Monte carlo simulation of regional aerosol: transport and kinetics

Chemical Transformation and Removal Kinetics

In the Monte Carlo model, a semi-empirical approach is employed for the simulation of pollutant transformation and dry and wet deposition. The approach is deterministic in that rate equations are used for the calculation of kinetic processes applied to each quantum. At a quantum’s release, it is assigned a species mass weight based upon the source's emission rate. Mass conservation is maintained at the quantum level by accounting for the mass that has been transformed and removed. The kinetic rate coefficients are dependent upon the chemical, meteorological, and geological environment of the quanta. Consequently, these coefficients vary with space and time creating variable transformation and removal rates.

The simulation of the kinetic processes is empirical, in the sense that the rate coefficient relationships are determined through a fitting process matching simulated to observed concentration and deposition fields. The main thrust of this approach is to make maximum use of existing meteorological, emission, receptor, etc. data in conjunction with a physical-chemical model to derive the best set of rate coefficient equations. This approach assumes that the correct pollutant transport and emission rates are known. It is hypothesized that given sufficient quality and quantity of data, it would be possible to identify a unique set of rate parameters, thereby simulating the physical/chemical processes.

- Simulation Over the Eastern US. Recently, a project was undertaken for the retuning of the Monte Carlo model for the simulation of the - system over the Eastern US during 1992. The following will present the resulting rate coefficient equations and some of the resultant simulated concentrations and wet deposition rates. The process by which the rate coefficient relationships were developed and their justification is given in Schichtel (32).

The rate of change of a quantum’s mass can be expressed as:

(2)

where , , and are the oxidation, dry deposition, and wet deposition rate coefficients of respectively. Similarly, the rate of change of the mass can be expressed as:

(3)

where and are the dry and wet deposition rate coefficients of respectively. It was assumed that all and mass arose from the simulated sources only, thus, background and were neglected.

The tuning process was conducted using the meteorological data generated from the National Weather Service’s Nested Grid Model (21, 22). The data were available on a stereographic grid with approximately 180 km size. Using the NGM data, a Lagrangian database was created by releasing particles from the center of each meteorological grid cell, east of Colorado. Approximately 350 sources were simulated in this manner. The three dimensional location of each particle, along with the specific humidity and temperature, were deposited in the Lagrangian database every two hours. Each particle was tracked for seven days, or until it was transported past the edge of the meteorological grid.

The NGM data contained a precipitation field. However, it was found that the precipitation field did not compare well with measured data. Deviations in the seasonal precipitation rates up to a factor of two were found depending on the season and location (32). The NGM precipitation field was replaced by hourly measured precipitation data from the US Control COOP Hourly Precipitation database (33). This was accomplished by averaging the precipitation rates over a two hour period, and spatially interpolating the data to the same grid used by the NGM meteorological data. The resulting database was then suitable for input into the kinetic module. A similar process was used for the creation of gridded total cloud cover data from hourly data in the National Solar Radiation Database (34).

Emissions were obtained from the 1985 NAPAP inventory (35). The original data were provided as point and area emissions on a 20 km grid. The data were transformed to the meteorological grid by summing all point and area emissions together for all emission grid cells which fell into the same meteorological grid cell. No attempt had been made to account for variations in 1992 emission from those in 1985.

The observed data utilized for the tuning process consisted of ambient fine sulfur aerosol concentrations from the IMPROVE (36) and the NESCAUM (37) monitoring networks, airport visibility data (38), ambient concentrations from the AIRS database (39), and total wet deposition data from the National Atmospheric Deposition Program/National Trends Network (40).

The final relations derived for the rate coefficients are:

= 0.001 + SR (0.02 + 0.05 SH/SHmax + 0.1*PF) (4)

SR = (1-.65TS)GSR (5)

= V_d / H (6)

V_d = .22 + 2 * SR (7)

= /10 (8)

= W_s2/H * P = Sqr(40000 / (SO₂CC + 2)) * P (9)

= W_s4/H * P = 250 * P (10)
where SR = ground level solar radiation [kW/m2]
TS = total cloud cover
GSR = ground level solar radiation in the absence of clouds [kW/m²]
SH = specific humidity [kg Water/kg Air]
PF = precipitation flag, i.e. PF = 1 if it rains PF = 0 otherwise
V_d = deposition velocity [cm/s]
H = mixing height [m]
W_s2, W_s4 = washout ratios for and respectively
SO₂CC = SO₂ column concentration [mg/m²]
P = precipitation rate [m/hr]
To illustrate how the meteorological variables affect the kinetic rate coefficients, a quantum was released from St. Louis on July 15 at 10:00 GMT, and tracked for four days. Figure 6 presents the quantum’s trajectory, and six of the airmass history variables necessary for the calculation of the rate coefficients. The resulting rate coefficients and a corresponding sulfur budget are presented in Figure 7. As shown, the quantum was released, and remained, below the mixing height for the initial fourteen hours. Thus, it was exposed to the ground and a dry depositions rate of approximately 5%/hr occurred. This resulted in about 38% of the to be dry deposited. During this time, the solar input cycled from 0 to 0.6 back to 0 kW/m² producing transformation rates from 0.1 %/hr to 3 %/hr. This caused approximately 12% of the to be transformed to . With the end of the solar input, the mixing layer collapsed and the quanta was above the mixing height. These conditions resulted in no deposition of the and little transformation. During the beginning of the second day, the ground level solar flux increased and the quantum encountered a light precipitating airmass. The precipitation caused total removal rates of the to be as high as 12%/hr removing 60% of the remaining mass over the next 12 hours. The precipitation also washed out a significant fraction of the . However, the enhanced transformation rates during precipitation, caused the overall mass to increase to 20% of the total sulfur. After this period the precipitation increased, and after two days of travel, nearly all of the and mass of the quantum had been dry and wet deposited.

Simulated concentrations. Using the above rate equations, the Monte Carlo model was run for all of 1992, generating 24-hour concentration and deposition fields over the Eastern US. The surface concentrations were calculated for receptor volumes defined by the meteorological grid cell and the mixing height. Figure 8, presents the average simulated concentration fields during the periods January - March, quarter 1, and July -September, quarter 3. Also, plotted for comparison are “observed” isopleth maps generated from the fusion of IMPROVE and NESCAUM data and airport visibility data (41). During quarter 1, the observed is nearly uniform over the east with a concentration of about 3 g/m³. The simulated concentrations have more spatial variability then the observed, but reproduce the general spatial pattern and magnitudes. The simulation deviates from the observed the greatest in a belt from eastern Ohio to the Chesapeake Bay, with simulated concentrations greater than 5 g/m³ compared to about 3 g/m³ for the observed .

The simulation during quarter 3, also reproduces the general spatial pattern and magnitudes of the observations. The highest observed concentrations are from southern Pennsylvania and north Virginia to western Kentucky and Tennessee ranging from 6 to greater than 7 g/m³. The simulated has a similar high concentration band, however, it is pushed slightly to the north and west of the observed band. The concentrations for both the observed and simulated progressively decrease in all direction from the high concentration band producing similar concentrations over New England (~ 3 g/m³), the South (~5 g/m³), and into the Mid west (2 - 5 g/m³).

The daily averaged simulated are seasonally compared to observations in correlation plots (Figure 10). Due to the coarse meteorological grid used, it is unreasonable to expect the model to simulate concentrations at a single receptor. Therefore, these comparisons were conducted by averaging the model results and the NESCAUM/IMPROVE observations over a New England and East Central spatial domains as defined in Figure 9. The simulated in New England compares well with the observations, with squared correlation coefficients varying between 0.56 during quarter 3 and 0.83 during quarter 2. The data points are centered around the one to one line, except for quarter 1 where the model simulation tends to underestimate the observed concentrations.

The correlation plots in the East Central domain have squared correlation coefficients varying between 0.35 during quarter 1 to 0.61 during quarter 3 with r² = 0.59 for the entire year. The simulation reproduced the average observed concentrations for quarters 1 and 4, but tended to underestimate them the rest of the year.

Simulated Total Wet Deposition Rates. The average simulated and observed total wet deposition rates for quarters 1 and 3 are presented in Figure 11. The observed deposition rates during quarter 1 range between 0.6 and 1.8 g/m²/yr with the largest depositions occurring in a band from New York around the Carolinas and into the deep South. The simulation reproduces this pattern, but tends to underestimate the deposition rates along the coast and at the model domain boundaries. For example, the simulated deposition rates in Florida are about 0.3 g/m²/yr compared to 1 g/m²/yr for the observations.

During quarter 3, the model results properly simulate the high deposition rates south of Lake Erie and Ontario, where they exceed 4.2 g/m²/yr, and the decreasing deposition gradient extending from this region. However, like quarter 1, the simulation underestimates the deposition rates along the east coast, Florida, and the western part of the model domain between a factor of 1.5 and 2.

Comparisons of the simulated total wet deposition rates to observed rates in New England and East Central for weekly data are presented in Figure 12. This figure was created using the spatial domains defined in Figure 9. The seasonal squared correlation’s for New England show high correlation ranging from 0.5 to 0.94 with an overall squared correlation of 0.79 for the year. The simulated deposition rates underestimated the observed by about 25% over the year, 1.71 g/m2/yr compared to 2.15 g/m²/yr for the observed rates. In the East Central domain, the seasonal squared correlation varied between 0.71 and 0.89 with a yearly correlation of 0.83. The yearly average deposition rates were 2 and 2.2 g/m²/yr for the simulated and observed data respectively.