EPA-600/4-76-031
June 1976
Environmental Monitoring Series
PRECIPITATION SCAVENGING OF
FOSSIL-FUEL EFFLUENTS
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and application of
environmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the ENVIRONMENTAL MONITORING series.
This series describes research conducted to develop new or improved methods
and instrumentation for the identification and quantification of environmental
pollutants at the lowest conceivably significant concentrations. It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors.
This document is available to the public through the National Technical Informa-
tion Service. Springfield, Virginia 22161.
-------
EPA-600/4-76-031
June 1976
PRECIPITATION SCAVENGING OF FOSSIL-FUEL
EFFLUENTS
By
M. Terry Dana, Dennis R. Drewes,
Donald W. Glover, and Jeremy M. Hales
Battelle, Pacific Northwest Laboratories
Richland, Washington 99352
Contract No. 68-02-1729
Project Officer
Herbert J. Viebrock
Meteorology and Assessment Division
Environmental Sciences Research Laboratory
Research Triangle Park, NC 27711
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
RESEARCH TRIANGLE PARK, NC 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research Labora-
tory, U.S. Environmental Protection Agency, and approved for publication.
Approval does not signify that the contents necessarily reflect the views
and policies of the U.S. Environmental Protection Agency, nor does mention
of trade names or commercial products constitute endorsement or recommenda-
tion for use.
ii
-------
ABSTRACT
The research program resulted in the development of a numerical model for
predicting precipitation scavenging of reactive pollutants from power-plant
plumes. Named SMICK (Scavenging Model Incorporating Chemical Kinetics), this
model calculates collection, liquid-phase chemical reaction, and (possible)
desorption of multiple plume-bound pollutants as they interact with falling
raindrops and are ultimately deposited at the ground. Calculations for any
specific aqueous-phase kinetics mechanism can be performed with SMICK by ex-
pressing the mechanism in appropriate subroutine form.
The program involved field experiments at the Centralia Steam-Electric Plant
in southwestern Washington State. These show that: 1) sulfate concentrations
in rainwater due to the plume's presence do not appear to be significant at
downwind distances less than about 10 km; 2) ammonium, nitrate, and soluble
(ortho) phosphate ion concentrations were at or near normal background lev-
els; and 3) the only species measured which showed plume-related depostion
patterns at all distances where samples were collected (0.4-11 km) were SO
and hydrogen ion.
Initial applications of the SMICK model included a previously-suggested
aqueous-phase oxidation mechanism for SO , which depends strongly on the
presence of dissolved ammonia. Comparisons of these predicted results with
the field measurements indicate that, for aqueous-phase ammonium concentra-
tions in the measured range, this mechanism is not adequate to account for
observed rates of sulfate formation.
111
-------
IV
-------
CONTENTS
Page
ABSTRACT iii
LIST OF FIGURES vi
LIST OF TABLES vii
ACKNOWLEDGEMENTS ix
SECTIONS
I. CONCLUSIONS 1
II. RECOMMENDATIONS 2
III. INTRODUCTION 3
IV. MODELING OF SCAVENGING PHENOMENA 7
V. EXPERIMENTAL DESIGN 15
VI. EXPERIMENTAL RESULTS 27
VII. COMPARISON OF SMICK CALCULATIONS WITH OBSERVED DATA 57
VIII. REFERENCES 65
IX. NOMENCLATURE 68
X. APPENDICES 70
-------
FIGURES
Page
1. Schematic of Physical Interactions Modeled by SMICK 13
2. The Centralia Steam-Electric Plant 17
3. The Centralia Precipitation Sampling Array 18
4. Precipitation Sampling Field Equipment 21
5. Rainfall Amount vs Time Measured at the Mobile Laboratory: Run Fl. 30
6. Rainfall Amount vs Time Measured at the Mobile Laboratory: Run F4. 30
7. Rainfall Amount vs Time Measured at the Mobile Laboratory: Run F3. 31
8. Ground-level Rainwater Concentrations Map for S02, Run Fl 36
9. Ground-level Rainwater Concentrations Map for S02, Run F3
Boxes Show Air Concentrations in ppb 37
10. Ground-level Rainwater Concentrations Map for S0~, Run F4
Boxes Show Air Concentrations in ppb 38
11. Measured Scavenging Rates for SOo (cf. (5)) as a Function of Down-
wind Distance 41
12. Ground-level Rainwater Concentrations Map for Sulfate, Run F3. . . 44
13. Ground-level Rainwater Concentrations Map for Sulfate, Run F4. . . 45
14. Ground-level Rainwater Concentrations Map for H , Run F3 46
15. Ground-level Rainwater Concentrations Map for H , Run F4 47
16. Cross-Plume Rainwater Concentrations Profiles in Tenino Area
(x^lO km) 49
17. Scanning Electron Photomicrograph of a Particle From Mobile Rain
Sample 71B_ The Diameter of the Particle is about 4.9 ym 55
18. Scanning Electron Photomicrograph of Particles from Mobile Rain
Sample 82C. The Diameter of the Spherical Particle in Right
Center is about 5.5 pm 55
VI
-------
TABLES
Page
1. Sampling Areas 20
2. Chemical Analysis Methods for Pollutants Measured and Nominal Low
Detection Limits 26
3. Plant Operation and Emission Data 28
4. Fixed Site Collection Times 29
5. Rainfall Rates Summary 29
6. Raindrop Size Frequency Distributions ... 33
7- Wind and Temperature Data .33
8. Scavenging Rates for SO,,: Comparison of Present Results with
Those of the Previous Studies ...... 40
9. Meteorological and Source Conditions Relevant to Scavenging Rate
Comparisons 42
10. Measured Air Concentrations of NH» and 0~ 51
11. Measured Air Concentrations of SO,., 52
12. Results of Electron Microscope Analyses of Filtered Rain Samples. . 54
13. SMICK Calculations with Scott and Hobbs Mechanism: Uniform Gas
Phase S0« Concentration ..... 59
14. Calculated Plume Centerline Rainwater Concentrations: SMICK (Scott
and Hobbs; K=0.0017 sec"1) and EPAEC (S02 only) 60
15. Sample SMICK Plume Centerline Rain Concentration Calculations with
Scott and Hobbs Mechanism: Run F3 Input Data P_ =311 ppm 62
co2
16. Cross-Plume SMICK Rain Concentration Calculations: Run F3S Tenino
(xVLO km) Area, K=0.0017 sec"1; ?_„, =1 ppb; P-- =311 ppm 64
Nn,, CD-
17- Rain Sample Concentration Data: Run Fl, December 6, 1974 71
18. Rain Sample Concentration Data: Run F3, December 10, 1974 72
19. Rain Sample Concentration Data: Run F4, December 12, 1974 73
20. Rain Sample Concentration Data: Run F5, December 15, 1974 74
21. Concentration Results-Mobile Samples (and Associated Fixed Sites):
Run F3, December 10, 1974 75
22. Concentration Results-Mobile Samples (and Associated Fixed Sites):
Run F4, December 12, 1974 76
23. Concentration Results-Mobile Samples: Convective Shower,
December 12, 1974 77
VI1
-------
Page
24. Values of the Equilibrium Constants (Equations 6b-14b) used in
SMICK calculations with the Scott and Hobbs Mechanism (T=298K). . .81
25. Input Data for Sample Calculations 94
26. Sample Calculations Showing AZ Variation.. 94
Vlll
-------
ACKNOWLEDGEMENTS
We are grateful for the cooperation and assistance of personnel of Pacific
Power and Light Company, operators of the Centralia Steam-Electric Plant.
In particular, Bob Davis and Gordon Strom of that company provided coordina-
tion which made our field operations on the plant property a success.
We appreciate the kind help of Professor Donald Adams of Washington State
University, who provided the wind and temperature data from the meteorology
tower.
Battelle-Northwest personnel who contributed significantly to the overall
project include Sam Henderson, R. E. Kerns, and J. W. Sloot.
IX
-------
SECTION 1
CONCLUSIONS
A predictive model for the scavenging of fossil-fuel effluents was
developed. The initial application of this computer model has been to the
sulfur compounds — SO and oxidized ionic species, particularly sulfate.
The development of the scavenging model incorporating chemical kinetics
(SMICK) came about by extension of the existing SO2~only model (EPAEC)
to include other species, and chemical reactions among them occurring with-
in raindrops during their trajectories through a plume. Comparisons of the
calculated results with experimental data, when a liquid phase oxidation
mechanism involving ammonia is employed, indicate that this mechanism is
insufficient to explain observed sulfate formation rates in the plume-and-
raindrops situations tested.
The data from the field experiments at the Centralia Steam-Electric Plant,
the first taken there during normal operation of the coal-fired facility,
show that: 1) sulfate concentrations in rainwater due to the plume's
presence do not appear to be significant at downwind distances less than
about 10 km; 2) ammonium, nitrite, nitrate, and soluble (ortho) phosphate
ion concentrations were at or near normal background levels; and 3) the
only species measured which showed plume-related deposition patterns at all
distances where samples were collected (0.4-11 km) were S0« and hydrogen ion.
The field program employed, in addition to fixed site collectors, short-time
sampling with mobile collectors. Some of the large-volume rain samples were
filtered and subsequently examined for insoluble particle content by scanning
electron microscopy (SEM). The mobile sampling can be useful for studying
short-time rainwater concentration fluctuations, and the SEM analysis shows
great promise as a means by which fossil-fuel effluent particles may be
identified and counted, and their scavenging properties determined.
-------
SECTION II
RECOMMENDATIONS
The modeling of scavenging of, reactive fossil-fuel plumes should be continued.
The basic framework of SMICK is amenable to the testing of many liquid-phase
reaction mechanisms, and — with the addition of an existing reactive plume
model — the role of gas-phase reactions can be examined as well. The major
effort of the current contract was concerned with development of the com-
putational algorithm; various scavenging and reaction mechanisms can be tested
without major modifications.
The findings of this phase of the research program demonstrate that there is
a strong need for experimental measurements of ammonia solubility in water at
low concentrations. Because of the evidence of marked deviations between
observations and calculated ammonia solubility, and because of the intrinsic
importance to several proposed mechanisms of SO- oxidation, these measurements
should be performed prior to any further speculation regarding the role of
ammonia in SO oxidation processes.
There is a need to examine the composition and scavenging efficiencies of
small insoluble effluent particles which escape the electrostatic precipita-
tors of fossil-fueled facilities. This can be accomplished through scanning
electron microscope analysis of filtered rain samples; in principle, the
rainwater content and particle size distributions can be determined at various
downwind travel times. These results can be compared with measured or infer-
red source strengths and particle size spectra to arrive at an assessment of
the scavenging rates.
Large-area mobile precipitation samples can be used in future research to
provide a reasonably good measure of the deposition patterns of many soluble
effluents. These also can be used to examine short-term concentration
fluctuations, and can provide the large water volumes needed for sensitive
insoluble particle analysis.
-------
SECTION III
INTRODUCTION
The atmospheric effluents which result from the use of fossil fuels gen-
erally consist of the gaseous products of ordinary combustion of organic
matter — CO, C02, NO, N02, HO, SO — and the more complicated and less
prevalent combustion products consisting of small amounts of various contamin-
ants. These contaminants — defined as not being a necessary aspect of
the fuel — are generally inorganic compounds and elements found in varying
amounts in the earth's crust, such as oxides of silicon, iron, aluminum,
and trace quantities of heavy metals. The relative abundances of the var-
ious pollutant species in the atmospheric effluent stream depend on the
type of fuel used and the means and efficiency of combustion and control
measures.
The atmosphere and biosphere have been capable of coping with large natural
occurrences of the release of typical fossil-fuel effluents. For example,
the natural emission of sulfur dioxide from volcanoes in normal times has
12 1
been estimates at about 10 grams annually. The increasing use of
fossil fuels by humanity throughout the ages has, however, brought us to
the point where the anthropogenic contribution of S0_ to the atmosphere is
2
of the same magnitude. The difficult question of when (if not already)
these and like emissions begin to affect health and quality of life must be
faced, in order that the overall impact of human activities may be recogniz-
ed in a timely way, and that alternatives to present developments may be
effectively considered.
The toxicities of the effluent gases (with the exception of HO) have been
well demonstrated; the task of removing them is a difficult one, however,
and the environmental question is basically one of assessing the damage
which comes about through their release, and then attaching limits to the
releases. Of course, the assessments direct the development of control
-------
mechanisms where necessary. The particulate etfluents are most
easily controlled; in fact, a large fraction of the mass of fly ash
particles resulting from most types of fossil-fuel combustion can now be
removed by electrostatic precipitation. Nevertheless, the effluent streams
contain large nimbers of very small particles, whose effect on the environ-
ment may be felt in a manner more indirect than that of the gaseous oxides.
The interactions between pollution and the atmosphere have been researched
extensively, particularly in the areas of transport and diffusion, dry
deposition, and precipitation scavenging. Each of these classes of inter-
actions has generally been treated independently of the others. Of late,
however, the overall fate of pollutants has been of concern; that is,
answers are being sought to the question of whether pollutants change
chemically and/or physically in time, and under what conditions. The
changes can profoundly affect the rates of removal from the atmosphere.
Probably because experimental measurements are easiest in fair weather,
the fate of certain pollutants is better understood for non-precipitating
conditions. For example, S00 conversion rate estimates and information
about the coagulation of particles have been derived from aircraft measure-
ments under fair weather conditions only.
The chemical conversions which may occur among fossil-fuel pollutant plume
constituents under precipitation conditions are probably more complicated;
in addition, experimental measurements are generally confined to the ground
level. There is evidence that the molar concentrations of sulfate in rain-
fall at ground level are comparable to those of SCL, even at relatively
3
short distances down wind of a coal-fired source. The implied rates of SO
conversion — if it is a gas-phase process — are much faster than those
proposed for fair-weather conditions. The possibilities are then open
for, among others, 1) an accelerated gas-phase reaction process, 2) a
liquid phase reaction process in precipitation, 3) a liquid phase
process in cloud droplets and subsequent efficient scavenging of cloud
drops by precipitation, or 4) combinations of the above.
-------
It is the objective of the present research to model precipitation scaveng-
ing of fossil-fuel effluents, accounting for chemical conversions as well as
the better known physical processes. The primary interest is with the SO
- sulfate reaction process, for the following reasons.
1) The fate of SO- is of particular concern because it is the
currently recognized major atmospheric effluent of fossil-
fuel combustion in terms of environmental impact.
2) Our experience with the modeling of scavenging of gases --
and particularly S0~ — makes it natural to begin a modeling
program including chemical reactions with the pollutant
whose scavenging properties are otherwise best understood.
In addition, the framework of the developed gas scaveng-
ing model is amenable to generalization to several species
with reactions occurring among them.
3) A number of atmospheric conversion processes have been pro-
posed which concern themselves with the S0~-sulfate problem.
It is important that means be developed whereby these processes
can be tested under conditions which exist in real fossil-fuel
plume and precipitation situations.
Furthermore, the atmospheric and plume conditions which promote chemical
conversions in precipitation conditions are not well known. That is, all
currently proposed mechanisms which involve catalysts by some plume or
natural constituent may prove inadequate to explain observed precipitation
concentrations of the materials of interest. Therefore, it is considered
an important objective of the study to conduct field measurements concerned
with a number of pollutants, some of which may be important as catalysts, and
some of which may be of significant environmental concern in their own
right.
This report documents the year's work, which consists of the development
of a model for the scavenging of chemically reacting plume-borne efflu-
ents and the conduct of a field measurement study of precipitation
-------
deposition of a number of pollutants from a large coal-fired power plant.
The report is generally in three parts, describing: the development of
the model (Section IV) , the methods employed in and preliminary results of
the field study (Sections V and VI), and some initial applications of the
model to real experimental situations from the current field study and
past ones (Section VII).
-------
SECTION IV
MODELING OF SCAVENGING PHENOMENA
During the contract period the mathematical modeling of scavenging phenomena
has progressed in two basic areas: improvements in the gas-scavenging code,
Q /
EPAEC, developed under a previous contract, and the formulation of a revised
code, SMICK, to calculate scavenging of multiple, reacting plume constitu-
ents. Progress in each of these areas is summarized individually in the
following paragraphs.
REVISIONS OF EPAEC
The computer code EPAEC (described briefly in Appendix C, and in detail
elsewhere"') calculates scavenging of n.aterial from a Gaussian plume
by raindrops falling along slanted trajectories; the drop-pollutant
interaction is considered to be purely a physical (or pseudophysical)
process, where no irreversible chemical reactions occur within the aqueous
phase. Although a first-order chemical reaction within the gaseous phase is
possible, modifications of the plume by the rain are assumed to be negligible.
Plume loft is incorporated into the model in rudimentary fashion only,
providing for either a linear loft rate, or a level plume established at a
predetermined virtual release height.
EPAEC was found to perform well for the conditions experienced in the pre-
3
vious study. Later applications under the present contract, however, showed
the numerical algorithm to be unstable under some circumstances, which often
involved high, confined plumes containing gases of relatively low solubility.
Because of this difficulty, a theoretical stability analysis of the basic
Runge-Kutta algorithm employed in the code was performed. Noting that the
Runge-Kutta algorithm is employed in EPAEC to obtain finite-difference
approximations (FDA's) to solutions of the basic drop-response equation
3K
-------
it was found that the algorithm should, indeed, tend toward instability
under the following conditions:
1 + Az
< 1
(2)
The nomenclature in the above forms is given as follows:
a = raindrop radius, H
c = concentration of pollutant in raindrop, moles/H
H' = solubility parameter, £3/mole
K = mass-transfer coefficient, moles/A t
v = terminal velocity of fall of raindrop, H/t
z
y., = mole fraction of pollutant A in local gas phase
Ab
z = distance above ground at stack base, £
Az = spacing of computation grid, £.
Inequality (2) indicates that, at set values of Az, numerical instability
should be expected for low solubilities (high values of H'), small rain-
drops, and high mass-transfer coefficients. Since the step size Az is
easily adjustable using the Runge-Kutta algorithm, a simple solution to this
instability problem is to provide for an automatic adjustment of the step
size within EPAEC to satisfy criterion (2). This measure, however, was
found to be only partially satisfactory. While successfully stabilizing
the algorithm, the modification often reduced Az to such small values that
execution times became prohibitively large. In view of this difficulty,
alternative solutions to the instability problem, involving other finite
differencing techniques, were explored. These included:
1) application of predictor-corrector methods,
2) application of successive perturbation techniques,
-------
3) use of second-order single-step algorithm suggested by
Liniger and Willoughby ,
and 4) application of the exponential method of Sherwood and Herman .
Techniques 2-4 were rejected, for the present, because they either
necessitated numerical differentiation of the forcing function y., , or were
AD
judged too clumsy for anticipated multicomponent applications in the SMICK
code, which was expected to employ a similar algorithm. Predictor-
corrector methods possess neither of these difficulties; on the other hand,
they do tend to become unstable under specific conditions. These condi-
tions are not so severe as those posed by the Runge-Kutta method; in
fact, many predictor-corrector methods possess the additional advantage of
being substantially more efficient than their Runge-Kutta counterparts.
Moreover, predictor-corrector methods provide a more direct estimate of
numerical deviations from the true solutions, allowing a relative econo-
mization of any step-size modification imposed for stability maintenance.
Because of the above features, a predictor-corrector algorithm was formu-
lated for use in a revised version of EPAEC. The technique chosen was a
modified version pf Hamming's method, which possesses a number of desirable
features, including a fourth-order truncation error and relative absence of
parasitic solution effects. Experience with this revised code to date has
indicated that the previous problems have been countered successfully under
most conditions, although a moderate reduction in Az is still necessary to
maintain stability under some circumstances. In addition to including a
provision for reducing Az, the new algorithm also increases Az where
conditions permit; this often results in a shorter computation time than
is required by the old method. Examples of Az modifications during inte-
gration are shown in Appendix D
In addition to incorporating the more sophisticated predictor-corrector
algorithm, EPAEC was also revised to account for plume loft in a more satis-
Q
factory fashion. The revised formulae of Briggs , used in the earlier
version of the code, are employed in a more exact manner to allow a "bent-
over" plume, rather than the more simplistic concepts employed by the
-------
earlier model. Conceptually, this was not a difficult modification.
Because of its effect on the relationships between the plume position and
raindrop trajectories, and the associated structure of the finite difference
grid locations, however, this modification has resulted in a substantial
restructuring of the computer program listing.
MULTICOMPONENT SCAVENGING FROM PLUMES: THE FORMULATION OF SMICK
As indicated previously, EPAEC focuses upon the scavenging of a single plume
constituent only. For an advanced analysis of pollutant scavenging in
power plant plumes it is desirable to possess an extended diagnostic model
which provides for the wet removal of multiple reactive components; the
computer code SMICK (Scavenging Model Incorporating Chemical Kinetics) has
been formulated to fulfill this requirement.
The basic properties of SMICK can be presented best by considering the basic
steady-state equations of continuity for the two phase (rain-air) system of
9
interest :
- V • p. v. + w, + r, = 0 (aqueous phase) (3)
Ax Ax A Ax
and
PA VA ~ w. + rA = 0 (gaseous phase), (4)
Av Ay A Ay
where
A = component index = 1,2,...,N
v , v = velocity vectors for component A in rain and air, £/t
Ax' Ay
r , r. = reaction rates for component A in rain and air,
Ax' Ay
moles/£3t
w. = interphase transport rate of component A, moles/£3t
p = molar concentration of component A in space moles/£3
A
10
-------
Both SMICK and EPAEC assume that w is negligibly small in (4), and
both require an externally-supplied solution of this equation* in order
to solve (3) for aqueous-phase concentrations. In contrast to EPAEC, which
assumes r = 0 and considers only a single component (N=l), SMICK allows
AX
for multiple components (N ^ 1)** and provides for arbitrary expressions to
describe aqueous-phase reaction behavior.
The simultaneous mass-transfer and chemical-reaction process characterizing
a particular microphysical model of interest in incorporated into the over-
all code as follows: first the mechanism is separated into its individual
steps, and is expressed in terms of simultaneous equations, which may be
algebraic expressions, first-order ordinary differential equations, or both.
(Note that (1), employed in EPAEC, is such an equation but the relative
simplicity of this code requires only one such expression.) Upon formu-
lating the defining equations, they are coded in a FORTRAN subroutine which
provides the appropriate microphysical parameters when interrogated by the
calling program, SMICK. With this arrangement, virtually any liquid-phase
reaction mechanism can be employed with the code in a relatively simple and
straight-forward manner.
The multiple interactions defined in the above fashion are treated by
solving simultaneously the corresponding multiple differential equations
using the predictor-corrector algorithm in a manner similar to that employed
for the single equation in EPAEC. All of the improvements described earlier
for EPAEC have also been adapted into SMICK, and one should note that SMICK
can be executed to produce the same results as EPAEC, if it is constrained
to the special case, N=l.
* This externally-supplied solution is presently the Pasquill-Gifford
bivariate-normal plume equation in both SMICK and EPAEC. This can be
modified easily, however, by simple exchange of the subroutines providing
the solutions.
**The present code limits N to 10 components. This number can be extended
easily by suitable replacement of DIMENSION statements, however.
11
-------
The physical interactions modeled by SMICK are presented schematically in
Figure 1. As indicated by this diagram, the code allows for the capture of
both gases and aerosols by the raindrops, in which the material may react,
desorb, or deposit on the ground.
It should be emphasized that the high versatility in describing liquid-phase
reaction phenomena in SMICK is not at the present time matched by a similar
capability for components in the gas phase. Since the plume model incor-
porated by this code is expressed in subroutine form, it may be exchanged
easily with one incorporating more sophisticated gas-phase reaction
kinetics. Such a measure, however, would result in intolerable increases in
execution time, and is judged to be unwarranted for this project. A more
viable alternative is to obtain computer solutions to an appropriate
reactive plume model, and store these solutions in a three-dimensional
matrix, corresponding to the spacially-varying concentration field. The
SMICK code can then be executed in its usual manner, after modifying the
plume subroutine to a three-dimensional interpolation, which calculates
plume concentrations at any point from the base data in the matrix. A
three-dimensional interpolation algorithm has been formulated to accomplish
the above function, and two reactive plume codes - formulated under other
programs - are ready for use in this capacity. The first of these is
PROTEUS, a moderately rigorous and versatile K-theory model; the second is
STRAC, a code based upon the simplified concept of a quasi-Gaussian plume,
which possesses the advantage of comparatively rapid execution times. At
the present time neither STRAC nor PROTEUS have been operated with SMICK in
this capacity.
Application of SMICK for Scavenging Analysis
The intention of this program is to apply SMICK as an aid in validating or
rejecting postulated mechanisms of microphysical scavenging behavior. In
accomplishing this, the particular mechanism of interest is expressed in
subroutine form and incorporated with the code as described earlier. The
code is then executed for conditions corresponding to field experiments
performed under this program, and pollutant concentrations in rain predicted
12
-------
Volatile
Material
in Raindrop
c
o
•H
J-J
a
CO
a)
Nonvolatile
Material in
Raindrop
FIGURE 1. SCHEMATIC OF PHYSICAL INTERACTIONS MODELED BY SMICK.
-------
by the code are compared with those actually measured. In such an appli-
cation it should be recognized that, while disagreement between observed
and predicted values provides a strong basis for rejection of a given
mechanism, the converse is untrue. This is simply because the complex
conditions leading to wet deposition of some materials allow the possi-
bility that the same final circumstances might be predicted by a variety
of diverse microphysical mechanisms. While this feature does not detract
totally from the above approach to mechanistic examination, one must be
cognizant of its existence in applying this method for analysis of the
scavenging process.
Several mechanisms for SO,., conversion and scavenging exist in the liter-
ature; it is the intent of this program to proceed systematically through
these mechanisms in the above manner in an attempt to ellucidate their
applicability. At the present only one mechanism, that of Scott and Hobbs .
has been examined in this fashion. A description of the application of
SMICK and this oxidation mechanism to experimental situations is included in
Section VII.
14
-------
SECTION V.
EXPERIMENTAL DESIGN
A field study was conducted to provide a data base of pollutant meas-
urements for use in the modeling analysis. For present purposes
these are primarily the sulfur compounds but it was desirous to
examine the precipitation (and some gas-phase) concentrations of a number
of inorganic pollutants, some of which are potentially important in the
sulfur oxidation process. In addition, the pollutants N0? , NO., , and
3-
PO, are candidates for possible future scavenging modeling. Initial
experimental examinations of insoluble pollutants were conducted as well;
this effort took the form of electron microscope analyses of particles
collected by filtration of rain samples.
The field work was undertaken in December, 1974, in the vicinity of the
Centralia Steam-Electric Plant in southwestern Washington State. Our
experience at that site includes two previous experimental series. Those
provided a considerable amount of SO- and sulfate concentration data
useful in conversion studies, but both were inadequate in certain respects.
A study under a previous EPA contract in the spring of 1972 was concerned
primarily with SO- scavenging and an initial assessment of the relative
concentrations of S0? and sulfate in rainwater. Concentrations of the
nitrogen compounds and phosphate were not measured at that time, and the
pH data were not extensive enough to characterize the H -SO, relation-
ship in individual samples. Some further work based on this study, includ-
ing a rather qualitative examination of SO^-sulfate conversion mechanisms,
11
has been published . A brief field effort in early 1974, involving
only one experiment, concentrated on sampler-surface S0_ desorption
(touched on briefly in the earlier study) and testing of new sampling
methods. Nitrogen compounds and pH were examined in more detail, but the
12
sampling array was limited to near the power plant . Both of the earlier
Centralia studies were affected by nontypical operating characteristics of
the power plant itself; only one of the two units was operating in each
case, and electrostatic precipitator problems limited output. The 1974
session was also affected by the addition of sulfur trioxide to the stack
15
-------
effluent stream; this was done to increase the precipitator efficiency
during the construction of new precipitators. In view of all these limi-
tations and special circumstances, the December 1974 field trip was mounted
in an effort to obtain more complete measurements under normal plant oper-
ating conditions.
The Centralia Steam-Electric Plant (Figure 2) is a dual-unit, coal-fired
facility with a maximum output of about 1300 Mw. The coal (mined locally)
is a subbituminous type with a sulfur content averaging 0.55%. The
topography surrounding the plant is rugged, with heavily forested hills
rising to within 50 m of the stack tops within a few kilometers in the
normal downwind direction. Precipitation sampling sites are easily located,
however, on clearcut areas within 7 km, and on natural prairie areas beyond.
The general wind direction during rainy weather (from the south) is such
that extensive background levels of the pollutants of interest are rare;
the plant and its environs are an excellent location for examination of
fossil-fuel effluent effects.
PRECIPITATION SAMPLING
The primary objective of the field study was to collect precipitation sam-
ples in the area under the plume at distances from 0.4 to 11 km downwind.
The fixed collection sites numbered about 90, and were arranged as shown in
Figure 3. With the exception of an area of heavy forest between sites 72
and 73, and a coal mining region generally east of 61 and south of 74, the
array covered the accessible area normally affected by the. power plant
plume. The sampling at the fixed sites was facilitated by the use of a
single precipitation collector, one which is designed to freeze, the collected
rainwater and thus preserve the sample to allow any number of separate chemi-
13
cal analyses to be performed . The sampler consists of a plastic funnel of
2
area 325 cm connected to a plastic bottle capable of holding and preserving
1.5 cm of rainfall up to 12 hours. In contrast to the 1972 study, where
separate collectors were used for S0?, sulfate, and H , the current samplers
were much easier to handle, thus allowing a far greater number to be
deployed by fewer personnel.
16
-------
FIGURE 2. THE CENTRALIA STEAM-ELECTRIC PLANT.
-------
143
124
.-.
125
126
123
131
129 "0
128 96
127
102
95
132
142
141
140
133134135 136
139
TENINO
122
121
120
94
97
93
99 98 100 92
101
32 83 103 84 85
81
80
87 88
89
90 91
68
69
67
70 60 71
73
BUCODA
75
-N-
66
76
63
CENTRALIA SAMPLING ARRAY
DECEMBER 1974
LINE
POWER
2km
FIGURE 3. THE CENTRALIA PRECIPITATION SAMPLING ARRAY.
18
-------
The fixed site array was divided into five areas listed in Table 1. These
coincide generally with the assignment of service personnel and approximate
lines of samplers at various distances downwind. (The Line A area is so
named because it coincides almost exactly with Line A of the 1972 field
study. The site numbers, however, are different.) For convenience, these
areas will be referred to by name in the balance of this report.
Mobile Sampling
Two mobile precipitation collectors were outfitted to provide short-time
(as short as one minute) samples at locations which could be selected as
the experiment went on. One of these is shown in Figure 4; it consists of
2
a large plastic-lined funnel of area 1 m mounted on cartop with plumbing
to the inside where samples could be bottled and frozen immediately. One
of the samplers had a provision for pH measurements (Figure 4); real-time
pH measurements allowed for location of the mobile site in relation to the
power plant plume, as observed by the presence of acid rain. In general,
the pH-capability sampler would locate the plume centerline, and its
operator could direct the other mobile sampler operator by radio to a suit-
Table 1. SAMPLING AREAS
Area
Name
BKG
Line A
Bucoda
Central
Tenino
Mean Distance
from Plant,
km
3.4*
0.4
4.5
8.0
11.0
Site Numbers
BKG
1-19
59, 60, 67-79
80-103
120-145
Upwind (south)
19
-------
FIGURE 4. PRECIPITATION SAMPLING FIELD EQUIPMENT, a) FIXED-SITE SAMPLING
STATION WITH ICE CREAM CARTON (TOP), S02 BUBBLER (LETT), AND
STANDARD FREEZING COLLECTOR, b) MOBILE pH MEASUREMENT
EQUIPMENT.
20
-------
-U,
FIGURE 4 (CONTINUED). c) MOBILE PRECIPITATION COLLECTOR (AT SITE 71),
WITH FIXED-SITE GEAR IN FOREGROUND. d) BATTELLE MOBILE
LABORATORY AT THE CONTROL CENTER, SITE 9.
21
-------
The mobile sampling capability was included in the experiment design for
several reasons:
1) Measured concentration in short-time samples could be compared
with accompanying fixed site concentrations to help identify
any sampler-related problems involving funnel-surface desorp-
tion of S02, speed of freezing, or dry deposition.
2) A series of short-time samples, taken sequentially, could give
indication of fluctuations in precipitation concentrations due
to plume movements, rain rate variations, or variable chemical
reactions.
3) The mobile samplers could simulate a line of fixed sites across
the plume through a rapid traverse of the plume (the time
between successive samples at different locations could be as
short as five minutes); an attempt at this was intended as a
test of the potentials for mobile samplers to replace fixed sites
in future field work.
4) The samples thus derived would in principle be less susceptible
to contamination by local sources of pollution; these could be
filtered for insoluble particle analysis.
Other Precipitation Sampling
A third type of bulk precipitation sampler was used at selected sites.
This was merely a plastic bag contained in a paper ice cream carton
(Figure 4). Intended for metals analysis by atomic absorption spectro-
photometry, the samples thus collected could be separate from the other
fixed site samplers. The latter required less care in handling because the
inorganic pollutant concentrations are presumably in the part per million
range, rather than the expected part per billion range for the metals.
22
-------
An attempt was also made to examine the insoluble particle content of indi-
vidual raindrops. This was done by exposing fine pored (0.1 micron) membrane
filters directly to falling rainfall, in locations which were expected to be
near the plume centerline. These filters were saved for examination by
electron microscopy.
AIR SAMPLING
A limited number of air concentration measurements were made by use of
battery driven bubbler-impinger samplers. These were designed primarily to
provide background measurements of SO 0 , and NH . The latter two have
been proposed as catalysts in the SO.-sulfate conversion process. Several
bubblers were also positioned downwind of the plant to provide spot checks
of ground-level SO concentrations in the plume. The absorbing reagents
(impinger solutions) were as follows: for SO tetracholormercurate (TCM);
for 0.,, an alkaline potassium iodide solution; and for NH.,, a slightly
acid solution. Analysis methods are discussed below.
SUPPORTING MEASUREMENTS
The Battelle mobile laboratory (Figure 4) was utilized to provide a field
base of operations at the center of Line A about 400 m north of the north
stack. At that site (site 9) were located two tipping bucket raingaup.es,
one with standard 0.25 mm (0.01 inch) increments, and one fitted with a
larger funnel to provide 0.04 mm increments. The rainpnuges provided an
overall measure of the rain amount and rate to characterize the runs, and
also provided a time record of rain for comparison of site 9 amounts with
the amounts collected in the fixed samplers at other locations. The
faster response gauge was useful in comparing rainfall rates with those cal-
culated from raindrop size spectra. The latter were derived from raindrop
spot size measurements from exposed sheets of water sensitive, calibrated
paper.
23
-------
No attempt was made to measure wind characteristics; we chose to rely on
wind data collected at stack top height on the plant-operated meteorology
tower, which was located approximately 1.3 km northwest of the power plant.
3
These data proved to be adequate for modeling input for the 1972 study :
the added expense of installing our own instrumented tower was thus not
deemed practical for the present study. The reliance on the plant data
caused some problems, however, as is noted in Section VI.
EXPERIMENTAL PROCEDURE
A supply of dry ice was kept available for stocking precipitation collectors
at a storage location in the town of Centralia. When a rain was forecast as
imminent, or when rain was already occurring at the start of a work day, the
collectors were stocked (normally taking about a half hour), and the Tenino
and Bucoda area personnel dispatched. The field director, who maintained
Line A and the background site, prepared those samplers and made a visual
check on wind and rain conditions at the mobile laboratory site. His radio
report to each sampling area operator either commenced sampler setout or
called a delay until conditions were proper. The Central area operator
(mobilt/pH operator) either deployed samplers in his area or proceeded to
mak° short-time collections. The second mobile sampler was employed in the
Bucoda area; short-time samples from this sampler were begun (with radio
guidance) after that area's fixed sites were deployed. At the conclusion
of rainfall, onset of darkness, or otherwise after at least 0.15 cm of
rainfall were collected, the recovery of fixed samples began. Mobile
sampling continued as long as useful thereafter, depending on the fixed-
site obligations of the mobile operators.
CHEMICAL ANALYSIS
All chemical analyses of rain samples and air samples of NH,. and SO were
conducted at the Richland laboratories of Battelle-Northwest. Standard
automated methods for analysis of SO , SO, , NO- , total inorganic fixed
- - + + 3-
nitrogen (TIFN=NO + NO 4- NH4 ), NH , and P04 are listed in Table 2,
24
-------
Table 2. CHEMICAL ANALYSIS METHODS FOR POLLUTANTS
MEASURED AND NOMINAL^ LOW DETECTION LIMITS
Pollutant
Method
Lower Detection Limit
3 9
(g-moles/cm ) x 10
SO (dissolved)
so4-
NO.
NO,
NH,
H
15
Modified West and Gaeke
Methylthymol Blue
Saltzmann
Phenoldisulphonic Acid
Phenoldisulphonic Acid
Glass Electrode
16
0.1
2.0
0.03
0.3
0.35
(±0.1 pH unit)
Sensitivities for individual analysis sessions vary; these are indicated
in the chemical analyses results in Appendix A.
along with nominal minimum detectable concentrations. The sensitivity
limits vary somewhat, depending on sample quality and variations in
reagents, etc.; actual sensitivity limits for given days of operation are
indicated in the tabulations of measured concentrations, Appendix A. It
was anticipated that NH, would be measurable with a specific-ion electrode,
but experience with the probe and improvements to the automated method
revealed better sensitivity for the latter. (The improvement noted came
about through the use of a longer colorimeter flow cell.) The bubbler solu-
25
-------
tions were examined in the same manner as the rain samples, with the
exception of the ozone bubbler samples, which were done in the field via
14
the alkaline potassium iodide method ' This method produces a stable
iodide which is later liberated in the form of iodine by the addition of a
reagent.
26
-------
SECTION VI.
EXPERIMENTAL RESULTS
Three experiments involving significant collection of precipitation were
accomplished at Centralia during the December, 1974, field trip. The
experiments to be discussed here are called Fl, F3, and F4. As a conven-
tion, we routinely number any experiment which results in some data col-
lection. In this connection, F2 was an aborted precipitation run — no
rain occurred, but air concentration measurements were made. A similar
situation arose in the case of Run F5, where only a few marginally small
rain amounts were experienced.
Relevant power plant operating data are listed in Table 3. During Fl the
electrical output was essentially at capacity; the other runs were accom-
panied by a slightly reduced firing rate for one of the units. The hourly
electrical output was very uniform during the times of precipitation sam-
pling in all cases, and it was assumed that the coal firing was uniform as
well (the coal consumption data are hourly averages computed from daily
totals).
There were slight variations among the runs in the number of fixed sampling
sites active. This is indicated in the concentration tabulations in
Appendix A. The timing of site deployment and recovery is shown in Table 4.
The times of sampling of all the sampling areas coincided well; this was
facilitated by having continuous radio contact between all personnel. It
should be noted — and this is important in interpreting the concentration
results — that the times of rainfall were generally different from the
total sampling times. For all practical purposes, rain fell over the
entire array for the duration of sampling on run F3. The other two runs
contained rain periods with dry times between. Figures 5-7 are plots of
rain amount vs time as measured at the mobile laboratory, site 9. A com-
parison of rainfall rates measured at site 9 and at apparent plume center-
line fixed sites in each area is shown in Table 5. Run Fl is not included
because the showers were scattered and rainfall amounts varied considerably
over the array. Very little rain fell at site 9, but significant amounts
27
-------
Table 3. PLANT OPERATION AND EMISSION DATAC
Hourly Averages
Run
Fl
F3
F4
Hours
PST
0900-1700
1200-1700
1000-1500
Electric Power
Mw
1197 + 1%
903 + 5%
934 ± 2%
Coal Burned
metric tons
680
535
530
Sulfur Emitted
gram moles/sec
32.9
25.4
25.7
From log sheets provided by plant management
± indicates extreme variations
Hourly average based on daily total
Based on a sulfur content of 0.56% derived from average of daily
assays taken between 12-1-74 and 12-11-74.
28
-------
Table 4. FIXED SITE COLLECTION TIMES
Inclusive Sampling Times - First Out to Last In
VD
Run
Fl
F3
F4
Line A
Date BKG (1-19)
12-6-74 0911-1603 0929-1503
12-10-74 1340-1645 1300-1558
12-12-74 1055-1516 0958-1413
Bucoda
(61-79)
0916-1524
1233-1643
1014-1447
Central
(80-102)
0935-1542
1245-1602
1012-1345
Tenino
(120-144)
0905-1643
1230-1632
1028-1431
Table 5. RAINFALL RATES SUMMARY
Runa
F3
F4
Rain Time Site 9 J
Fractionb J, mm/hr BKG Line
1.0 0.76 0.27/0.61 0.56/0
0.4 2.4 2.0/2.3 2.6/2.
(Collected rainc)/J
A Bucoda
.73 0.97/0.81
7 1.3±.2/1.3
(Site 9 gauge)
Central
0.93/0.79
2.6/2.4
Tenino
1.2/0.80
1.7+.5/2,
Rainfall too scattered on Run Fl for meaningful analysis.
Fraction of average total sampling time during which rain fell.
At apparent plume centerline.
-------
L4
1.2
e
E
_r LO
<
LU
I 0.8
0.6
0.4
0.2
RAINFALL, RUNF1
MEASURED AT SITE 9
1030
1100
1130
1200
1230 BOO
TIME, PST
1330
1400
FIGURE 5. RAINFALL AMOUNT VS TIME MEASURED AT THE MOBILE LABORATORY:
RUN Fl.
RAIN SIZE SAMPLES *
* *
RAINFALL, RUN F4
MEASURED AT SITE 9
0900
1000 1100 1200 DOO 1400
TIME, PST
1500
FIGURE 6. SAME AS FIGURE 5, EXCEPT FOR RUN F4.
30
-------
MOBILE SAMPLES
SITE 71
SITE 101
SITE 82
SITE 59
RAIN SIZE SAMPLES
E
E
0
RAINFALL, RUN F3
MEASURED AT SITE 9
1230
1300
1330
1400
1430 1500
TIME. PST
1530
1600
I-H;UKK 7. RAINFALL AMOUNT vs TIME MEASURED AT THE MOBILE LABORATORY: RUN
-------
(which were largely unobserved with regard to rate and timing) fell else-
where, both east and west of the apparent plume heading. Comparison of
raingauge data with collected rain amounts was then of little value for
run Fl. The two rainfall rates shown in Table 5 for each site were computed
by assuming that the time of rainfall (40% of the total sampling time for
F4, and 100% for F3) at the site indicated was the same as that at site 9.
The good agreement between the numbers suggests that the rainfall was uni-
form over the entire array for both runs. Furthermore, in run F4, which
contained two or three (depending on the time span of sampler deployment)
separate rain periods, it is apparent that the assumption that the rain
started and stopped everywhere simultaneously was an acceptable one.
Several raindrop size spectra samples were collected during each of runs F3
and F4. Table 6 lists the spectra chosen from these to represent the rain
character of the run as a whole. The choice was made on the basis of a com-
parison of the rainfall rate from the spectrum to that of the raingauge for
the entire run. In F3, the spectra at different times were similar, reflect-
ing the continuous, prefrental nature of the rainfall. The F4 spectra
varied considerably between showers and were more bimodal in character. The
chosen spectrum for F4 came from the rain shower which provided the majority
of the rain amount to the samples.
Wind and temperature data are listed in Table 7. Run Fl is broken into
times roughly coinciding with the rain showers at site 9. The same approach
would have been taken for F4, but the data from the plant tower were lost
due to instrument malfunction. The only recourse was to visual observations
and evidence that the peak SC>2 concentrations in rain occurred at a plume
heading of about 180°. The wind speeds during the day (as reported at the
Olympia National Weather Service station (17 km north of the plant)) which
had that heading were averaged to arrive at an approximate value for the
run time of F4.
As is always the case with field experiments of this type, the ideal run
is seldom accomplished. This is clear from the overall conditions of the
present experiments. The foremost criterion for a successful run is con-
tinuous and uniform precipitation over the sampling array; this was met
32
-------
Table 6. RAINDROP SIZE FREQUENCY3 DISTRIBUTIONS
Run
Diameter, cm
F3
F4
0.022
0.005
0.040
0.030
0.080
0.270
0.041
0.115
0.220
0.055
0.100
0.130
0.074
0.260
0.080
0.100
0.300
0.110
0.140
0.135
0.080
0.185
0.005
0.058
0.255
0.012
CO
LO
Tabulated values are the fraction of the total number of drops of a diameter less than the diameter
listed, and greater than the previous diameter listed.
Table 7. WIND AND TEMPERATURE DATAa
Run
Fl
F3
F4
Time Span*5
1030-1100
1215-1300
1315-1345
1200-1700
1200-1600
Speed
cm/sec
580
380
350
705
^500d
Direction0
degrees
177
176±10
190
174
•VL806
Temperature
C
9.5
10.5
10.0
9.0
6. f
From 15 minute average data collected at approximate stack top height on plant-
operated tower.
b Approximate times of rainfall periods.
c Azmuth from true north.
d Plant tower data missing. Average of hourly speeds from Olympia NWS station
which had a direction near 180°.
e Visual estimate, and from resulting deposition patterns.
f Site 9 surface measurement.
-------
satisfactorily in F3, but the average rainfall rate was low, requiring long
sampling times. In addition, the wind speed was generally above average for
storms of the region. The strong wind apparently did not seriously affect
fixed collector collection efficiencies (see Table 6), but it is undesirable
for Line A measurements, since many raindrops collected there undoubtedly
undercut the power plant plume. The rainfall of F4 was uniform in extent
and timing, but not in rate, and there were dry periods. The latter
increases the potential for dry deposition to be a significant component of
the concentrations. Run Fl breaks down on several points as a successful
run, because the wind direction was probably variable in addition to the
rainfall problems noted above.
The times and locations of mobile samples are indicated in Figures 6 and 7.
Due to a vehicle breakdown and intermittency of rainfall, none were accom-
plished during Fl. The emphasis was on sequential sampling during F3, once
the apparent plume centerline was located. A traverse of the Bucoda area
was accomplished at the start of F4 (in place of Central area fixed site
deployment) but the eventual letup of rain limited the Central area
traverse. Because none of the runs was associated with an extended period
of rain, a downwind traverse of mobile sampling (to beyond 11 km downwind)
did not occur. After the close of fixed sampling on F4, a series of mobile
samples were collected along Line A, during an energetic convective shower.
An attempt was made to traverse the plume, but as the shower progressed,
the wind direction shifted in the same direction as the traverse. Thus,
the samples generally were all taken near or under the plume center. This
series of samples did explore the effects of variable rainfall rate and
possible changes in concentration with time within an isolated shower.
POLLUTION CONCENTRATIONS IN RAINWATER
The fixed site rain concentrations of the species SO , SO , NO , NO ,
NH. , H , and PO, ~, along with rainfall amounts are listed in Tables 17-20
of Appendix A. Similar results, plus rainfall rates, for the mobile
samples are listed in Tables 21-23. Where the rainfall amount entry is
blank, no sample was collected. Otherwise, blank entries mean that no
34
-------
analysis was performed. The symbol "NE" indicates that not enough sample
was left for analysis for that pollutant. Where a number is preceded by
"<", the value is generally the lower detection limit for that pollutant
during the analysis session. The symbol "£." means that the value following
is an upper limit, resulting from uncertainty in true concentration due to
instrument or calibration problems.
Sulfur Dioxide
On the basis of our experience at Centralia it was clear that the rainfall
deposition pattern for SO should be the most discernible. Therefore, for
the purpose of locating the plume and identifying samples upon which to
emphasize the analyses for other species, we performed the S0? analysis
first for all samples. The scavenged "plume" of SO is clear in all three
runs (though somewhat off the array in Fl): Figures 8-10 are maps of the
sampling array with numbers representing S0_ concentration in g-moles/1.
The isolines are somewhat schematic in view of the sampling coverage limi-
tations, but represent reasonable confidence regarding the shape of the
deposition pattern. Ground-level air concentrations of S0~, as measured by
i i ^- •}
the bubblers, in parts per billion (1 ppb - 5 x 10 g-moles/cm ) are indi-
cated within the boxes.
A useful indicator of the overall removal rate of S0~ from a plume is the
cross-plume integrated flux or scavenging rate. This quantity, generally
the rate of removal per unit time per unit distance downwind, is defined by
W J*° C(y) J dy * 1 £ (C. J. + C.+1 J.+1) |y. - y.+1 (5)
w —oo i"_L * '
where C is the concentration, J. is the rainfall rate (cm/sec), y. is the
i i i
cross-plume coordinate of sample position i, and n is the total number of
rain collectors along the cross-plume line. These were calculated from the
present data by selecting lines of samplers that crossed the plume, and
taking average values of J and y-separation in each case.
35
-------
S0?, RUNF1
39
(g - mole/cm ) x 10
2.0
0.7
0.7
0.2
0.6
1.3
0.5
0.8 0.6 1.5
1.0-
'0.3
1.0
0.3 0.2
1.0
0.2
1.0
0.5 1.4
1.1 \1.1
0.2
0.2
1.0
\
0.9
2.7
1.4
0.1
0.2 0
0 _ 0
O.E
n i
I 2.1 //
3.3
0.5 0
FIGURE 8 GROUND-LEVEL RAINWATER CONCENTRATIONS MAP FOR S02, RUN Fl
(SEE FIGURE 3 FOR IDENTITIES OF SAMPLING SITES).
36
-------
SCL, RUNF3
39
(g - mole/cm ) x 10
0.1
0.1
20 ^ '" o 0
\ 6.3 0
10
22
12
18
1.6
3.6
0.1
0.2
0.1
13 ' ' 2.1 0.2
10 I 28 \ 3.3 0
17
25
\5 33 33
\ \ rm'12
0
0.1
0.2
FIGURE 9. SAME AS FIGURE 8, EXCEPT FOR RUN F3.
CENTRATIONS IN ppb.
BOXES SHOW AIR CON-
37
-------
10
10
2.9 23
8.7 4.5
0.7
1.8 1.5
0.8
0.4 0.5
0.7
0.3
10
11
11
0.7
5.8
11 ' ' 12.3
16.1
0.1
0.2
0.2 0.8
17.1
18.2
7.8
\ 9.0
0 0 0
0
SO RUN F4
3 Q
(g - mole/cm ) x 10
0.2
0.2
0.2
0.4
0.4
0.1
0.3
FIGURE 10. SAME AS FIGURE 8, EXCEPT FOR RUN F4.
CONCENTRATIONS IN Dt>b.
BOXES SHOW AIR
38
-------
The results are listed in Table 8 and plotted as a function of downwind
distance x in Figure 11. For comparison, the results of previous similar
3 12
experiments ' are shown in the graph. Relevant source and meteorologi-
cal conditions are listed in Table 9. The great variability in scaveng-
ing rates for SCL at the nearest distance has been discussed before in
connection with the "C" series runs . In brief, this variability results
mainly from undercutting of the plume by small raindrops and near miss of
the SO^ source by other raindrops. In the former case, very low concen-
trations will result if a significant amount of the mass of rain is in
sufficiently small drops. In the latter case, the reversible scaveng-
ing behavior of SO-, being sensitive to sharp changes in air concentration
and rain acidity, can potentially result in a wide variety of rain concen-
trations to the ground. The tendencies for sampler-surface desorption may
be sharply variable as well - this is a function of plume characteristics.
All these factors are dependent on the mean wind speed, and the wind shear,
as well as on the character of the rain size spectrum. These are all
difficult to measure and characterize as an average for the 1 1/2 to five
hour sampling times.
The bunching of the scavenging rates within a reasonably narrow range at
x ~ 2.5 km is probably the result of the tendency for the power plant
plume to assume a similar character whenever rain is falling. This reflects
the fact that S0? scavenging is essentially "equilibrium scavenging": the
ground-level rain concentrations of S0_ are those which are in equilibrium
with the ground-level air concentration of S0~. The scavenging rates then
reflect the nature of the SO,., plume at ground level. These essentially
equilibrium scavenging rates become slightly more bunched if one "normalizes"
them in terms of Q/u (to which the air concentration is proportional), and
J. Some of the remaining variability would result from experimental error
(discrete concentration measurements) and topographic effects, which might
tend to misshape the plume when the mean plume heading varies slightly.
Sulfate
The analysis for sulfate ran concurrently with that for SO . Because of an
apparent laboratory contamination problem associated with the Fl samples, and
39
-------
Table 8. SCAVENGING RATES FOR S02: COMPARISON OF
PRESENT RESULTS WITH THOSE OF THE PREVIOUS STUDIES
Scavenging Rate at Distance x Downwind3
Exper-
iment
C-2
C-3
C-4
C-5
C-A
F3
FA
(gram-moles /cm sec)
Reference
3
3
3
3
12
Present
Present
0.4-0.5 km
26.
0.14
b
1.1
45.
3.0
27.
2.4-3.4 km
86.
33.
98.
b
b
b
b
4.5 km
b
b
b
b
b
100.
42.
x 109
7.0 km
b
b
136.
b
b
54.
b
10-11.5 km
66.
83.
233.
21.
b
115.
129.
Range of distances downwind reflect differences in mean line distances due to
different sampling arrays and mean plume headings.
Insufficient measurements.
40
-------
1UUU
rs
O
i — i
5 100
to
E
_o
"o
E
i
•& 10
u_r
i—
o:
o
o
1 1.0
o
on
0.01
^c'4
C/F4
C-4 c_3 ^*^M
L» — t r 5 Q
• C-2
1 x ^^ IT'S
n ° F4
8P, c"3
F4 C-5
^
• F3
°C-5
o C-3
l i
0. 1 1.0 10
x, km
FIGURE 11. MEASURED SCAVENGING RATES FOR S02 (cf.(5)) AS A FUNCTION OF DOWNWIND DISTANCE x
-------
Table 9. METEOROLOGICAL AND SOURCE CONDITIONS
RELEVANT TO SCAVENGING RATE COMPARISONS
Run
F3
F4
C-2
C-3
C-4
C-5
C-A
Date
12-10-74
12-12-74
2-28-72
3-1-72
3-5-72
3-9-72
3-1-74
Ja
mm/hr
0.
2.
3.
0.
2.
1.
3.
8
4
5
8
3
3
0
Wind Sulfur Source
Speed Strength
cm/ sec gram-moles /sec
705
500b
541
547
800
345
141
25.
25.
9.
13.
16.
11.
9.
4
7
3
5
1
7
8
Background
Rain
PH
5.
5.
5.
6.
5.
4.
4.
3
4
7
1
4
7
8
Average at raingauge site.
Estimate - see Table 7.
42
-------
instrument difficulties during the F3 sample analysis sessions, a complete
examination of sulfate concentrations was not accomplished on the first
analysis "pass." A selection of F3 and F4 samples were then rerun for
sulfate. Sulfur dioxide tends to oxidize to sulfate in bulk (unfrozen)
water samples*; it was necessary then to reexamine the S0? concentration to
assess the extent of conversion of previously measured SCL. The sulfate
concentration is then determined by subtracting the (molar) concentration
of converted S0~ from the total (molar) sulfate concentration. Often con-
version was substantial, so that the resulting sulfate concentration
determined for the sample was the difference between two rather large
numbers. The error resulting from this process then effectively cut
accuracy such that values for sulfate concentration listed in the tables
whose magnitudes are below about 10 micromoles/1 are uncertain by at least
50%. In general the sulfate molar concentrations were not large compared
with those of S0~, so that it was difficult to define a deposition plume
for sulfate. This fact is shown in the sulfate concentration maps, Figures
12-13, where the only discernible plume-related sulfate deposition appears
at downwind distances greater than about 7 km.
Hydrogen Ion
The hydrogen ion deposition patterns in Figures 14-15 generally reflect the
presence of the power plant plume (as defined by the S0? deposition pattern).
Among the species whose concentrations were determined, sulfate is the only
apparent significant contributor to rain acidity above the background level.
Comparisons between the H and SO, concentrations for runs F3 and F4 show
considerable scatter; whereas H = 2SO, would suggest that the acidity is
due to sulfate in the absence of other ionic species. A good deal of this
scatter could be accounted for in terms of inaccuracies in measuring sulfate,
as noted above; in addition, there may be other species in the rain, not
measured, which could contribute to the acidity. Local sources of effective
materials at certain sites may have influenced some of the data. Of
^Samples were normally unfrozen for up to 8 hours during each analysis
session. They were refrozen, however, between analysis sessions.
43
-------
11
20
3
29
21
v 6 7 SO/, RUNF3
on
26 „ 9 (g - mole/cm ) x 10
12
6 5 12
2 0
FIGURE 12. GROUND-LEVEL RAINWATER CONCENTRATIONS MAP
FOR SULFATE. RUN F3. SEE FIGURE 3 FOR
IDENTITIES OF SAMPLING SITES.
44
-------
10
9
2
19
20
11
20
20
8
20 9
17
13
IJ
21
14 14 IJ 20
15
8
18
13 6 18
0 J 22
S04=, RUN F4
3 o
14 (g - mole/cm) x 10
20
12
4 21
13 24
12
FIGURE 13. SAME AS FIGURE 12, EXCEPT FOR RUN F4,
45
-------
0 1 0
14
SO
60
\
11
15
16
37
66
68
40
17
\
10
15
H , RUN F3
(g - mole/cm3) x 109
63 13 4
69 29 42 35
16
53
\
21
39
V
69 64 71
23
\
\
16
40
\
\
5
21
19
18
46
31
0 63
FIGURE 14. GROUND LEVEL RAINWATER CONCENTRATIONS MAP FOR H+, RUN F3.
-------
,'' °'
13 8
2 20
0
6
13
13
22
1ft 20
12 16 17
19 H+, RUNF4
22
6
3
4 6 3
(g - mole/cm3) x 109
22
26
17
6
4 40 11 2
19 23
40
12
FIGURE 15. SAME AS FIGURE 14, EXCEPT FOR RUN F4.
47
-------
particular interest in this regard, however, is the H -SO, relationship
for the line of sampling sites in the Tenino area, at a distance of about
10 km downwind. These data, for Run F3, are shown in Figure 16.
The balance between H /2 and SO, concentrations is quite good in this
area, except at two sites, presumably in the center of the plume, where
the SO- concentrations appeared very low. It is true that "negative
washout" of S0? can occur where the rain acidity is so high that SO- solu-
3
bility is significantly reduced . From all available data for this run,
however, this does not appear to be the case, inasmuch as the H concen-
tration at these points was not sharply larger than at other sites where
significant SO appeared. For example, the same sites in Run F4 pro-
vided a quite continuous broad flat peak in the concentrations of S0?,
sulfate, and H . In this case, the H concentration was less than half
of the sulfate concentration, as it appeared to be for most of the other
sites for that run.
Nitrogen Compounds and Phosphate
There appeared to be no power plant plume-related trend to the measured
- + - 3-
rain concentrations of N0_ , NH, , NO , and PO, . The concentrations
ranged from near detection limits to values which are not excessive in
comparison with levels observed in unpolluted areas . There were
some sites which consistently showed significant concentrations, however.
In the case of nitrite, these sites were generally those located near a
railroad or highway. The rain in some farming areas, particularly at
sites near active cattle pastures and corrals, contained elevated ammonium
ion concentrations. One group of sites showed high NH, concentrations
during run F4. This cannot be explained in terms of power plant effects
or agricultural sources; it is conceivable that some contamination was
present in the area operator's vehicle or some source of contamination
traversed the road along which the samples were collected.
48
-------
40
30 -
20
10 -
s*
o
0
40
o
I—
<
\—
^
LLJ
O
O
O
30
20
10 -
0
121
122
123 125 127
SAMPLING SITE
129
131
FIGURE 16. CROSS-PLUME RAINWATER CONCENTRATIONS PROFILES IN TENINO AREA (x VLO km)
49
-------
AIR CONCENTRATIONS
Table 10 lists the air concentrations measured by the bubbler/impinger
methods for NH and 0~. Concentrations of the bubbler solutions for NH
were determined in the same manner as the rain samples, the solutions
having been acidified to form NH, . The 0., concentrations were determined
by manual means in the field, on the same day as collection. The lower
concentrations, in the region of 1 ppb, are approximate, due to doubts
about the validity of the bubbler technique for measuring very low concen-
25
trations . Other error indications are the result of uncertainties in
bubbler solution volume flow rates, and actual time of sampling due to
bubbler malfunction. S0~ air concentrations a:
deposition maps, Figures 8-10 and in Table 11.
bubbler malfunction. S0~ air concentrations are indicated on the SO,.,
INSOLUBLE PARTICLE ANALYSES
Several fine pored membrane filters (0.1 micron Millipore) were exposed to
falling raindrops at the same time as mobile rain samples were collected.
These were expected to represent the rain falling through the central
portion of the power plant plume, as indicated by pH measurements on the
scene and, later, the S0~ concentration in the rain samples. The filters
were exposed so that raindrop spots on the filters would not overlap; thus,
the images of particle-containing drops were expected to be visible on the
filter when later examined by electron microscopy. Such examination was
conducted for several of the filters, but it was found that the images were
not readily apparent. In addition, there were insufficient numbers of
particles on the filters to make extensive scanning electron microscope
(SEM) with X-ray fluorescence analysis practical. It was then decided to
effectively concentrate the particles from the rain on filter surfaces by
filtering the bulk rain samples, using the same filter material. This
increased the surface density of particles by about a factor of one hundred
(considering the volumes of rain samples on hand), but, of course, individ-
ual raindrop images were not obtained.
50
-------
Table 10. MEASURED AIR CONCENTRATIONS OF NH3 AND
NH,
Run
Fl
F2
F3
F4
F5
Site
BKG
BKG
8
11
BKG
BKG
7
BKG
8
71
Time of
Sample,
min
412
357
186
182
271
261
162
350
241
b
Volume of
Air Sample
£
490
480
240
260
310
300
185
400
275
b
Concen-
tration
ppb
3.9
14. 5a
135. a
4.2a
8.3
11.4
5.2
17.3
2.9
b
Time of
.. Sample,
min
412
357
b
b
271
261
b
350
b
258
Volume of
Air Sample
£
535
430
b
b
310
300
b
400
b
530
Concen-
tration
Ppb
3.5
<1.
b
b
1.7
<1.
b
<1.
b
<1.
May be contaminated by previously used impinger.
No measurement.
51
-------
Table 11. MEASURED AIR CONCENTRATIONS OF SO,
Run
Fl
F2C
F3
F4
F5
Sitea
BKG
8
9
10
11
71
10
11
71
84
BKG
6
10
71
126
BKG
5
10
71
74
124
BKG
6
8
59
Time of
Sample,
min
412
246
250
259
264
349
186
184
182
216
330
271
155
49
182
184
261
161
222
198
171
203
350
245
245
258
Volume of
Air Samples
a
490
340
360
360
375
385
260
270
240
320
300
310
175
55e
250
250
300
185
255
190
195
260
400f
280
275
550
Air Concentration
(gram-moles /cm 6)
x 1012
<.04
<.04
<.04
<.04
<.04
<.04
<.04
<.04
<.04
0.71
0.28
<.04
0.05
0.25
2.3
1.2
<.04
<.04
<.04
<.04
.04
0.52
<.04
<.04
<.04
.08
j
PPbc
<1
<1
<1
<1
<1
<1
<1
<1
<1
16.
6.3
<1
1.7
5.6
55.
28.
<1
<1
<1
<1
1.0
12.
<1
<1
<1
1.8
, See Figure 3. BKG 4.3 km upwind of power plant.
Lower limit for bubbler/impinger. method arbitrarily chosen as 1 ppb.
, Based on Run temperature (Table 7 ) and a pressure of 1 atm.
Non-precipitation run.
Represents only a small portion of the total run time.
Bubbler flow rate not measured; estimated as same as others.
52
-------
Four of the mobile rain samples were so treated, and the filters examined
by SEM. (One of the fixed-site samples — exposed for several hours rather
than several minutes — was also filtered and examined, but the particle
content, presumably from local sources, was very high. Apparently, fixed-
site containers will accumulate quantities of organic matter or other debris,
which does not seem to affect the inorganic, nonmetallic pollutant analyses,
but does lead to large insoluble particle counts.) Table 12 lists the
samples treated and some rough estimates of the count of particles per cm
of rain. These order-of-magnitude estimates are based on counts of parti-
cles residing on small, known filter areas and known filtered rain amounts.
Sample 66A (run F4; see Table 22) is considered to be a background sample as
far as the power plant plume is concerned, because of its relatively high
-9 3
pH (5.19) and low SO- concentration (1.0 x 10 g-moles/cm ). The particle
density on the filter was about one hundredth that of the samples collected
under the plume.
A large fraction of the particles appearing on the under-plume filters were
spherical or very nearly spherical in shape. These particles are assumed
9 f)
to be products of an efficient, high-temperature combustion process . The
numbers present indicate that the source is probably the power plant. Such
particles were observed on filter 66A, but in very much fewer numbers.
Examples of the particles observed on two of the filters are shown in
Figures 17 and 18.
The chemical composition of several of the particles on each filter was
examined by X-ray fluorescence techniques, using the primary electron beam
of the microscope to promote X-rays whose energies could be characterized
in terms of the elements present. The most abundant element among those
analyzable by this method (generally of molecular weight above that of Na)
was usually Si, though some angular particles were primarily iron. The
spherical particles had X-ray spectra which were more alike than those of
the angular particles (some of the latter of which were sand; i.e., nearly
all Si), with a primary emission peak corresponding to Si, secondary Al,
and small peaks representing Ca, Fe, K, and Ti. The non-spherical particles
were generally more diverse in content, with S, Zn, Ce, K, Zr, Mg also
53
-------
Table 12. RESULTS OF ELECTRON MICROSCOPE
ANALYSES OF FILTERED RAIN SAMPLES
Run
F3
Samplec
7 IB
Volume
ml,
133
Volume of Water
in Count Sample
7.1 x 10
-4
Estimated Particle
Concentration
Total
50,000
Spherical
5,000
F3
82C
66
3.9 x 10
-3
20,000
5,000
F4
66A
78
0.042
500
See Tables 21 and 22 for soluble pollutant concentrations.
Remaining after soluble pollutant analyses.
Volume of water from which count was made = total volume x
area of microscope photograph
total area of filter (17.35 cmz) Magnification
Particles diameter ^ ly ; Order-of-magnitude estimate.
54
-------
FIGURE 17. SCANNING ELECTRON PHOTOMICRO-
GRAPH OF A PARTICLE FROM
MOBILE RAIN SAMPLE 71 B. THE
DIAMETER IS ABOUT 4.9 ym.
FIGURE 18. SCANNING ELECTRON PHOTOMICRO-
GRAPH OF PARTICLES FROM
MOBILE RAIN SAMPLE 82 C. THE
DIAMETER OF THE SPHERICAL
PARTICLE IN RIGHT CENTER IS
ABOUT 5.5 ym.
55
-------
appearing, but had much less Al than the spherical particles. Some of the
angular particles showed more X-ray "hash" of low energy which corresponds
often to a significant amount of carbonaceous or organic materials.
Despite the concentrating effect that the bulk filtration had on the indi-
vidual particle analyzability, the volumes of water associated with the
mobile samples were still too small to provide a sufficiently densely coated
filter for practical particle sizing and determination of particle size
distributions. This would be possible, given the same conditions as were
extant during the present experiments, with the filtration of about 1 liter
of water. A good deal of the funds budgeted for the SEM work was used in
the unsuccessful examination of the rain-spot filters. Future work with
bulk rain sampling should give promising results, particularly in relating
the fly ash particle counts and masses with other pollutant concentrations,
and in determining the particle size distribution of scavenged insoluble
particles. An interesting comparison can possibly then be made with the
particle size spectrum which escapes the precipitators; the results could
reveal much about the scavenging efficiencies of various-sized fossil-fuel
effluent particles.
56
-------
SECTION VII.
COMPARISON OF SMICK CALCULATIONS WITH OBSERVED DATA
The Scavenging Model Incorporating Chemical Kinetics was applied to input
conditions corresponding to the major experiments of the two Centralia
field series. The sulfur dioxide oxidation mechanism employed in the
calculations was the liquid-phase ammonia-involved model of Scott and
Hobbs . Whereas these authors considered the liquid phase in the atmos-
phere in general, constant atmospheric concentrations of SO , NH , and C0_,
and fixed values of the equilibrium constants for the dissolution of these
gases, the present treatment assumes that SO is scavenged reversibly from
a specified plume to raindrops which are distributed in size. The reac-
tions involved in the Scott and Hobbs mechanism (described in Appendix B)
are included in the concentration calculations, performed at various levels
(as required by the SO,, scavenging regime) along the raindrops' trajector-
ies. The result is a test of the ammonia model's ability to account for
the observed values of rain concentration of sulfate at ground level.
It should be noted that some assumptions have been made, in lieu of more
precise knowledge of some of the contributing factors. Perhaps the most
important of these is the inclusion of a constant value for ammonia gas-
phase concentration and solubility constants. The latter value is based
27
on measurements made at high gas-phase concentrations only . In addition,
the equilibrium constants used in the intermediate reactions (involving
NH,,, NH, , CO , HCO- , CO , and H ) are taken as constants, and at values
listed in the literature for a temperature of 25 C. Moreover, there is
considerable dispute about the magnitude of the assumed first order rate
constant K for the final step, the oxidation of SO., to SO, . Various
treatments of the laboratory data have led to a range of values for this
constant. We have considered two values f°r K which differ by an order of
magnitude; these correspond to that suggested by Scott and Hobbs, and to
a value given later by McKay. Appendix B contains a listing of equili-
brium constants and the reactions involved.
57
-------
Preliminary to the application of SMICK to a plume situation, several cal-
culations were performed assuming a constant value for the gas-phase con-
centrations of SO . The results, listed in Table 13, correspond with tho
calculations performed by Scott and Hobbs , but in this case, the liquid
phase consists of raindrops distributed in size, and the solubility of SO
is dependent on gas-phases mixing ratio y0/^ . Since ynri is held constant,
bO2 ^2
the liquid phase concentration is in equilibrium with the gas phase, and is
also constant for all drop sizes. These calculations were performed for the
smaller values of K = 0.0017 sec only. (As will be shown below, an in-
creased K will result in predicted increases in. the sulfate concentrations,
but not significantly those of S02, NH, , or H .) It should be noted that
the concentrations reported as "SO " (both observed and calculated) in this
report are the sum of the concentrations of aqueous-phase SO , HSO» , and
SO,, . This is necessitated by the measurement technique for dissolved SO-,
which determines all three species. The magnitude of the concentration of
HSO~ generally predominates the others, the first dissociation process
-> 4
being nearly complete under most circumstances .
Table 13. SMICK CALCULATIONS WITH SCOTT AND HOBBS10
MECHANISM: UNIFORM GAS PHASE SO CONCENTRATION.
Gas phase
\-\JLL\-
NH3
5
0.007
0
0
dlUJ-. d
ppm
co2
311
311
311
311
L. -LUL1 y
so2
5
0.007
0.007
5
K, sec
0.0017
0.0017
0.0017
0.0017
Concentration, (gram
S02a
1.9 x 105
260
10.2
287
so4=
1.0 x 104
13.9
0.03
0.03
moles /cm )
NH.+
4
2.3 x 105
332
0
0
x 109
H+
0.47
0.47
10.7
282
3See note a, Table 14.
58
-------
The wide range of concentrations resulting from variations in y and y
IN il *5 o \J 2.
illustrates the strong influence of ammonia predicted by this model. The
third case listed — the Scott and Hobbs case for the general atmosphere —
presents the most realistic results in terms of a non-plume precipitation
situation. Clearly, the presence of ammonia is required in this mechanism
for any significant sulfate to appear, but the process also leads to rather
large ammonium ion concentrations and high pH.
The SMICK calculations, employing the Scott and Hobbs mechanism, a
bivariate-normal plume and actual experimental input data, are listed in
Table 14. Each concentration is a plume centerline value which is the bulk
rain concentration, averaged for rain size spectra measured during the runs.
Relevant input data for the current experiment series ("F" runs) are listed
in Tables 3-9; input data for the earlier "C" runs are listed in Table 9
3
and in another report . For comparison are given the observed peak concen-
trations at the various downwind distances, and the results of calculations
employing the EPAEC-mode of SMICK (SO- scavenging only). The latter
calculation involves no decay function ror the SO- in the plume (no con-
version or prior scavenging is considered); thus, the results may be
better compared with the sum on the observed S0_ and sulfate concentra-
tions, rather than with the observed S0? concentrations alone. The value
of K used for Table 14 calculations is that measured by Van den Heuval
28
and Mason . Two values of y were used: 7 ppb, as used by Scott and
3 29 30
Hobbs, and considered to be a reasonable background level ' ; and 1 ppb,
perhaps a more realistic one for the Centralia area (as indicated by mea-
surements, see Table 10). All the calculations employed the Smith and
31
Singer values of the plume standard deviations; the centerline values
are comparable, but may not necessarily reflect the real plume situation,
there being room for considerable variation in plume centerline concentra-
tions if the parameters are changed slightly.
In the case of either value of y , the predicted NH4 content and pH are
significantly higher than any measured. This follows as a direct result
of the reactions included in the mechanism. The large predicted concentra-
tions of dissolved SO result directly from the increased alkalinity, in-
asmuch as the solubility of SO- is a strong decreasing function of excess
H concentration. The latter is calculated at each vertical calculation
59
-------
Table 14. CALCULATED PLUME CENTERLINE RAINWATER CONCENTRATIONS: SMICK
(SCOTT AND HOBBS; K = 0.0017 sec"1) AND EPAEC (S02 ONLY).
o
Concentration, (gram moles/cm ~) x 10
Run
C-2
C-3
C-4
C-5
F-3
F-4
a(so2
X
km
3.1
11.5
2.4
10.9
3.5
7.0
11.2
11.2
4.5
7.0
11.2
4.3
7.0
10.2
273
86
924
278
407
223
147
183
519
356
256
485
321
232
so/
(350)
(107)
(1950)
(545)
(640)
(340)
(225)'
(298)
(841)
(564)
(403)
(710)
(460)
(331)
S°4=
1.3 (6.2)
2.4 (9.0)
2.3 (14)
6.1 (31)
1.7 (8.8)
2.5 (12)
3.3 (15)
4.1 (19)
2.2 (11)
2.9 (14)
3.6 (17)
1.9 (9.4)
2.4 (11)
3.0 (14)
= 1 ppb (7 ppb)
NH.+ H+
4
287
105
932
300
419
239
16&
203
531
371
274
497
336
249
(438)
(191)
(2060)
(680)
(736)
(438)
(324)
(406)
(944)
(669)
(511)
(807)
(558)
(430)
2.5 (0.6)
1.0 (0.3)
10. (2.5)
3.2 (0.8)
4.0 (1.0)
2.5 (0.5)
1.6 (0.4)
2.0 (0.5)
5.0 (1.3)
4.0 (0.8)
2.5 (0.6)
5.0 (1.0)
3.2 (0.6)
2.5 (0.5)
S02 ( EPAEC )C
64.
24.
102.
36.
60.
35.
23.
23.
68.
47.
34.
92.
63.
47.
Observed0
S0_a S0.= NH.
24 4
11.0 30.
4.0 16.
23.0 d
13.0 d
15.0 52.
12.0 20.
12.0 15.
6.2 9.4
33. 14. 2
28. 28. 3
22. e 33. 2
9.0 24. 19
18. 17. 5
11. 22. 6
d
d
d
d
d
d
d
d
.4
.6
.5
.5
.0
.5
H+
20
8
50
32
50
32
25
^30
79
79
63
32
32
32
) + HSOo" + SO ~
aq J 3
The peak values of all the species are not necessarily recorded at the same sampling site.
'No
decay of S0_ with travel time (compare with observed SOo +
No measurement.
^
'"Centerline" value was anomalously low. See Figure 16.
-------
step as are all the other relevant concentrations. The predicted values
for sulfate concentration are of the same order of magnitude as those
observed, but clearly the means by which they are arrived at (including
high NH, and low H values) are inappropriate. It should be noted that
the observed deposition patterns for sulfate for the F runs are somewhat
different from those of the C runs, when the plant operating character-
istics differed. In general, the F run deposition patterns, representing
the scavenging under normal operating conditions, were not clearly plume-
related at distances less than about 10 km.
The effect of varying K and y was investigated by considering the center-
line concentrations for run F3. Table 15 lists the results for various
combinations of y (0, 1, and 7 ppb) and K (0.0017 and 0.013 sec ).
The latter value of K is that of MacKay, who arrived at that figure through
32
a reassessment of older data . The computations show that the absence of
ammonia (which may also be considered an extreme case for low ammonia solubility)
results in very low sulfate at all distances (the combination y.TTI =0, K=0.0017,
Nn_
was not run, but it is clear from the other examples that this would only reduce
the sulfate concentration), though the H and SO- concentrations are more in
agreement with the observations. The S0? concentration is, however, essentially
the EPAEC-mode value, and the H value is equal to that of HSO , the predomin-
ant constituent of dissolved S0?.
The presence of ammonia, on the other hand, results in very high ammonium
ion and low H concentrations, with not much variation between the two
values of K. From this it is highly apparent that ammonia solubility is not
accounted for appropriately by this mechanism.
The quite good agreement between the EPAEC-mode (or SMICK with yNR =0) results
and the total of measured sulfur species may be evidence that some other pro-
cess is occurring in the liquid phase, converting previously scavenged S02 to
sulfate. The accuracy of the S09 scavenging-only model is considered to be
"factor-of-twc", however, so this result may be fortuitous. It should also be
noted that the relationship between sulfate and H concentrations does vary
between runs (cf. Figure 16 comparing runs F3 and F4).
61
-------
Table 15. SAMPLE SMICK PLUME CENTERLINE RAIN CONCENTRATION CALCULATIONS
WITH SCOTT AND HOBBS MECHANISM: RUN F3 INPUT DATAa.
PC02 = 311 ppm.
3 9
Concentrations, (gram moles/cm ) x 10
x,
km
PNH
4.5
7.0
10.2
PKH3.
4.5
7.0
10.2
PNH3 =
4.5
7.0
10.2
PNH3 =
4.5
7.0
10.2
so2b
3 = °
66.
46.
34.
1 ppb
519.
356.
256.
1 ppb
512.
352.
251.
7 ppb;
841.
564.
403.
sv
K = 0.
0.20
0.28
0.37
K = 0.
2.2
2.9
3.6
K = 0.
16.
20.
25.
K = 0.
11.
14.
17-
NH.+
4
013 sec"1
0
0
0
0017 sec"1
531.
371.
274.
013 sec"1
551.
396.
305.
0017 sec"1
944.
669.
511.
H+
66.
47.
35.
5.0
4.0
2.5
5.4
3.9
5.4
1.3
0.8
0.6
See Tables 3-9.
5See note a, Table 14.
The cross-plume deposition pattern, as calculated by SMICK for run F3 for the
Tenino area (x ^ 10.2 km) and for one set of y and K values, is shown in
31 3
Table 16. The Smith and Singer sigma values appear to have been a reason-
able approximation to the real plume spread in this case. The computed
scavenging rates (cf. (5)) show that the same variations in magnitude occur as
with the centerline concentrations.
62
-------
Table 16. CROSS-PLUME SMICK RAIN CONCENTRATION CALCULATIONS: RUN F3,
TENINO (x = 10.2 km) AREA. K = 0.0017 sec"1; PNH = 1 ppb;
PCO = 311 ppm.
39
Concentration, (gram-moles/cm ) x 10
Position
Centerline
± 5°
± 10°
± 15°
± 20°
± 25°
Calculated
so2»
256
179
51
6.1
0.2
^0
so4=
3.6
3.1
1.6
0.2
^0
*Q
™4+
274
197
70
27
22
22
H+
2.7
1.9
0.7
0.3
0.2
0.2
-2'
7 fiC
/ . o
15.7
1.8
0.1
<0.1
d
SO
23
15
16
d
d
d
Observed0
.- NH.+
4 4
1.5
2.5
2.8
1.0
d
d
H+
56.
31.
12.
14.
5.4
7.8
e 9
Scavenging Rate , (gram moles/cm sec) x 10
1430. 27.
1880.
20.0 85.0
168.
988.
Average of concentrations measured at approximate ± 5 degree spacings from
plume centerline at Station 126 (alternate stations, except for centerline,
where the average of three stations is listed).
See note a, Table 14.
CIncludes two anamolously low concentrations, see Figure 16. Thus the scav-
enging rate may be low.
No measurement.
Q
See equation (5). Observed rates calculated from data in this table only.
No deposition evident.
63
-------
While the above results were derived from calculations made using T=25C, it
appears that the results for lower temperatures depart further from the
observations, especially in terms of NH, concentration. Limited runs using
equilibrium constants corresponding to T=10C show an increase in ammonium
ion concentration; this results from current knowledge of ammonia solubility,
which is that the solubility increases with decreasing temperature. The
resulting sulfate concentrations are probably increased as well, but these
do not give much support for the efficacy of the ammonia-based model.
Thus, it appears that the Scott and Hobbs mechanism alone cannot account for
the Centralia observations. Its effectiveness depends to a large degree on a
significant solubility of ammonia in water, and a subsequent fast ionization.
The equilibrium constants used for these processes are imperfectly known —
more research needs to be done in the area of solubility of ammonia and other
gases. One cannot discount the possibility that ammonia plays a central role
in more polluted environments than that of Centralia, but it appears that
evidence for its action should include rather high ammonium ion concentrations
and high pH in rainfall.
64
-------
SECTION VIII
REFERENCES
1 Stoiber, Richard E., and Anders Jepson. Sulfur Dioxide Contribu-
tions to the Atmosphere by Volcanoes. Science 182: 577-578, 1973.
2 Kellogg, W.W., R.D. Cadle, E.R. Allen, A.A. Lazrus, and E.A. Martell.
Science 175; 587, 1972.
3 Dana, M. Terry, J.M. Hales, W.G.N. Slinn, and M.A. Wolf. Natural
Precipitation Washout of Sulfur Compounds from Plumes. EPA-R3-73-047,
Battelle, Pacific Northwest Laboratories, Richland, WA, p. 202, June
1973.
4 Dana, M. Terry, J.M. Hales, and M.A. Wolf. Natural Precipitation
Washout of Sulfur Dioxide. Battelle, Pacific Northwest Laboratories,
Richland, WA, BNW-389, p. 148, February 1972.
5 Liniger, W. and R.A. Willoughby. Efficient Integration Methods for
Stiff Systems of Ordinary Differential Equations. SIAM J. Numer.
Anal. 7: 47-66, 1970.
6 Sherwood, C.C. and M. Berman. An Exponential Method for the Solution
of Systems of Ordinary Differential Equations. Comm. of ACM 17:
699-702, 1974.
7 Carnahan, B., H.A. Luther, and J.O. Wilkes. Applied Numerical
Methods. New York, Wiley, 1969.
8 Briggs, G.A. Some Recent Analyses of Plume Rise Observations. Proc.
2nd Int. Clean Air Congress, Washington, D.C., 1970.
9 Hales, J.M. Fundamentals of the Theory of Gas Scavenging by Rain.
Atm. Environment 6: 635-659, 1972.
10 Scott, W.D., and P.V. Hobbs. The Formation of Sulfate in Water
Droplets. J. Atmos. Sci. 24: 54-57, 1967.
11 Dana, M. Terry, J.M. Hales, and M.A. Wolf. Rain Scavenging of S02
and Sulfate from Power Plant Plumes. J. Geophys. Res. 80 (30):
4119-4129, 1975.
12 Dana, M. Terry and D.W. Glover. Precipitation Scavenging of Power
Plant Effluents: Rainwater Concentrations of Sulfur and Nitrogen
Compounds and Evaluation of Rain Sampler Desorption of S02- Pacific
Northwest Laboratory Annual Report for 1974 to the USAEC Div. of
Biomedical and Environmental Research, Part 3: Atmospheric Sciences.
BNWL-1950 PT 3, Battelle, Pacific Northwest Laboratories, Richland,
WA, February 1975.
65
-------
13 Dana, M. Terry, J.M. Hales, 'C.E. Hane, and J.M. Thorp. Precipitation
Scavenging of Inorganic Pollutants from Metropolitan Sources.
EPA-650/3-74-005. (NTIS: PB-237 374/4G1) Battelle, Pacific North-
west Laboratories, Richland, WA, June 1974, p. 136.
14 Saltzman, B.E. Determination of Oxidants (Including Ozone): Alkaline
Potassium Iodide Method. In Selected Methods for the Measurement of
Air Pollutants. USHEW, Public Health Service. Cincinnati, Ohio, May
1965, pp. E-l/E-6.
15 Scaringelli, P.P., B.E. Saltzman, and S.A. Frey. Spectrophotometric
Determinations of Atmospheric Sulfur Dioxide. Anal. Chem. 39: 1709,
1967.
16 Lazrus, A., E. Larange, and J.P. Lodge. New Automated Microanalysis
for Total Inorganic Fixed Nitrogen and for Sulfate Ion in Water. In:
Trace Inorganics in Water. ACS Advances in Chemistry Series No. 73.
Washington, D.C., American Chemical Society, 1968.
17 Saltzman, B.E. Anal. Chem. 26: 1949, 1954.
18 Granat, L. On the Relation between pH and the Chemical Composition in
Atmospheric Precipiation. Tellus 24: 550-560, 1972.
19 Petrenchuk, O.P., and V.M. Drozdova. On the Chemical Composition of
Cloud Water. Tellus 18: 280-286, 1966.
20 Yaalon, D.H. The Concentration of Ammonia and Nitrate in Rainwater
Over Israel in Relation ot Environmental Factors. Tellus 16: 200-
204, 1964.
21 Tarrant, R.F., K.C. Lu, C.S. Chen, and W.B. Bollen. Nitrogen Content
of Precipitation in a Coastal Forest Opening. Tellus 20: 554-556,
1968.
22 Junge, C.E. The Distribution of Ammonia and Nitrate in Rainwater over
the United States. Trans. AGU 39: 241-248, 1958.
23 Jones, M.J. Ammonium and Nitrate Nitrogen in the Rainwater at Samaru,
Nigeria. Tellus _2_3: 459-461, 1971.
24 Herman, F.A. and E. Gorham. Total Mineral Material, Acidity, Sulfur,
and Nitrogen in Rain and Snow at Kentville, Nova Scotia. Tellus j^:
180-183, 1957.
25 Katz, Morris. Analysis of Inorganic Gaseous Pollutants. In Air
Pollution. New York, Academic Press, 1968, pp. 55-58.
26 McCrone, W.C., R.G. Draftz, and J.G. Delly. The Particle Atlas.
Ann Arbor, Ann Arbor Science Publishers, 1967.
66
-------
27 Morgan, O.M. and 0. Maass. Investigations of the Equilibria Existing
in Gas Water Systems Forming Eloectrolytes. Can. J. Res. 5_: 162-199
28 Van den Heuval and B.J. Mason. The Formation of Ammonium Sulfate in
Water Droplets Exposed to Gaseous Sulphur (sic) Dioxide and Ammonia.
Quart. J. R. Meteor. Soc. 89: 271-275, 1963.
29 McConnell, J.C. Atmospheric Ammonia. J. Geophys. Res. 78 (33):
7812-7821, 1973.
30 Healy, T.V., H.A.C. McKay, A. Pilbeam, and D. Scargill. Ammonia and
Ammonium Sulfate in the Troposphere over the United Kingdom.
J. Geophys. Res. 75. (12) : 2317-2321, 1970.
31 Smith, M.E., and I.A. Singer. An Improved Method of Estimating Con-
centrations and Related Phenomena from a Point Source Emission.
J. Appl. Meteorol. 5: 631-639, 1966.
32 McKay, H.A. The Oxidation of Sulphur (sic) Dioxide in Water Droplets
in the Presence of Ammonia. Atmos. Environ. 5: 7-14, 1971.
67
-------
SECTION IX
NOMENCLATURE
Units: £ = length; t = time; m = mass (moles); p = pressure (atm);
r = radius; none = dimensionless.
a £ Raindrop radius
a ' al* 32' a3 Coefficients of Polynomial solved in Scott and
Hobbs Mechanism
A Subscript identifying pollutant
b Subscript denoting "bulk"
3
c m/H Pollutant concentration in raindrop
•j
GSo2 m/£ S0_ concentration in raindrop
3
C m/£ Pollutant concentration in bulk rain
H' £ /m Modified Henry's-law constant
i Index integer
J lit Rainfall rate
K t First order rate constant
2
K m/£ t Overall mass transfer coefficient based on gas
^ driving force
n Total number of precipitation collectors
affected by plume, crosswind
P p Partial pressure
p
Co? p Partial pressure of Co_ in atmosphere
p
NH,. p Partial pressure of NH in atmosphere
O
rAx m/£ t Reaction rate for pollutant A in rain
•Q
rAy m/£, t Reaction rate for pollutant A in air
t t Time
vAx £/t Velocity vector for pollutant A in rain
68
-------
Ay £/t Velocity vector for pollutant A in air
vz £/t Terminal velocity of fall of raindrop
w^ m/£ t Counterphase transport rate of pollutant A
W m/£t Cross-plume integrated flux (scavenging rate)
x £ Downwind distance; also subscript denoting
liquid phase
y Crosswind distance; also subscript denoting gas
phase
y.. Mole fraction of pollutant A in gas phase
£i.D
y
NH~ Mole fraction of pollutant NH_ in gas phase
•y
so_ Mole fraction of SO in gas phase
z £ Vertical dimension
Az £ Vertical separation of calculation steps
Molar li
•
in space
3
p m/ H Molar liquid phase concentration of pollutant A
3
p m/£ Molar gas phase concentration of pollutant A in
" space
a H Plume dispersion parameter in y direction
a H Plume dispersion parameter in z direction
afl r Plume dispersion parameter in azimuthal
direction
-------
SECTION X
APPENDICES
Page
A. TABULATIONS OF RAINWATER CONCENTRATIONS, DECEMBER
1974 71
B. THE SCOTT AND HOBBS SO OXIDATION MECHANISM 79
C. A DESCRIPTION OF THE EPAEC GAS SCAVENGING MODEL 88
D. EXAMPLES OF Az ADJUSTMENTS ACCOMPLISHED BY 93
REVISED EPAEC MODEL
70
-------
APPENDIX A
TABULATIONS OF RAINWATER CONCENTRATIONS, DECEMBER 1974
EXPLANATION OF SYMBOLS
The site numbers refer to sampling locations shown in Figure 3. The
concentration data are broken into groups corresponding to the general
sampling areas identified in Table 1. Where the rainfall amount column is
blank, no sample was taken. Blanks in other columns indicate that no
analysis was performed. The symbol "<" means that the measured concentra-
tion was less than the lower detection limit for that particular specie
during the analysis session. "_<" means that the following number is an
upper limit on the concentration, as a result of analysis difficulties or
other uncertainty. "NE" means that there was not enough water sample remain-
ing to conduct the analysis.
71
-------
TABLE 17: RAIN SAMPLE CONCENTRATION DATA: RUN Fl, DECEMBER 6, 1974
Concentration (gram-moles/cm ) x 10
NJ
Site
#
BKG
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
59
60
61
62
63
64
65
66
67
Rainfall
cm
.04
.12
.11
.13
.11
.12
.10
.11
.10
.08
.08
.09
.09
.08
.08
.08
.08
.08
.08
.06
.80
.77
.83
.70
.45
.41
.22
so2
0.2
<.2
0.1
0.2
5.2
0.1
0.1
0.2
0.4
0.3
1.2
1.2
0.8
0.2
0.2
0.1
0.2
<. 1
<. 1
0.4
<.l
<. 1
<. 1
<.l
<.l
<. 1
<.l
S04 N02" N03"
5.6
4.6
3.8
5.0
5.1
2.9
4.3
4.8
2.6
0.04
0.06
0.08
0.08 2.1
<.03
0.15
0.31 3.5
NH4+
1.9
2.7
4.4
3.6
4.0
4.5
3.3
3.2
4.4
2.0
2.9
H+
NE
66.
44.
42.
4.5
91.
60.
3.6
39.
91.
32.
91.
8.5
41.
7.4
54.
7.4
9.3
3.6
1.9
55.
66.
58.
63.
55.
78.
40.
po43'
NE
0.054
0.027
0.134
NE
NE
NE
0.074
0.142
-------
Table 17 (Continued)
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
,22
,16
.25
.27
.35
.39
.61
.93
.96
.56
.51
.17
.26
.30
.33
.31
.23
.22
.17
.16
.33
.35
.24
.25
.49
.36
.27
.09
0.1
<.l
0.2
0.5
0.9
2.1
3.3
0.
0.2
0.2
1.1
1.1
1.0
0.9
2.7
1.8
1.4
1.4
1.0
0,
24.
29.
28.
33.
16.
20.
0.19
0.05
0.30
0.07
0.03
0.04
0.04
0.04
0.03
0.04
0.05
0.13
0.21
0.03
<.03
0.07
0.14
0.06
0.06
0.03
<.03
0.07
<.03
<.03
0.04
<.03
0.13
9.0
3.6
2.7
2.2
2.8
2.8
2.7
3.8
2.5
4.6
4.0
4.8
2.6
3.3
3.5
2.2
2.0
2.3
2.4
2.0
7.1
2.5
3.7
4.0
4.0
4.3
1.7
0.5
6.8
53.
3.2
14.
60.
53.
74.
69.
53.
81.
40.
51.
62.
40.
87.
87.
76.
79.
81.
100.
83.
85.
45.
60.
87.
83.
100.
74.
79.
0.047
0.033
0.049
0.026
0.030
0.020
0.080
0.078
0.117
0.070
0.022
0.044
0.155
-------
Table 17 (Continued)
101
102
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
.54
.53
.48
.38
.39
.32
.33
.10a
.32
.31
.33
.35
.34
.13a
.38
.42
.45
.38
.31
.29
.073
.45
.62
.03a
<.l
0.3
0.2
1.3
0.6
0.5
0.8
0.6
.5
1.0
0.3
1.0
1.0
1
0.
0.
2.0
0.9
16.
15.
34.
43.
0.07
0.08
0.07
0.10
0.06
0.07
0.09
0.07
0.12
0.07
0.09
0.08
0.04
<.03
0.04
0.05
0.12
0.09
0.05
<.03
0.10
0.17
0.04
0.13
1.8
1.8
1.2
2.3
3.6
3.0
3.4
3.4
2.7
1.9
2.7
2.4
2.6
2.4
3.4
3.9
0.9
3.4
3.5
2.3
2.2
4.7
1.4
4.7
10.
48.
83.
71.
56.
4.<
56.
7.(
91.
91.
11.
63.
76.
112.
87.
102.
16.
96.
12.
85.
76.
8.:
96.
33.
69.
NE
:.02
0.024
0.019
0.020
0.018
0.041
0.020
0.022
0.022
a Incomplete sample due to water freezing in neck of funnel.
run.
Thus, the rain analyzed came early in the
-------
TABLE 18: RAIN SAMPLE CONCENTRATION DATA: RUN F3, DECEMBER 10, 1974
Concentration (gram-moles/£) x 10
Site
#
BKG
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
59
60
61
62
63
64
65
66
67
Rainfall
cm
0.06
0.14
0.11
0.16
0.17
0.13
0.15
0.16
0.16
0.14
0.14
0.13
0.15
0.13
0.29
0.27
0.26
0.30
0.34
0.33
0.33
0.29
so2
<.l
o!2i
0.25
0.17
0.21
0.23
1.6
5.7
9.8
10.6
3.4
0.25
0.10
< . 1
32.7
<.l
<.l
<.l
<.l
<.l
<.l
<.l
S04 N02"
0.13
0.10
0.04
18. 0.10
10.
12. 0.09
7.1
<2.
6.6 <.03
<2. 0.04
4.4 0.03
<.03
0.03
0.03
<.03
<.03
N03"
4.5
3.1
3.0
2.3
2.8
1.6
1.8
2.9
1.9
2.1
2.1
0.8
2.4
1.0
NH4+
1.7
<.5
<.5
5.' 5
•=.5
1.'7
3.4
3.0
1.1
<.5
5l.O
1.2
1.6
2.0
H+
7.4
4.9
4.5
5.1
53.
5.1
65.
25.
34.
69.
36.
12.
9.1
12.
69.
5.9
7.6
45.
6.2
20.
6.0
44.
PD
A
0.033
0.029
0.059
5.01
<.015
-------
Table 18 (Continued)
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
0.34
0.36
0.32
0.29
0.28
0.23
0.23
0.23
0.21
0.20
0.18
0.18
0.26
0.35
0.32
0.27
0.34
0.31
0.33
0.13
0.29
0.22
0.28
0.32
0.31
0.22
0.19
0.32
0.28
15.1
33.0
11.8
0.48
24.9
17.4
9.7
27.7
3.3
0.10
0.10
0.15
2.1
13.2
0.21
14.
8.3
14.
3.4
13.
6.8
<2.
28.
13.
11.
7.8
9.3
9.0
0.03
<.03
<.03
<.03
<.03
0.04
0.05
0.23
0.05
0.10
0.15
0.10
0.08
0.05
<.03
0.06
0.03
0.08
0.07
0.15
0.13
0.17
0.12
0.07
0.10
0.23
0.08
0.04
0.09
1.5
1.7
0.9
2.2
2.1
2.1
0.3
0.8
2.8
1.9
1.9
1.5
2.2
1.9
2.2
1.9
<0.4
1.2
1.9
5.9
5.6
2.4
1.2
<.5
6.7
6.9
2.3
3.6
1.8
9.0
4.0
4.5
3.5
1.7
4.9
5.5
7.4
2.1
58.
8.5
74.
76.
28.
23.
50.
36.
11.
4.2
3.5
58.
26.
20.
74.
34.
47.
40.
8.3
5.4
10.
26.
24.
16.
20.
15.
18.
68.
8.5
0.016
0.044
0.088
<.015
0.063
0.038
0.030
0.018
-------
Table 18 (Continued)
101
102
120 0.41 <.l 0.09 1.2 1.8 4.8
121 0.35 0.11 0.09 1.6 1.0 22 0 050
122 0.18 3.6 8.0 0.15 1.3 4.7 14
123 0.34 12.4 8.0 0.03 0.9 1.1 20 5012
124 0.44 17.9 8.8 0.06 42
125 0.37 0.17 33. 0.05 1.8 0.9 60'
126 0.42 1.6 31. 0.06 0.8 2.3 63 < 01
127 0.36 21.5 4.9 0.04 0.5 1.4 45. 0.026
128 0.37 13.5 22. 0.04 0.8 2.5 20
129 0.34 5.3 13. 0.03 0.5 2.2 16 0 024
130 0.37 <.l 23. 0.0.3 1.3 1.0 10 0 020
131 0.37 <.l 0.06 6.9
132 0.36 <.l 0.12 6.5
133 0.38 <.l 0 06 51
134 0.07 <.i o.n e'o
135 0.40 <.l 0.10 5 1
136 0.37 0.10 0.04 7*8
137
138 0.31 <.l 0.09 5 9
139 0.20 0.15 0.11 1.9 1 5 19*
140 0.30
-------
TABLE 19: RAIN SAMPLE CONCENTRATION DATA: RUN F4, DECEMBER 12, 1974
Concentration (gram-moles/cm ) x 10
oo
Site
BKG
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
59
60
61
62
63
64
65
66
67
Rainfall
_ cm
0.38
0.40
0.42
0.42
0.40
0.40
0.41
0.40
0.40
0.41
0.39
0.38
0.44
0.42
0.45
0.41
0.43
0.39
0.37
0.39
0.17
0.49
0.51
0.48
0.47
0.41
0.35
0.25
so2
<.10
o.'ie
<.10
<.10
<.10
<.10
<.10
<.10
2.1
9.4
14.9
18.7
1.5
6.9
4.6
1.8
<.10
<. 10
<. 10
<.10
0.77
<. 10
<. 10
<.10
<.10
<. 10
<.10
0.20
so4=
14.
54.
54.
<3.
9.4
9.0
21.
17.
8.9
12.
18.
14.
7.3
54.
54.
11.
18.
8.2
<3.
14.
7.3
6.0
<3.
N02"
0.05
0.11
0.08
0.10
0.04
0.11
0.03
0.03
0.04
<.03
0.08
0.03
0.04
0.03
0.06
<.03
0.04
0.14
0.06
0.08
0.28
0.11
0.04
0.06
0.16
0.11
0.11
0.13
N03-
0.9
0.7
2.0
1.0
1.0
0.4
0.8
1.1
0.5
<.5
o.'g
2.0
0.7
1.1
0.7
<.4
<.3
2.8
NH4+
1.8
1.2
<.5
1.0
0.9
1.4
1.2
< 5
i.'o
1.9
1.0
<.5
<.6
0^7
18.1
3.7
6.5
4.0
H+
4.3
5.5
4.8
4.0
36.
5.0
6.8
8.9
25.
78.
55.
81.
49.
17.
15.
11.
6.5
6.6
8.3
3.2
44.
6.9
10.
7.1
8.1
6.6
8.9
23.
po43-
<.01
<.01
0.024
0.046
0.047
-------
Table 19 (Continued)
68 0.22 0.10 10. 0.14 21.
69 0.17 <.10 20. NE 10.
70 0.18 0.22 15. 0.21 1.5 8.1 8.3
71 0.14 7.8 20. NE 3.6 19.0 15.
72 0.10 9.0 24. NE 4.0 19.5 27. 0.050
73 0.08 0.22 NE 4.2 18.6 5.9
74 0.06 0.18 22. NE 44.
75 0.07 0.15 14. NE 6.9 9.6 16.
76 0.08 0.40 23. NE 1.9
77 0.07 0.36 15. NE 3.5
78 0.09 0.10 NE 9.3
79 0.09 0.25 26. NE 8.1
80 0.42 18.2 7.0 <.03 0.6 2.2 30.
81 0.38 17.1 17. 0.04 <.5 5.0 26. <.01
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98 0.37 16.1 16. <.03 <.5 4.5 26. 0.011
99
100
-------
Table 19 (Continued)
101
102 0.34 12.3 5.9 <.03 1.7 3.7 21.
120 0.22 <.10 s4. 0.07 1.1 4.4 23.
121 0.24 <.10 11. 0.21 5.1
122 0.23 <.10 0.07 1.3 4.9 17. 0.039
123 0.25 0.73 12.2 0.06 1.2 2.8 10.
124 0.26 5.8 21. <.03 1.7 2.0 16. <.01
125 0.22 10.5 22. 0.06 0.7 2.9 20.
126 0.23 10.8 22. 0.04 0.7 6.5 24.
127 0.21 10.9 22. 0.08 1.3 3.7 26. 0.022
128 0.18 10.1 13. 0.08 2.3 3.9 17.
129 0.17 8.7 22. 0.06 1.5 4.0 17.
130 0.17 4.5 11. 0.04 0.6 3.5 12.
131 0.16 2.9 0.06 12.
§ 132 0.16 2.3 9.5 0.12 1.7 7.0 7.6
133 0.16 1.8 9.4 0.19 2.1 7.4 7.9 0.030
134 0.13 1.5 19. 0.29 2.4
135 0.12 0.76 16. 2.4 4.7 7.2 0.016
136 0.11 0.40 16. 0.18 6.3
137 0.10 0.47 15. NE 24.
138 0.09 0.71 23. NE 1.6
139 0.10 0.25 22. NE 32.
140 0.06 0.31 19. NE 3.3
141 0.05 0.73 NE 8.5
142
143
144
145
-------
Table 20. RAIN SAMPLE CONCENTRATION DATA: RUN F5; DECEMBER 15, 1974
Station Rainfall Concentration (gram-moles/cm X 10 ^
Bkg
80
81
82
84
98
102
126
127
128
129
130
133
136
cm
0.02
0.02
0.03
0.02
0.01
0.02
0.02
0.02
0.02
0.02
0.02
0.03
0.02
0.04
so2
<0.05
0.69
1.3
4.7
0.47
0.61
0.45
4.6
6,3
3.1
1.5
0.22
1.2
0.34
S04 N02 NO
<3.
30.
33.
45.
27.
22.
32.
41.
57.
77.
55.
40.
33.
48.
- + + 3-
NH H PO
8.9
24.
21.
71.
11.
45.
42.
33.
29.
110.
79.
96.
4.6
50.
81
-------
Table 21. CONCENTRATION RESULTS
DECEMBER 10. 1974
- MOBILE SAMPLES (AND ASSOCIATED FIXED SITES): RUN F3,
CD
to
Site Desig.
71 A
B
C
D
b E
59° A
71 (Fixed)
82 A
B
C
D
82 (Fixed)
101 A
B
C
84 (Fixed) ^
100 (Fixed)
a At start of
Time
PST
1350
1400
1415
1430
1445
1515
1302
1501
1510
1520
1533
1252
1437
1442
1449
1258
1324
sample
Duration
min
10
15
15
10
3
15
182
6
7.5
7
5
170
1.5
1.
3
178
174
Rainfall
Rate
mm/hr
0.38
0.74
0.62
1.6
0.74
1.1
1.0
1.0
0.31
1.0
0.68
1.1
3.4
6.3
0.44
1.1
1.0
Concentration (gram-moles/cm '.
SO
2
38.
22.
11.
3.3
8.2
13.
33.
28.
52.
57.
60.
9.7
<0.1
<0.25
<0.25
3.3
0.21
c
d
S°A
4
88
41
40
21
99
24
14
56
59
56
73
28
22
26
21
11
9
0.2 km
/-\ f\ i
NO
0.
< .
0.
< .
< .
< .
< .
<.
< .
< .
0.
0.
0.
0.
S of
»T C
09
03
10
03
03
03
03
03
03
03
05
04
03
09
Site
NO ~
3
5.5
5+.8
2.0
1.9
<0.6
1.3
0.9
0.6
<1.3
<1.2
<0.6
0.8
1.9
<0.6
1.4
1.9
1.9
101
i r\ n
) x 10
NH H (pH)
4
3 . 3+ . 6
<1.7
<0.5
<0.8
<0.8
<0.8
2.4
1.4
<0.8
<0.8
<0.8
2.3
<0.8
<0.8
<0.8
1.8
2.1
48.
20.
31.
11.
85.
26.
76.
68.
87.
96.
129.
74.
31.
46.
7.
47.
8.
(4.32)
(4.70)
(4.51)
(4.97)
(4.07)
(4.59)
(4.12)
(4.17)
(4.06)
(4.02)
(3.89)
(4.13)
(4.51)
(4.34)
2(5.14)
(4.33)
5(5.07)
-------
Table 22. CONCENTRATION RESULTS - MOBILE SAMPLES (AND ASSOCIATED FIXED SITES): RUN F4,
DECEMBER 12, 1974
oo
Site Desig
66 A
66 Fixed
71C F
G
H
I
71 Fixed
73 A
73 Fixed
82 E
F
Time'
PST
1017
1030
1032
1245
1300
1315
1054
1046
1103
1221
1258
a
Time of beginning
Based on time of
f\ -i IT i
Duration Rainfall
Rate
min mm/hr
3
247
2
10
15
8
198
2.5
183
^3
3.5
sample
rainfall of
2.
2.
6.
0.
0.
0.
1.
3.
1.
M).
1.
Site 9
6
7b
9
76
16
24
5b
5
lb
6
3
Concentration (gram-moles/cm
so2
1.0
13.
21.
11.
3.8
7.8
4.1
0.22
22.
14.
S°4=
19
6.0
25
72
58
45
20
8
-
47
57
NO ~ NO ~
0.14 0.
0.11 2.
<.03 " 1.0+_.
0.06 5.
5.
5.
3.
0.
4.
2.
<.03 4.
5
8
6
7
5
8
6
9
2
5
6
NH/
4.6
4.0
<1.2
5.3
2.6+1
4.3+1
19.
<0.8
19.
4.9
3.6
) x 10
H
6.5
8.9
21.
96.
59.
33.
15.
22.
5.9
78.
71.
(pH)
(5.19)
(5.05)
(4.68)
(4.02)
(4.23)
(4.48)
(4.82)
(4.65)
(5.23)
(4.11)
(4.15)
Actually 0.15 km E of Site 71
-------
Table 23. CONCENTRATION RESULTS - MOBILE SAMPLES: CONVECTIVE SHOWER, DECEMBER 12, 1974
03
Site, Desig. Time
PST
10
12
14
16
18
20
16
Time
b
A
A
A
A
A
A
B
at
1429
1438
1443
1449
1458
1506
1517
beginning of
Duration
min
3
0.5
2
3
3
b
4
Sample
Rainfall Concentration (gram-moles/cm ) x 10
Rate = + +
mm/hr SO SO NO NO NH H (pH)
^ ~I ^J .3 TE
2.1 <0.1 23 <.03 3.6 6.3 14. (4. 85)
30. 3.6 19 <.03 2.3 4.8 31. (4. 51)
6.4 22. 34 <.03 3.3 10. 76. (4. 12)
2.5 14. 40 <.03 3.8 11. 48. (4. 32)
1.5 65. 25 <.03 0.2 2.8 126. (3. 90)
b 64. 24 <.03 <.6 1.2 78. (4. 11)
0.30 18. 46 - 2.5 <.8 96. (4. 02)
Unknown
-------
APPENDIX B
THE SCOTT AND HOBBS MECHANISM
This mechanism is based on the assumption that the oxidation of
SO is catalyzed by ammonia. The following equilibria are involved:
S02(gas) + H20 * S02*H2° (6a)
S02'H2° 1 * H+ + HS03~ (7a)
i __
HSO > H + SO ~ (8a)
NH3(gas) —> NH3-H20 (9a)
NH 'H 0 v NH, + OH~ (lOa)
C02(gas) > C02-H20 (lla)
C02-H20 >• H+ + HC03~ (12a)
HCO~ > H+ + CO = (13a)
HO > H+ + OH~. (14a)
The corresponding equilibrium constants are defined:
(6b)
K. = [H+][HS00 ]/[SO_«H00] (7b)
Is 3 22
K9 = [H+][SO.=]/[HSO "] (8b)
£- S J J
\a= lm3'E20]/Pm3 (9b)
K, = [NH, ][OH ]/[NH. 'H00] (lOb)
la 4 4 /
-FT- _ fpn «HOl /P (11 Ti^
K = [H+][HC03~]/[C02-H20] (12b)
K2c = [H+][C03~~]/[C02-H20] (13b)
KW = [H+][OH~]. (14b)
The numerical values used in the calculations (Section VII) are listed in
Table 24. 05
-------
In the above expressions the brackets indicate molar concentrations, and
P is partial pressure in the atmosphere.
The S02 oxidation is assumed to be first order:
d[S04=]
- = K[S03-]. (15)
dt
For the sake of electroneutrality,
[H+] + [NH4+] =
IHS03-] + 2[S03=] + 2[S04=] + [HC03~] + 2[C03=] + [OH~] (16)
By rearranging expressions (6) -(14) and defining
cs()2 = [HS03-] + [S02-H20],
(16) yields and equation for [H+] :
a3[H+]3 + a2[H+]2 + a^H+1 + aQ = 0, (17)
where
a3 = -1 -K^K^P/^ (18)
a =
0 = 2[SO.=] -K- -K, Kn P.TTJ K- /K ,...
2 4 Is ha la NH3 Is w (19)
an = K.. ccr. + 2[SO,=] K, + K, IL P_n + K (20)
1 Is S02 4 Is nc Ic C02 w
a = K. (2K0 ccn + K K, P_. + K ) (21)
o ks 2s SO,., he Ic CO w
It has been assumed that [CO,,"] is negligible. The subroutine to
SMICK which solves (17) for [H ] uses a cubic equation-solving subroutine
and then calculates [HSO^} , [S03=] , [NH^t^O], [NH4+] , [CO^I^O] ,
[HC03~], [C03=], and
[H+] = [H+] - [HSO ~] =
ex -"
2([S03=] + [S04=] + [C03=] + [HC03~] + [OH~] - [NH4+] . (22)
The value for [S0,=] is found by integrating (15), and the value of
T"
is found by integrating the differential equation for absorption by
the falling drop ((!))•
c
O«J ^
86
-------
The use of this mechanism in SMICK does not follow Scott and Hobbs
procedure of assuming an S07(gas) - SO- (aqueous) equilibrium. Furthermore,
the introduction of ammonia into the system leads to the possibility that
IHSO ~] > [H+]; thus [H+] can be negative. This occurrence makes it
*-* 6-2C
necessary to continually recalculate the value of this variable for use
where needed elsewhere in SMICK.
Table 24. VALUES OF THE EQUILIBRIUM CONSTANTS (Equations 6b-14b) USED
IN SMICK CALCULATIONS WITH THE SCOTT AND HOBBS
MECHANISM (T=298K).
Constant
Kls
K2s
\a
Kla
"he
K
Ic
K
2c
K
Value Used
0.00124 moles/cm3
1.27 x 10
6.24 x 10
0.057
-5
1.77 x
3.4 x 10
-5
w
4.45 x 10
4.68 x 10
1.01 x 10
-10
-14
-20
moles/cm3
moles/cm3
moles/cm3
moles/cm3
moles/cm3
moles/cm3
moles/cm3
(moles/cm3)2
87
-------
APPENDIX C
A DESCRIPTION OF THE EPAEC GAS SCAVENGING MODEL
A detailed description of the model, with computer listing and glossary of
terminology, appears in another report . This appendix is to provide a brief
review of the gas scavenging theory involved, and a summary of the calculation
scheme of EPAEC, as background for the discussion of the modifications to
EPAEC in Section IV.
The basis for all washout calculations for present purposes is the property
that a gas will be absorbed or desorbed depending upon whether the concentra-
tion driving force is to or away from the falling raindrop. Mathematically,
this may be expressed by the form
NAo = -Ky(YAb - YAe> ' (23)
where N is the flux of pollutant A away from the drop at its interface.
fT.O
The subscripts A and o denote "pollutant A" and "interface," respectively, in
9
a manner consistent with terminology used in a more detailed account ; the sub-
script b denotes "bulk," indicating that the entity being described is an av-
erage or "mixed-mean" value, and y is the mole fraction of A in the local
rVD
gas phase. y is the mole fraction of A in the liquid phase expressed in
/\,6
gas-phase terms to allow incorporation with y as a driving force; it is,
AD
more precisely, the gas-phase mole fraction that would coexist in equilibrium
with the mixed-mean mole fraction in the liquid phase x .
The above equilibrium relationship can be expressed by the functional form
> ' (24)
In the event that the gas-phase fraction is linearly dependent on the liquid-
phase value, the above equation reduces to
= HXAb ' (25)
88
-------
where H is the Henry's-law constant. This may be expressed in terms of liquid-
phase concentration by the essentially equivalent form
y^ = H'c , , (26)
Ae Ab
where H1 = H/c is a modified Henry's-law constant, c being the total liquid-
x 3 x
phase concentration (nominally 1/18 moles/cm ). Nonlinear behavior in the
equilibrium relationship is accounted for in the numerical model simply by
applying the form (26) and varying H1 as the computation progresses.
To calculate washout concentrations one must rewrite (23) in terms of the
liquid-phase concentration, c . ' (abbreviated to c in Section IV). The result
Ab
is
dCAb 3Ky
—— = My - H'c ) , (27)
dz v a *Ab Ab '
z
where K is the overall mass-transfer coefficient, and v is the terminal fall
Y z
velocity of a raindrop of radius a.
Mass-transfer coefficients for gas absorption by falling drops have been the
subject of extensive discussion in the literature. The EPAEC model considers
limiting conditions, namely those posed by gas-phase limited and stagnant
drop behavior. Behavior of real systems should fall somewhere between these
extreme cases.
The gas-phase mass-transfer coefficient k can be estimated from the well-
known Froessling equation
2k a -2av „ ^
. f
Ay y Ay
where D is the diffusivity of SO in air, c is the total concentration in
Ay ^ Y
the gas-phase, and v is kinematic viscosity of air. For stagnant drop condi-
tions the liquid-phase mass transfer coefficient can be calculated using pub-
lished solutions to the continuity equation describing diffusion into a sphere.
Although the resulting values are concentration dependent, they can be approx-
imated by constant values corresponding to drops experiencing ramp-function
89
-------
changes in surface concentrations, thus,
5D c
k = X (29)
x a
where D is the molecular diffusivity of SO in water.
Equations (28) and (29) may be used to calculate the overall coefficient for
the limiting cases using the well-known relationship
K = — . (30)
y H c,, n
k k
x y
An appropriate form for the gas-phase concentration field y (x,y,z) must be
nJD
furnished prior to calculation of numerical results. At the present time
this procedure uses Pasquill-Gif ford bivariate -normal plume equation,
modified to account for pseudo first-order chemical reaction (of rate k) and
background. This is given as
^)2] + exp [-1/2
y z
/ i,
exp(- =-) + Y^'bkg
The solubility of a gas in water often depends upon concentration in a highly
non-linear fashion. This is true for SO : our previous laboratory measure-
ments have shown the relationship to be given approximately by the form
CSO = [S°2]a + [HS03]
2 + ~\ / + 2 o
1 - FH O 1 + V[H O 1 + 4K [SO ] /H
g L 3 Jex L 3 ex lLOW2Jg/ , (32)
H°
where the bracketted terms denote concentrations in moles/liter and [HO ]
is the concentration in solution of hydrogen ions donated by sources other
than the SO ionization reaction. H and ^ are the ("true") Henry's-law
and dissociation constants, respectively, for reactions (33) and (34):
90
-------
[SO.]
r •
2 aq
[HSO~][H_0+]
'
Here it is important to note that [SO_] denotes dissolved, undissociated SOn,
2 aq 2
while the total dissolved SO is denoted by c .
2. 9
Execution of the EPAEC model consists simply of solving (27) numerically, in
*
conjunction with subroutine - provided values of K , v , y , and H' , for a
y z Ab
number of selected drop sizes, and then distributing the results according to
the raindrop spectrum to obtain average concentrations. The computer code
performs this function and allows for a number of sophistications including
nonlinear solubility behavior, lofting plumes, and nonvertical rainfall.
Prior to entering the integration algorithm, an "equilibrium scavenging" test
is performed. This test determines whether or not the expected liquid-phase
concentration (within a drop of particular radius a) at the receptor loca-
tion is that which would be in equilibrium with the local gas-phase concen-
tration. In the former case, the liquid-phase concentration c is set to
f\D
CAb = ^b/H'(W ' (35)
where y' is the air concentration at the receptor location, and the integ-
.A-D
ration is bypassed. Otherwise, (27) is integrated along a slanted raindrop
trajectory (determined by the wind speed and the terminal fall velocity),
* *
beginning at a selected point above the plume and ending at ground level or
other selected receptor elevation. The integration is performed with a fourth'
order Runge-Kutta algorithm, which provides values of c in response to pro-
"^
*The identity of the gas essentially only appears in H'. Thus an H' sub-
routine for any gas may be used.
**Where y is 0.01 of the centerline value with the t>lume rise and rain-
drop trajectory accounted for.
91
-------
vided values of dc /dz. The exprected errors from this method are oi the
r\D
order of the fractional grid spacing to the fourth power.
The receptor-level value of c is returned to the main program, and the pro-
Ab
cess is repeated for the rest of the preselected raindrop sizes. The average
concentration is determined by summing over the raindrop spectrum via
N 3
I F.D. c,.
i i Abi
c = , (36)
avg N
I F.D
1 1 X
where D_^ and F^ are the diameter and frequency of the ith drop size, and N
is the total number of different drop sizes.
The input chemical and meteorological conditions are printed out, along with
concentrations for each drop size and the average. If desired, the program
can calculate at a number of cross-plume (y-dimension) positions and perform
calculations of cross-plume integrated flux ("scavenging rate", cf. (5)) at
any desired number of downwind distances.
92
-------
APPENDIX D
EXAMPLES OF AZ ADJUSTMENTS ACCOMPLISHED BY REVISED EPAEC MODEL
As in the original version of EPAEC, the starting point for integration
^XT' *T' 2T^ is calculated prior to beginning the integration. This starting
point is that where the drop first encounters an air concentration of the
pollutant which is some fraction of that at the plume centerline (e.g., y
£\D
{V V V - °-01 YAb(V V H)' or ZT-H + % V; (V V V is a
function of windspeed u , raindrop terminal velocity v (and thus radius a),
z
source height H, sampling distance x , a (x), and plume loft. The stack
B cp
intercept, or distance above the source point that the drop crosses the
(x, y) plane, is given by
-X V
S = B Z -H . (37)
u
If S < O ("undercutting"), z is set equal to S+H, and Az, the computational
grid i
loft,
grid spacing in the z direction, is zT/40. If S > O and there is no plume
z 3o~
(38)
= H + S + XTV2/u , (39)
- T • (40)
and A „ _ . T
Thus, the number of integration levels N, is
N ~ -ZT/A.z . (41)
Table 25 lists input data for two sets of computations where I was the pol-
lutant gas (the solubility properties are similar to those of SO ); the only
difference between them is the source strength Q. Table 26 shows some of the
results. The values of N actually used (in the non-equilibrium scavenging
93
-------
cases) vary both above and below the initially set values. More integration
steps are required for the higher value of Q because the solubility of I2
(as with SO2) is lower (higher value of H1); but the equilibrium scavenging
criterion is met for larger drop sizes, for this particular set of input
conditions at least. In the case of the more-stably-integrating lower Q
runs, the spacing could be reduced for all appropriate drop sizes.
Table 25. INPUT DATA FOR SAMPLE CALCULATIONS
Source Strength Q
Source Height H
Sampling Distance X
Wind Speed u
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
3ORT NO.
EPA-600/4-76-Q31
3. RECIPIENT'S ACCESSI Or* NO.
4. TITLE AND SUBTITLE
PRECIPITATION SCAVENGING OF FOSSIL-FUEL EFFLUENTS
5. REPORT DATE
June 1976
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
M. Terry Dana, Dennis R. Drewes, Donald W. Glover, and
Jeremy M. Hales
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Battelle-Pacific Northwest Laboratories
Battelle Boulevard
Richland, Washington 99352
10. PROGRAM ELEMENT NO.
1AA009
11. CONTRACT/GRANT NO.
68-02-1729
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory
Office of Research & Development
Environmental Protection Agency
Research Triangle Park, N.C. 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final 6/24/74-11/28/75
14. SPONSORING AGENCY CODE
EPA-ORD
15. SUPPLEMENTARY NOTES
16. ABSTRACT
A numerical model for predicting the precipitation scavenging of reactive pollu-
tants from power plant plumes has been developed. The model, called SMICK (Scavengin
Model Incorporating Chemical Kinetics), calculates collection, liquid-phase chemical
reaction, and desorption, if any, of multiple plume-bound pollutants as they interact
with falling raindrops and are ultimately deposited on the surface. Calculations
for any specific aqueous-phase kinetics mechanism can be performed with the model
by expressing the mechanism in appropriate sub-routine form. The model has been
tested against field experiment data.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
*Scavenging
*Raindrops
*Mathematical Models
Reaction Kinetics
Electric Power Plants
13H
04B
12A
07D
10B
13. DISTRIBUTION STATEMEN1
RELEASE TO PUBLIC
19. SECURITY CLASS (ThisReport)
UNCLASSIFIED
21. NO. OF PAGES
105
20. SECURITY CLASS (This page)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
------- |