwide, and China
produced 35% of the world’s steel in 2007 and 47% in 2010 (IEA statistics). Particles
emitted from iron and steel plants are known to be rich in iron. Iron mass fraction in
steel plant fly ash is variously reported as 26.0% [Tsai et al., 2007], 27.6% [Geagea et
al., 2007], or 35.0% [Prati et al., 2000], giving a geometric mean of 29.3%. These
values are substantially higher than the 4.0%, 2.0%, and 4.8% of mineral dust, coal and
coal-burning fly ash, respectively (the latter two values are averages of data provided in
Tables 3.A1 and 3.A2). Due to a lack of other information, this study applied the 29.3%
value mentioned above for calculating the Fe emission from steel plants. Evaluating
the degree of the iron enrichment requires information related to the amount of coal
used in the steel industry, but this information is not available in the regional emission
inventory that we used. So, this study estimated the proportion of coal used in the iron
and steel industry according to the energy consumption data from IEA (for details see
Appendix 3.A). For this calculation we used data for China due to its dominance in the
East Asia steel industry. We found that the iron and steel industry accounted for
approximately 46% of the total coal consumption in the industry sector in 2007.
Additional details required for estimating anthropogenic iron deposition include
the total mass of emissions and the size distribution of fly ash particles. The mass
fraction of iron in fly ash was 4.82% (Table 3.A2), from which the mass of fly ash
emissions was calculated using equation (3.5). Zhang et al. [2005] provided size spectra
of fly ash for Chinese power plants. Their data show that freshly emitted fly ash
particles have a size range exceeding that of ambient aerosols; however, very large
particles fall out quickly and thus are of little consequence to long-range transport. The
upper size limit for fly ash particles that can effectively enter the atmosphere is ~20 μm
[Mcelroy et al., 1982]. Applying this upper limit to the size distribution provided by
Zhang et al. [2005] led us estimate that only 20% of the mass of fly ash calculated from
SO2emissions should be counted as effective emissions.
It was also necessary to determine the size distribution of fly ash particles, which is
generally not provided in emission inventories. In accordance with Luo et al. [2008] and
Ito and Feng [2010], we assumed that the particle size distribution of fly ash has two
modes -- fine mode and coarse mode -- and that each mode can be represented by a
log-normal function [cf. Chen et al., 2013]. We used the parameters for fly ash size
distribution provided by Mcelroy et al. [1982] and set the modal diameter at 0.11 μm
and standard deviation of the size distribution at 1.4 for the fine mode. For the coarse
mode, these parameters were set at 2.2 μm and 1.8. We assigned 1.5% of the fly ash
mass to the fine mode and the remainder to the coarse mode according to that outlined
by Mcelroy et al. [1982]. It should be note that the removal efficiency of fly ash for the
facilities reported by Mcelroy et al. [1982] is 99.7%, similar to the present-day
efficiency. This may be an indication of the maturity of the removal technology and
justify our use of somewhat dated data. The bimodal size distribution constructed using
the above parameters is relatively consistent with those reported by Yi et al. [2006] for
fly ash emissions from power plants in China.
3.2.3 Measurement data
Observational verification of the simulation results is somewhat limited,
particularly in terms of particle composition. This study obtained aerosol measurements
at two stations (locations shown in Figure 3.1), which were suitable for model
verification. The Magong station (23°34’09”N, 119°33’58”E) is on the Pescadores
Islands, in the Taiwan Strait. These islands are relatively remote (approximately 45 km
and 130 km from Taiwan and Mainland China, respectively), sparsely populated
(population just under 100,000), and without major pollution sources. The Magong
station can therefore be regarded as a regional background station. The Wanli station
(25°10’46.8”N, 121°41’23.57”E) is located in a small town (population ~ 22,000) on
the northern coast of Taiwan, where a northerly wind prevails for most of the year
except during summer. The Wanli station is therefore also representative of regional
background situations during the winter monsoon.
Both stations are equipped with standard air pollution instruments operated by the
Environmental Protection Administration of Taiwan (TEPA). This study used data on
the mass concentration of PM smaller than 10 μm (PM10), as measured using tapered
element oscillating microbalance samplers. Near the two TEPA stations are also located
two Taiwan Aerosol Observation Network (TAON) aerosol measurement sites, operated
by the Research Center for Environmental Changes, Academia Sinica [cf. Chou et al.,
2010]. The Cape Fuguei site (25°17’53.8”N, 121°32’11.4”E.) is at the northern tip of
Taiwan, about 20 km northwest of Wanli, and the Xiaomen Islet site (23°39’7.4”N,
119°31’7.1”E) is about 10 km northwest of the Magong site. These sites measure PM10
and PM2.5mass concentrations and chemical composition by collecting samples on
filters. Analysis of chemical composition includes soluble ions and metal elements, as
well as organic and elemental carbon. Fractions of the measured PM10 composition
contributed specifically by mineral dust and fly ash were unavailable. Mineral dust and
fly ash are generally considered insoluble in water; therefore, we can calculate their
mass concentration by subtracting the soluble species (such as sulfate, nitrate, sea salt,
and soluble organic carbon) and other main insoluble species (elementary carbon and
insoluble organic carbon) from the PM10values. Volcanic ash is another common
insoluble aerosol; however, it is probably rare in this area in 2007.
3.3 Simulation Results
We estimated the annual atmospheric input of iron into the ocean based on
conditions in 2007. This year was selected partly due to the availability of emissions
data and concurrent observational data, and partly because no strong climate anomalies
occurred in the region during that year, and it could therefore be considered
representative of recent climate conditions. One noticeable climate anomaly that did
occur in 2007 was a weak-to-moderate La Niña in the latter part of the year. Note that
the strength of the El Niño or La Niña events are defined as sea surface temperature
anomalies (positive or negative) for the region 5qN5qS, 120qW170qW, reaching the
following thresholds for five consecutive months: weak: 0.5 to 0.9, moderate: 1.0 to 1.4,
and strong ≥ 1.5 (see http://www.esrl.noaa.gov/psd/enso/mei/). This was also the first
year in which China’s demand for coal exceeded its domestic supply, according to
statistics presented by the World Coal Association for 2009
(http://www.worldcoal.org/resources/coal-statistics/). In the same year, China’s SO2
emissions peaked with declines in the following years occurring mainly in the energy
sector, rather than the industry sector [Klimont et al., 2013].
3.3.1. Meteorological aspects
The transport of aerosols from the East Asia continent to the marginal seas is
largely controlled by the Asian monsoon system which exhibits significant seasonal
variations. Northerly winds, which generally originate in populated lands, dominate the
marginal sea regions in autumn and winter; while southerly winds, which originate
mostly from the oceans and thus tend to be cleaner, prevail in summer. The precipitation
fields show a SSE to NNW (i.e., open sea to inland) gradient in all seasons, with
maximums and minimums occurring in summer and winter, respectively. Winds and
precipitation in the simulations were compared with the NCEP Reanalysis II data
(representing observations). From a statistical analysis on the 5-day average data over
the entire domain, the space and time correlation coefficient reached 0.89 for winds and
0.78 for precipitation. These high correlations indicate good model performance. Model
bias was shown to be relatively small in the case of winds but more significant with
regard to precipitation. Precipitation bias was the highest in spring (+24%) and lowest
in winter (-2%), and the annual mean was overestimated by 11%. More details of the
meteorological aspects are provided in Appendix 3.C.
3.3.2. Simulated aerosol emission and concentration
Emissions and transport of mineral dust and fly ash were simulated using the
CMAQ model, according to meteorological conditions generated from MM5 (Figure
3.4). As discussed earlier, we used iron mass fractions of 4.00% and 4.82% to convert
dust and ash emissions to iron emissions. Most (89%) of the dust iron in East Asia
originates in the desert areas of northern China and Mongolia, with the remainder (11%)
from nondesert areas including agricultural lands. Total iron emissions in this region are
26.0 Tg yr–1from mineral dust and 7.2 Tg yr–1from fly ash. Industrial coal-burning
accounts for 64% of total fly ash iron emissions, while residential use and power plants
contribute 27% and 9%, respectively. The large fraction originating from residential
sources appears anomalous, as only a minor fraction of total coal consumption is for
residential use (less than 14%, according to IEA 2007 statistics). This anomaly reflects
the fact that recovery devices are seldom used in the residential coal burning.
The emission and transport of mineral dust are controlled primarily by
meteorological factors. The East Asian monsoon system, in particular, determines the
emission strength and low-altitude transport, whereas the high-altitude westerlies
facilitate long-range transport in the free troposphere. Our goal was to evaluate the
deposition of atmospheric nutrient into the oceans, so we focused on the NWPO and
NSCS and excluded iron deposition on land. Figure 3.5 presents simulations of seasonal
variations in the near-surface mass concentration of mineral dust (all sizes) in spring
(MarchMay), summer (JuneAugust), autumn (SeptemberNovember), and winter
(DecemberFebruary). Due to their short lifetimes, aerosols tend not to move very far
from their geographically fixed sources. As a result, the general patterns in the four
seasons seem to be similar at a first glance. More obvious seasonal differences tend to
be observed when focusing on more distant areas, such as over open oceans. Over the
desert regions, airborne dust concentrations are generally greater in spring than in winter,
partly due to stronger synoptic activity and the reduced snow cover. Over the NWPO,
particularly the ECS, the highest dust concentrations occurred during spring. Over the
particularly the ECS, the highest dust concentrations occurred during spring. Over the