EPA600/3
84/040
January 1984
REGIONAL ACID DEPOSITION:
DESIGN AND MANAGEMENT PLAN FOR A COMPREHENSIVE MODELING SYSTEM
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
REGIONAL ACID DEPOSITION:
DESIGN AND MANAGEMENT PLAN FOR A COMPREHENSIVE MODELING SYSTEM
By
The NCAR Acid Deposition Modeling Project
National Center for Atmospheric Research
P. 0. Box 3000
Boulder, Colorado 80307
Project Officer
Kenneth L. Demerjian
Meteorology Division
Environmental Sciences Research Laboratory
Research Triangle Park, N. C. 27711
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research
Laboratory of the U.S. Environmental Protection Agency (EPA) and approved
for publication. Approval does not signify that the contents necessarily
reflect the views and policies of the EPA, nor does mention of trade names
or commercial products constitute endorsement or recommendation for use.
-------
ABSTRACT
This report presents a design and management plan for the development
of a state-of-the-art Eulerian model for the study of regional acid deposi-
tion phenomena. This plan directly addresses the findings of the report
Regional Acid Deposition: Models and Physical Processes (NCAR, 1983). It
is a plan for the integration of recent advances in mesoscale meteorology,
tropospheric chemistry and advanced computing into a scientifically defen-
sible, state-of-the-art regional acid deposition modeling system. The pro-
posed use of an established, proven, mesoscale meteorological model will
significantly improve our understanding of the role of transport in acid
deposition. The inclusion of fundamental chemical process equations will
make it possible to test our understanding of the fundamental transforma-
tion processes via comparison with observed data. The focus on statistical
and uncertainty analysis will aid the interpretation of modeling results,
and hence facilitate the proper assessment of the source-receptor rela-
tionship. The modularity of the proposed model system allows the easy and
timely incorporation of new results of research sponsored by EPA or other
agencies.
This report first reviews the major physical processes of regional
acid deposition and then describes the structure of the two principal sub-
systems, the meteorology and chemistry systems. Concepts and some proposed
preliminary steps for model integration and validation are next discussed,
with a final section on the management plan. The need for interdiscipli-
nary interaction and cooperation, specialized working groups, modeling sym-
posia, and the recommended internal management structure are all presented.
This is the second of two reports prepared by the National Center for
Atmospheric Research (NCAR) for the Environmental Protection Agency (EPA)
under Interagency Agreement No. AD49F2A203 extending from July 1, 1982 to
May 31, 1983.
111
-------
CONTENTS
Abstract ill
List of Figures vi
List of Tables vii
EXECUTIVE SUMMARY 1
1. INTRODUCTION TO THE PROBLEM 3
1.1 The Physical Picture 3
1.2 Existing Models and Components of Models 5
1.3 The Chemistry of Acid Generation in the Troposphere 7
1.4 Acid Deposition Model Development and Testing 9
2. OVERVIEW OF THE SYSTEM 10
2.1 The Meteorology System 10
2.1.1 Development of Meteorological Components 14
2.1.2 Model Validation 16
2.2 The Chemistry System 19
2.2.1 The Chemistry-Transport Module 19
2.2.2 Initial and Boundary Conditions 21
2.2.3 Emissions 22
3. SYSTEM INTEGRATION AND VALIDATION 23
4. MANAGEMENT PLAN 26
4.1 Manpower Needs 26
4.2 Internal Management Structure 28
4.3 External Interactions 30
4.4 Facilities for Model System Development 32
5. REFERENCES 32
-------
LIST OF FIGURES
Page
Figure 2.1 Components of Recommended Acid Deposition Modeling
(ADM) System 11
Figure 2.2 Recommended Domain for Preliminary Development and
Testing of the ADM System 15
Figure 2.3 The Chemistry-Transport Module 20
Figure 3.1 Recommended ADM System (Physical and Computational
Components) 24
Figure 4.1 Proposed Management Structure for ADM System Development.. 29
VI
-------
LIST OF TABLES
Page
Table 2.1 Basic Variables Predicted by the Dynamic Model and Derived
Variables for Input into the ADM 17
Table 2.2 Summary of Meteorological Model Experiments Recommended
During Validation Phase of the Project 18
Table 3.1 Status of Components of Proposed ADM System 27
vii
-------
EXECUTIVE SUMMARY
This document presents a design and management plan for the develop-
ment of a state-of-the-art Eulerian model for regional acid deposition.
This plan is based on the conclusions presented in the companion document
entitled Regional Acid Deposition: Models and Physical Processes (NCAR,
1983).
The key conclusions of (NCAR, 1983) are:
• There are fundamental weaknesses in existing models of regional
acid deposition, particularly in upper-air transport and disper-
sion, omissions of detailed chemical reactions, cloud physics,
and the treatment of terrain and surface effects.
• Marked improvements are now possible due to recent advances in
mesoscale meteorology and tropospheric chemistry; the construc-
tion of a comprehensive regional acid deposition model is now
feasible.
• The development of such a comprehensive model requires a clearly
focused, multidisciplinary group effort under strong scientific
leadership.
• The Eulerian framework is most suitable for representing the
essential physical and chemical processes in regional acid
deposition.
We present here a plan for the integration of recent advances in meso-
scale meteorology, tropospheric chemistry, and advanced computing into a
scientifically defensible, state-of-the-art regional acid deposition model-
ing system. A model system developed according to this plan would differ
from current models in the following ways:
• It would be based on an established, proven, mesoscale meteoro-
logical model and its analysis techniques.
t It would use fundamental chemical process equations to predict
the relevant transformations; thus, by comparison with observed
data, it would be possible to test our understanding of the fun-
damental processes.
• It would incorporate the details of both wet and dry deposition.
• Effort would be focused on analyzing the sensitivity of model
predictions to uncertainties in chemical initialization and
parameter!' zati ons.
-------
• Effort would also be focused on the proper statistical interpre-
tation of the model predictions in the context of regional cli-
matology.
t A user-oriented post-processor would facilitate the interpreta-
tion and application of the model results.
• It would be modular and highly flexible and would thus allow the
easy incorporation of new results of research sponsored by EPA or
other agencies.
-------
1. INTRODUCTION TO THE PROBLEM
The Acid Deposition Modeling Project of the National Center for Atmo-
spheric Research has prepared a document entitled Regional Acid Deposition:
Models and Physical Processes (NCAR, 1983). It reviews the physical and
chemical phenomena that give rise to acidic precipitation and deposition on
regional scales from the viewpoint that mathematical modeling of these phe-
nomena is necessary and feasible. While the review is not exhaustive and
all-encompassing, it is lengthy. Accordingly, we will introduce the prob-
lem by briefly summarizing the contents and conclusions of that report.
The rise of the acid rain (more accurately, acid deposition) problem
to public awareness in the United States and Canada has occurred very ra-
pidly and recently. Both public and scientific awareness and activities in
Europe, especially Scandinavia, preceded those in North America. Indeed,
acid rain is not a new phenomenon; many of the causes and controlling fac-
tors and some of the consequences were recognized 100 years ago. Features
of the acid rain problem that are new are (a) our perception of the quanti-
tative questions that must be answered to gain a full understanding of the
essential chemical and meteorological processes, and (b) our ability to in-
vestigate the questions with field- and laboratory-measurement programs and
with mathematical models. Similarly, from the point of view of those con-
cerned with the effects of acid rain, there now exist reasonably logical
and mature formulations of (c) relationships between ecological systems
(and physical structures) and acid deposition that can be investigated
quantitatively. Also, as noted above, public awareness of the potential
effects and probable causes of acid rain is new, as is the understanding
that some kinds of pollution traverse political boundaries.
In the bulk of (NCAR, 1983), we examine the full range of meteorolo-
gical and chemical processes that are involved in the overall phenomenon;
that is, the production and deposition of acidified rain, snow, fog, mist,
and dry deposition of acid anhydrides over important inhabited regions such
as the east central United States and Canada. We pay particular attention
to issues in the study of acid rain through mathematical models. While the
scientific questions dictate the kinds of field measurements, laboratory
experiments, and model development to be undertaken (all of which are ne-
cessary), we are particularly interested in how to develop and employ cred-
ible models. By credible models we mean those that are built on basic phy-
sical and chemical processes and that can test hypotheses and guide the de-
sign and assessment of field measurement programs with the eventual goal of
predicting acid deposition rates and source-receptor relationships and of
providing reliable estimates of the effects of emission control strategies.
1.1 The Physical Picture
Understanding and modeling the acid rain phenomenon requires one to
3OO SW 3STH STREET
CORVAUJ& OREGON 97333
-------
recognize a wide range of physical and chemical processes and their inter-
actions. Briefly, these are (a) emissions of materials that cause and reg-
ulate acidity in precipitation and deposition, (b) meteorological motions
that transport and dilute the emitted substances laterally and vertically,
(c) the variety of physical and chemical transformations that alter the
physical phase and chemical properties (e.g., valence or oxidation state)
of the emitted substances, and (d) the meteorological factors and surface
adhesiveness that lead to deposition of the transformed substances. A less
well-recognized set of questions surrounds those properties of the Earth's
surface that control the rate of uptake of dry materials (e.g., gaseous S02
and/or airborne particles).
Because the principal acids in precipitation are sulfuric (H2S0lf) and
nitric (HN03), we are most concerned with emissions of sulfur and nitro-
gen. However, the hydrocarbons and their oxidation products are important
reactants in the chemistry which ultimately leads to HN03 and H2SOlf. Esti-
mates of anthropogenic emissions of S02 (mostly from coal- and oil-burning
electrical power plants and metal-smelting plants) and of NOX (mostly NO
and NQ2 from high temperature combustion processes, including those in auto
and truck engines and power plants) are reasonably reliable for the world's
industrialized countries. Much less credible, but probably less important,
are estimates of natural emissions of organic sulfur gases and of natural
NOX compounds. Natural sources of gaseous NH3 and particulate NH^"1", gas-
eous hydrocarbons, airborne mineral dusts, and lightning-produced NOX
must also be estimated reliably. Minor contributions to precipitation
acidity from HC1 and organic acids are often negligible.
Whether the key emissions are anthropogenic or natural, they are in-
jected into the atmosphere at or near the Earth's surface, usually within
the planetary boundary layer. Accordingly, boundary layer meteorology is
at the core of the acid rain problem. The physics of turbulence and con-
vection, diurnal variation in surface heating, terrain geometry, and sur-
face and boundary layer hydrology exert strong control over the initial
dispersion of the emitted substances. Further, during the time these sub-
stances spend in the boundary layer, their physical environment (e.g., tem-
perature, pressure, humidity, available sunlight) and proximity to surfaces
and to other pollutants such as aerosol particles control the rate and type
of chemical transformations that occur—and they are markedly different
from those that are favored above the boundary layer in the free tropo-
sphere. There is perhaps only one important acid precursor or regulator,
NOX from lightning, that does not begin its atmospheric life in the
boundary layer, although background tropospheric ozone is central to all
tropospheric chemistry.
In dirty or clean air, in the boundary layer and above, chemicals
react with each other. The precise rates and types of reactions depend
strongly on the local pressure, temperature, available sunlight (both
direct and scattered), the presence of liquid and vapor H20, and on the
local chemical composition (i.e., the spectrum of available chemical co-
reactants). In Chapter V of (NCAR, 1983), we organize our discussion into
categories of homogeneous reactions (gaseous and liquid) and heterogeneous
processes and by principal categories of chemical species. Key consider-
ations include the exact rates of transformation (oxidation) of S02 and
-------
NOX into H2SOit and HN03, the major pathways of transformation, and the
essential controlling agents.
In an oxidizing atmosphere such as that of the Earth, the oxidation of
S02 and NOX to \\2^k and HN03 is inescapable, given enough time in the
atmosphere. Practically, however, it is very important to know what frac-
tion of all of a region's emissions is oxidized and deposited within the
region and what fraction of the total is transported long distances (at
high altitudes, for example) for eventual deposition onto territories hun-
dreds or thousands of kilometers from the sources. This is to say that a
credible description and model of this physical system must include quanti-
tative treatment of material transport and transformation above the bound-
ary layer. Similarly, the factors that limit the rate of surface deposi-
tion and uptake of gases (dry deposition) must be treated quantitatively.
These include near-ground turbulence, the condition and type of the surface
(e.g., vegetation, soils), and the chemical stickiness and reactivity of
the relevant substances on the surfaces.
1.2 Existing Models and Components of Models
In Chapter III of (NCAR, 1983), we define, describe, and compare two
distinct types of models used for studying long-range transport of air pol-
lutants: Eulerian grid models and Lagrangian trajectory models. Also, be-
cause of different goals and problems facing air pollution meteorologists
and chemists, it has been useful to develop and employ distinctly different
models for air quality modeling (AQM) and acid deposition modeling (ADM).
For ADM, we conclude that the three-dimensional nature of the problem and
the importance of simulating with adequate generality specific source dis-
tributions and eventual control strategies require an Eulerian framework.
Existing ADM's have already contributed to our understanding of the
acid rain problem, but a number of phenomena have not been treated fully
yet, largely because of the relative youth of the ADM field. Reasonably
well-based treatments of each phenomenon have been attempted, but not in-
side one model; that is, the best available mathematical parameterizations
have not been coupled together. Individual models tend to be strong in one
respect, but very weak in others. A number of fundamental weaknesses that
are widespread, or even ubiquitous, can be mentioned. For example, exist-
ing acid deposition models do not allow for mixing of pollutants above the
boundary layer. Similarly, no recognition is given to different types of
precipitation (rain, snow, dew, etc.) or to the temperature and pH that
characterize precipitation scavenging and acid formation. No cloud-chem-
ical processes have been considered so far. Further, fundamental (or ele-
mentary) chemical reactions have not been treated with sufficient detail.
Instead, linear overall transformation rates have been employed (for exam-
ple, the rate of conversion of S02 to sulfate has been set equal to x% per
hour without regard to mechanisms or controlling factors, although seasonal
dependence of x is sometimes permitted). No published model has yet in-
cluded reasonably complete chemical reaction schemes, and nitrogen oxides
are usually omitted entirely. Similarly, dry deposition of pollutants has
been simulated with fixed deposition velocities, and dependences on winds,
surface topography, moisture, and vegetation types have been ignored. Sub-
grid-scale inhomogeneities in emissions, transport, chemical reaction types
-------
and rates, and deposition have not been included. Consequently, and also
because of lack of field data, the verification of ADM's has not progressed
very far. Few data and the use of nonmechanistic model parameterizations
have led to more model "tuning" in the past than is desirable in the fu-
ture.
Air quality modeling and regional meteorological modeling are better
developed fields than acid deposition modeling. The former have a longer
history and a greater data base than the latter. Fortunately, techniques
and results from AQM and regional meteorological modeling are valuable for
ADM. For example, the experience and results of AQM researchers in dealing
with large numbers of chemical reactions can be tapped. Schemes to classi-
fy and to reduce systematically the numbers of independent chemical reac-
tion equations offer help to ADM (see especially Chapter V of (NCAR,
1983)). Also, methods of incorporating emissions into AQM's and the AQM
emissions data base itself are largely applicable in ADM.
The relative maturity and quantitative nature of regional meteorologi-
cal modeling as a field can be of enormous benefit to those who seek to de-
velop new, more general and realistic acid deposition models. In Chapter
III of (NCAR, 1983), we present an overview of regional meteorological mod-
els. A brief history of their goals, methods, and capabilities is outlined
and the principal components of these models are identified. Briefly,
these are the mathematical or numerical aspects and the more physical fea-
tures. In the former category, we review the essential features of the
spatial grids in these models, the various numerical methods employed to
solve the governing partial differential equations, the lateral boundary
conditions, and the overriding need for adequate data analysis and data
initialization. In each consideration, much of the task at hand in ADM,
namely to model accurately the dispersion and transport of pollutants, is
closely related to the main purposes of regional meteorological modeling.
Thus, the progress and methods in the latter field can be tapped as future
ADM's are contemplated and designed.
Similarly, the ways in which the physical aspects of regional meteoro-
logical models have been improved and tested will be of benefit to ADM de-
velopment. These physical aspects include the transports of heat, moisture
and momentum at the Earth's surface, in the planetary boundary layer and
free troposphere, and the energy sources and sinks that govern the trans-
ports. Also included are phase changes of water and the interaction of ra-
diation with clouds and the surface. The fact that these physical phenome-
na occur on many disparate spatial scales, including scales shorter than a
model's grid spacing, necessitates parameterizations—relating the cumula-
tive effect of subgrid-scale phenomena on the fluid flow, for example, to
the model-resolvable scales of motion. Parameterizations of surface pro-
cesses, of planetary boundary layer processes, of condensation and evapora-
tion processes, and of radiative effects of layered clouds in current mod-
els are also reviewed, and strong indications of areas ripe for progress
are identified. While many of the simpler parameterizations of physical
processes now in use in regional meteorological modeling (RMM) are attract-
ive in the early stages of acid deposition model development, it is encour-
aging to note the progress in RMM toward tractable, improved parameteriza-
tions.
-------
Another important consideration in the field of RMM that will be di-
rectly usefifl in ADM is that of objective measures of model skill, i.e.,
the accuracy of model predictions. In Chapter III of (NCAR, 1983), we
review several standard quantitative measures of forecast skill and also
summarize the state of the art of RMM's to forecast (precipitation, for ex-
ample). Clearly, the ability of acid deposition models to forecast deposi-
tion patterns (say, annual totals) or deposition amounts in distinct events
must be measured objectively. The methods used in RMM will serve as good
guides at first.
Because of the great potential for transferring methods and parameter-
izations from RMM to ADM, we have reviewed components of the former models
in some detail, principally in Chapter IV of (NCAR, 1983). First, the need
for objective analysis is recognized—irregularly-spaced initial meteorolo-
gical data must be transformed to provide initial conditions on a model
grid. The techniques, quality, computational costs, and history of ob-
jective analysis methods are summarized and several case studies are dis-
cussed. The related need for data initialization is discussed in similar
detail. General physical considerations, mathematical analysis, and exper-
ience with meteorological models can indicate general spectral and transi-
ent characteristics of data-caused noise. In specific applications (e.g.,
for a specific regional topography and synoptic situation), there is both
sound theory and practical experience to guide the choice of initialization
procedure. Accordingly, unneeded computational costs can be avoided.
Also, as is true in all methods to solve differential equations, boundary
conditions must be specified. Principal techniques now in use in RMM's
(spatial damping (or sponge) conditions, wave-radiation conditions) and
bounded derivative schemes are reviewed with various applications in mind.
Numerical methods and mathematical principles for objective analysis, data
initialization, and boundary conditions are also reviewed in Chapter IV of
(NCAR, 1983). Once again, the available general theory plus the experience
of RMM researchers constitute a well-based foundation for ADM development.
On a more physical side, the essential RMM components mentioned above,
surface physics and effects, planetary boundary layer physics and effects,
and the thermodynamic and radiative physics and effects of clouds and pre-
cipitation are also reviewed. The methods and problems extant in the field
of RMM are very close to those that will prevail in ADM.
1.3 The Chemistry of Acid Generation in the Troposphere
As mentioned earlier, the chemical phenomena and reaction sets in ex-
isting acid deposition models are far from complete. This is so for many
reasons, including the fact that the importance of long-range transport of
pollutants has been perceived by the public and its agencies. Accordingly,
much work in ADM has focused on the meteorological aspects of transboundary
transport. Also, it is true that mechanistic information on the actual
chemical processes that transform S02 into sulfuric acid and NOX into ni-
tric acid has appeared very confusing and incomplete until recently. Also,
the chemistry of acid generation is more complicated than that of regional
chemical oxidants; the former involves gas-phase and aqueous reactions,
while the latter is due to gas-phase reactions alone.
-------
Accordingly, our discussions and review of the chemistry of.acid gen-
eration in Chapter V of (NCAR, 1983) are focused at first on the essential
chemistry itself rather than the chemistry now in the existing ADM's. The
main categories of the review are gas-phase reactions, aqueous-phase and
heterogeneous processes, and photodissociative processes. In any credible
ADM, it is extremely important—in fact, essential—to think in terms of
reaction mechanisms as opposed to depending completely on parameterizations
of overall reaction or transformation rates. For example, it is inadequate
to know only the rate at which substance A is transformed to substance C in
a mixture as complex and variable as a regionally-polluted atmosphere. In-
stead of the overall process
we require, instead, knowledge of elementary reaction mechanisms exempli-
fied by
A + B k ^ c + D,
where the rate k is specific to the two reactants A and B and to reaction
conditions (pressure, temperature) and the chemical identities of C and D.
Only in this way can a rigorous mechanistic understanding be developed
wherein the overall rates of the key transformations and their sensiti-
vities to pollutant and ambient chemical concentrations are predictable.
Without it, we would continue to be prey to unknown errors and to criticism
of the type that now applies to
S02 rate = x% per hour^ $0^=.
For example, this simple and widespread parameterization is inherently
linear: the rate of production of S04= is proportional to the gaseous
S02 concentration. In reality, the supply of the chemicals that actually
oxidizes S02 to SQk~ might be limited in certain locations, and little or
no SO^3 production could take place even when large amounts of S02 are
available there. Similarly, the S02-to-SOi+= conversion rate probably de-
pends on the exact species that is accompanying the oxidation so that the
rate, x, is not constant but varies with time. Obviously, analogous funda-
mental considerations apply to the conversion of NOX to nitric acid, to
the production of photochemical oxidants like ozone and peroxyacetylni-
trate, and to the production of S02 from biogenic organic sulfides, for
example.
The main goals of the very detailed presentations in Chapter V of
(NCAR, 1983) are to identify from available research results the principal
elementary reaction mechanisms and key species in the gas-phase, aqueous-
phase, and heterogeneous reactions that cause and control acid generation.
From a complex and encyclopedic list of chemicals and reactions, a smaller,
more concise list of chemical variables and processes must be distilled to
develop a tractable and useful ADM. From fundamental principles, labora-
tory data or photochemistry and kinetics, laboratory simulations of complex
systems and field data, we can explain the essence of acid generation.
These shortened lists of species and processes (elementary reactions when
possible) will require further testing, such as zero-dimensional sensiti-
8
-------
vity calculations. In some cases, such as gas-phase species (i.e., hydro-
carbons), the grouping into representative categories has been done in AQM
research previously, so only refinements will be needed for ADM develop-
ment. In other cases, such as solution-phase chemistry, it is not yet com-
pletely clear how to achieve conciseness in the reaction list while still
simulating the essential features and rates of reactions. This is partly
because the role of in-cloud chemistry in generating acids has been appre-
ciated only recently.
Certain clear indications of how to proceed in ADM development do ap-
pear in the course of our review. For example, because all gas-phase pro-
cesses that lead to S02 oxidation are initiated by the gas-phase HO radical
(in daylight, of course), it is clear that the major processes that control
HO concentrations must be embodied in the minimal reaction set for the
ADM. Similarly, because of its role in NOX chemistry and because it is a
major source of HO, tropospheric 03 must be calculated accurately. In the
liquid phase, it will be necessary to simulate behavior of 03, H202, HO,
H02, N03, and probably 02~ and N205. Fortunately, there is a large and
talented group of chemists working worldwide on precisely the reactions of
interest and there are several international panels who meet regularly to
prepare critical reviews of progress in chemical kinetics, so the funda-
mental data necessary in ADM development are forthcoming or are largely
available already.
1.4 Acid Deposition Model Development and Testing
In Chapters IV, V, and VI of (NCAR, 1983), we face many of the issues
that arise in the design of a comprehensive model, i.e., one which includes
coupled meteorology and chemistry. The key meteorological and chemical
processes that are identified in the earlier chapters of the report are
stated more concisely in Chapter VI, and certain other phenomena and prac-
tical considerations are introduced into the discussion. For example, we
discuss the apparent importance of dry deposition of acidic gases and par-
ticles, the available methods for its measurement, the controlling physics
and chemistry, and how an ADM might treat dry deposition. We also intro-
duce in Chapter VI the questions and facts concerning surface emissions of
pollutants and natural sources of acid precursors and of those species that
regulate acid generation. Other general features, components, and ques-
tions in ADM development are also reviewed and summarized in Chapter VI.
These include model resolution, subgrid-scale processes and how to begin to
treat them, mathematical and numerical techniques for large comprehensive
models, cloud considerations in models with coupled chemistry and physics,
and issues in model validation and sensitivity analysis.
While there are many issues and potential problems involved in the de-
velopment of a comprehensive acid deposition model, it is clear that this
field is ripe for progress. The two principal disciplines that are in-
volved, meteorology and atmospheric chemistry, have made dramatic if sep-
arate progress recently. Early attempts to include meteorological and
chemical processes in integrated models have been useful already, and the
experience of the contributing scientists can be tapped. Computational
facilities and methods are up to the task. With appropriate amounts of
enthusiasm, realism, resources, and teamwork, a new, greatly improved
-------
generation of acid deposition models can be born.
2. OVERVIEW OF THE SYSTEM
2.1 The Meteorology System
To understand the phenomenon of acid deposition, one must understand
the atmospheric processes of horizontal and vertical transport, turbulent
mixing, and cloud and precipitation formation, in addition to the complex
chemistry involved in the formation of acidic material. Because of the
transport of gases and aerosols out of the boundary layer by clouds, it is
necessary to consider the motions in the entire troposphere in order to de-
termine the transport of chemical species on regional scales. From a sci-
entific point of view, the processes of scale interactions, boundary layer
and radiative effects, and cloud and precipitation formation are among the
more exciting areas of atmospheric research. From an assessment point of
view, the correct modeling of horizontal and vertical transport, turbulent
mixing, wet and dry removal, and the possible nonlinear interactions among
these processes is essential in evaluating potential control strategies.
The necessary meteorological components of an acid deposition modeling
system are shown in Figure 2.1. Input data are routinely available from
surface stations, radiosondes, satellites, and aircraft. These different
data, at irregularly-spaced points, form the basis for a three-dimensional
analysis of the primary meteorological variables (pressure, temperature,
water vapor, winds) at regularly-spaced points on a three-dimensional grid.
An initialization procedure adjusts these data to a dynamically-consistent
set of initial conditions for the dynamic model. This model integrates the
equations of motion, the continuity equation for dry air, the theromodyna-
mic equation, and the continuity equation for water forward in time to pro-
duce the temporal variation of the meteorological variables in three dimen-
sions. A dynamic model processor then converts these primary data to the
meteorological data required by the acid deposition model (ADM). The ADM
utilizes these meteorological data to estimate the transport and transfor-
mation of the chemical species that contribute to wet and dry acid deposi-
tion. Finally, an ADM processor is required to convert the ADM data to es-
timates of total acid deposition over the temporal and spatial scales of
interest.
As reviewed in (NCAR, 1983), considerable progress has been made over
the last decade in developing the meteorological components of Figure 2.1.
Thus, the time is right for further development and application of these
components to the acid deposition problem. Because of the tremendous
amount of previous work, we feel it is possible to produce a state-of-the-
art meteorological system suitable for applications within a three-year
period. In addition, construction of the system with modular components
(such as analysis and initialization procedures and physical parameter! za-
tions) paves the way for future improvements as scientific advances occur
in the components.
While the modular framework for a meteorological model which will
provide the necessary meteorological variables as input to the ADM is at
the core of our design, another essential component is the validation (or
10
-------
Figure 2.1
Components of Recommended Acid Deposition Modeling (ADM) System
INPUT DATA
Sources: Surface stations, radiosonde, satellite, aircraft
Types: Pressure, temperature, horizontal Mind components,
water vapor, cloud cover, precipitation
ANALYSIS
Objective estimation of meteorological variables at
regularly-spaced points over domain
INITIALIZATION
Objective modification of analysis to achieve dynamically
consistent set of Initial conditions for model
DYNAMIC MODEL
Provides temporal variation of 3-0 fields of primary
meteorological variables
• Horizontal Mind components
• Vertical motion
• Temperature
• Pressure
• Water vapor content
• Precipitation
11
-------
Figure 2.1
Components of Recommended Acid Deposition Modeling (ADM) System
(Continued)
DYNAMIC MODEL PROCESSOR
Uses primary meteorological variables from dynamic model
to estimate meteorological data required by ADM
• Time- and space-filtered primary variables
• Fraction of cloud cover
• Cloud liquid water content
• Radiation
• Height of planetary boundary layer
• Intensity of vertical mixing
• 3-D trajectories
ACID DEPOSITION MODEL
Produces temporal variation of 3-D fields of chemical species
and acid deposition. Solves conservation equation for
N species and M reactions:
at
- w
aw'Q'
az az
+ Sources
+ Sinks
(time rate of change of species
Q at a fixed point)
(horizontal transport by resolvable
and subgrid-scale flow)
(vertical transport by resolvable
and subgrid-scale flow)
(emissions plus reactions)
(deposition plus reactions)
12
-------
Figure 2.1
Components of Recommended Acid Deposition Modeling (ADM) System
(Continued)
ADM PROCESSOR
Uses variables from ADM to prepare estimates of total
add deposition (wet and dry) over region
on time scales of interest (episodes,
monthly, seasonal, annual)
13
-------
testing) of the meteorological model with the goal of determining the in-
herent uncertainties associated with each component, as well as with the
entire system. Application of the complete modeling system in decision-
making must recognize the stochastic (uncertain) nature of any simulation
or prediction, and determination of the stochastic component introduced by
the meteorology is necessary before application of the model to test alter-
native control strategies. Therefore, the required research on the meteor-
ological component of the ADM system consists of two parts: (1) model de-
velopment and (2) model validation.
2.1.1 Development of Meteorological Components
Three major components of regional numerical meteorological models can
be identified. Computational aspects include the design of the horizontal
and vertical grids and the accuracy of the numerical approximation to the
differential equations. Physical aspects include the parameterization of
important energy sources or sinks, such as radiation, condensation, and
evaporation, and frictional dissipation. Finally, a meteorological model
requires initial data and methods of analysis and initialization. These
three components are discussed in considerable detail in (NCAR, 1983) in
Section 4 of Chapter III and in Chapter IV, so a further review is not
required here. Instead, we recommend a specific combination of existing
model components for the first-generation ADM system. We emphasize that
all of the components proposed here can be added in a modular way, and can
therefore be replaced by alternate components. In addition, specific de-
tails of the model, such as location of the domain, number of layers, or
horizontal resolution, are reasonable estimates only; the system should be
designed so that these features can be varied easily for future testing or
application. The specifications given here correspond to a medium-resolu-
tion model that can be exercised in preliminary tests; it is likely that
higher-resolution versions will be required for some applications.
The prototype model recommended for initial development and testing
includes 15 layers, a 41 x 41 horizontal grid with a resolution of 80 km,
and covers the domain illustrated in Figure 2.2. The relatively coarse
grid will allow extensive testing of the model with real data in order to
obtain estimates of the model's accuracy and the uncertainties associated
with each simulation. Following these preliminary tests, subsequent tests
of the model with horizontal resolution of order 20-40 km are recommended
if the computer power is available.
The parameterization of planetary boundary layer (PBL) processes
should be developed within the framework of a medium-resolution (- five
layers in the lowest kilometer) PBL model in which the vertical fluxes of
heat, moisture, and momentum are calculated directly. Not only are such
models very general, treating both stable and convective situations and the
transition between these states, they are also suitable for providing di-
rect estimates for the ADM of the intensity of vertical mixing, the height
of the PBL, and the vertical profiles of wind speed, temperature, and hu-
midity. The PBL formulation must be coupled with a surface energy budget
and include a diurnal cycle.
The parameterization of cumulus clouds can be simple at first, and
14
-------
Figure 2.2
Recommended Domain for Preliminary Development and Testing
of the ADM System
15
-------
based on empirically-determined vertical profiles of heating and evapora-
tion. The total convective precipitation can be related to the total mois-
ture convergence. An alternative, which could be available in the model as
an option, is to neglect the parameterization of convective clouds and in-
stead resolve condensation and evaporation explicitly. The effect of lay-
ered nonprecipitating clouds can be considered in the surface energy budget
by relating the amount of these clouds to the layer-mean relative humidity.
In addition to developing the analysis and initialization routines and
the dynamic model itself, considerable effort will have to be directed to-
ward the development of a processor, which will utilize the basic meteoro-
logical data to derive estimates of the parameters needed in the ADM. The
basic and derived data are listed in Table 2.1. While calculation of some
of the derived parameters (such as time- and space-filtered horizontal and
vertical velocities) is straightforward, others, such as cloud liquid water
content, will require research into ways of parameterizing them in terms of
the basic variables.
2.1.2 Model Validation
As discussed in Section 8.2 of Chapter VI of (NCAR, 1983), there are
two phases to the recommended validation strategy. The first phase is to
evaluate each component separately, under simplified conditions, to esti-
mate the uncertainty associated with that component. The uncertainty of a
particular component can be estimated by varying the input data and physi-
cal parameters. From comparative numerical experiments, statistical mea-
sures such as the variance and bias can be calculated and statistical tests
performed to estimate the significance of varying each model component.
The validation of the complete meteorological model should be done us-
ing various conventional and nonconventional methods of skill discussed in
Section 4.2 of Chapter III of (NCAR, 1983). The validation should be done
separately for separate synoptic weather types, as outlined in Section 8.2
of Chapter VI of (NCAR, 1983). The initial effort should be concentrated
on those synoptic types that contribute most to annual acid deposition
(e.g., Niemann et al. (1979)). For the preliminary validation, we suggest
three distinct synoptic types: (1) winter storm precipitation events, (2)
summer convective precipitation events, and (3) summer fair-weather stag-
nation events. In the winter storm type, precipitation is expected to be
widespread and associated with frontal lifting. In the summer convective
precipitation type, the precipitation is expected to be locally heavy,
smaller in scale than the winter case, and associated with weak frontal
systems or no fronts at all. The summer stagnation case is expected to be
associated with light winds and little or no precipitation.
During the initial effort, we recommend the study of five cases be-
longing to each of the first two synoptic types above and two cases belong-
ing to the third type, for a total of twelve cases (Table 2.2A). For each
case, a number (approximately 12) of 48-h forecasts should be run in which
several important components of the model are varied in a systematic way
(Table 2.2B). The statistical aspects of each regional forecast (mean,
variance, etc.) can be computed from the 12 forecasts, and the uncertainty
associated with each synoptic type estimated from these statistics. In
16
-------
Table 2.1 Basic Variables Predicted by the Dynamic Model
and Derived Variables for Input into the ADM
Basic Variables
Horizontal wind components
Vertical wind component
Pressure
Temperature
Water vapor
Precipitation amount
Surface fluxes of heat,
moisture, and momentum
Derived Variables
Filtered (in space and time)
wind components - for transport
Turbulence intensity
3-D trajectories
Cloud type, amount, and depth
Liquid water content
Radiation fluxes
Height of planetary boundary layer
Relative humidity
17
-------
Table 2.2 Summary of Meteorological Model Experiments
Recommended During Validation Phase of the Project
A. Three Synoptic Types:
(1) Winter storm precipitation (5 cases)
(2) Summer convective precipitation (5 cases)
(3) Summer fair-weather, stagnation situation (2 cases)
Total number of cases: 12
B. Variations of Meteorological Model Forecasts:
(1) PBL parameterization
a. Medium resolution, explicit
b. Bulk aerodynamic formulation
(2) Cumulus parameterization
a. Function of total moisture convergence
b. None (explicit calculation of condensation, precipitation)
(3) Surface processes
a. No fluxes of heat or moisture
b. Fluxes computed according to surface energy budget
(4) Radiation (longwave and shortwave)
a. None
b. Interactive with layered clouds
(5) Initialization
a. Analysis, no initialization
b. Analysis, nonlinear normal mode initialization
(6) Comparison of two independently developed models
18
-------
addition, when the accuracy of forecasts is compared, according to the
pairs of forecasts (Table 2.2B), statistically-meaningful statements can be
made concerning the impact of varying the individual components. For exam-
ple, the above set of forecasts will yield 12 forecasts with two different
PBL formulations, 12 forecasts with two different initializations, etc. To
our knowledge, such organized sensitivity tests with complete diagnostics
(measures of skill) have not been conducted. The result will be a well-
documented summary of the accuracy and uncertainty associated with the
meteorological component of the ADM system.
2.2 The Chemistry System
2.2.1 The Chemistry-Transport Module
The overall design of the recommended chemistry-transport module can
be understood from the simplified diagram given in Figure 2.3. The module
receives, as input for each grid square and each of the 15 atmospheric lay-
ers above it, data on air transport velocities, extent of cloud cover, the
cloud depth, liquid water content, air temperature, pressure, and relative
humidity, amount of precipitation and its form (liquid, solid), and the
nature of the surface coverage. There are six different functions involved
in the operation of the module itself (Figure 2.3). Each of these requires
input data from other modules. The ultimate output of the chemistry-trans-
port module is the amount of wet and dry deposition of the acidic species,
including H+ ions.
We may consider briefly the flow of information to and from this mod-
ule with reference to Figure 2.3. The first of the chemical modules (1)
receives initial chemical species data, emissions data, and meteorological
data from three modules. The initialization data and emissions data mod-
ules are keyed to respond uniquely to time of the day, day of the week, the
season, and the location of the given grid scale of interest. Selected me-
teorological data related to cloud cover, liquid water content, depth of
clouds, and the temperature also feed into the chemical module (1) and are
required to estimate the distribution of reactants and products between the
gas phase and the liquid phase of the cloud water or rain water. In module
(1), this information is combined to calculate the initial distribution
(liquid and gas phase) of all species at any given time and position in
space.
Three other modules select required rate data: (2) chemical rate con-
stants for all reactions for the given input of temperature, pressure, and
humidity; (3) the photochemical rate constants for the given input of tem-
perature, time of the day, day of the year, cloud cover, and altitude; and
(4) dry deposition rate data (Chapter VI, Section 5 of (NCAR, 1983)) for
the given input of temperature, type of ground cover (including moisture on
surface), time of the day, and cloud cover. These modules feed the concen-
tration calculation modules (5a) and (5b).
Other meteorological data required for the transport and precipitation
calculations at each grid square and atmospheric level, the wind fields
(wind velocity components), and the precipitation extent and form (liquid
or solid), enter module (5). A subgrid-scale module generates a suitable
19
-------
Figure 2.3
The Chemistry-Transport Module
Initialization Data
for Chemical Species
Meteorological Data:
Wind fields, turbulent mixing
rates, cloud cover, radiation,
temperature, humidity, liquid
water content, type of ground
covers, precipitation
(1) Partitioning of reactants and products
(2) Selection of thermal reaction rate constants
(3) Selection of photochemical rate constants
(4) Selection of dry deposition rates for each species
Acid Deposition Model (ADM):
Solution of transport-transformation equations
(5a) Concentrations of reactants and products in gas phase
(5b) Concentrations of reactants and products in liquid phase,
droplet size distributions and numbers
Amount of wet deposition
and concentrations:
H+, HS03-, SO/,
MO,-, NH/, Ca**,
Mg , others
Subgrid-scale
corrections
Amount of dry deposition
of HN03, S02, HjSO,,
aerosol (NH,,HSO,J,
CH20, etc.
20
-------
correction function which also provides input to module (5) to allow for
the chemical heterogeneity of certain air masses in which point sources of
S02 and NOX are poorly mixed with the bulk of the ambient air containing
reactive hydrocarbons, aldehydes, etc. With these combined data, the mod-
ule (5) solves the coupled differential equations for the transport-trans-
formation and provides concentrations of the reactants and products as a
function of time. The concentration of the transient species with very
short lifetimes (less than about 1 sec), e.g., HO, H02, and R02 radicals,
is calculated directly from the steady-state relation for the given species
at every chosen time interval on the order of minutes.
The output of the chemical transformation-transport module provides
the amount of dry deposition of HN03, H2SOit, aerosol (NH^HSO^), S02, N02,
CH20, etc. and the amount of wet deposition of H+, HS03~, SQ^=, N03~,
NHi/, Ca++, Nig"*"1", and possibly other species of special interest.
The recommended chemical reaction schemes are discussed in detail in
Chapter V and in Chapter VI, Section 3, of (NCAR, 1983). The final choice
of mechanism will depend upon the results of a preliminary study (first six
months) of possible simplified reaction mechanisms based upon the Atkinson
et al. (1982) and the Killus and Whitten (1982) gas phase schemes. In the
most unfavorable case, gas phase and solution phase reactions required for
the chemical module may be about 100 in number, but we expect that some
significantly lower number will provide an adequate accuracy. For each of
the 15 atmospheric layers above each of the grid squares, our system design
requires the storage of the concentration of at least 20 different gas
phase species at each of selected time intervals: S02, NO, N02, 03, H202,
HN03, RH, CH20, CH3CHO, N205, N03, CH3C002N02 (PAN), NH3, H02N02, H2S, CO,
HO, H02, CH3C002, R02, together with aerosol components: H2SOi+, NH4HS0lt,
NHi+NO^ inorganic metal ion-containing species MN(II) and Fe(III), and
graphitic carbon; there are also at least 15 different liquid phase spe-
cies, the concentrations of which must be carried in the memory: S(IV),
S(VI), H+, N03-, NHi/, H202, 03, CH20, HOCH2S03H, H02N02, HO, H02, R02,
Mn(II), Fe(III), and possibly other species, as well as the liquid droplet
size distribution.
The output of the chemistry-transport module includes the integral of
the amounts of each acidic material (H2SOH, HN03, H+, NHttHSOH, S02, N02,
HS03~) deposited within each grid square for each of the many simulations
designed to duplicate the varied meteorological conditions encountered in
the eastern United States.
2.2.2 Initial and Boundary Conditions
Initial distributions and boundary conditions of key gases are likely
to have significant impact on the outputs of the acid deposition model.
Studies of complete chemical systems involving many species are required to
provide estimates of the quantitative impact of variations in the initial
concentrations of each species on the final solutions over time periods of
a few hours to a few days. It is likely that the distributions of species
with lifetimes longer than a day will be affected significantly by their
initial values and boundary conditions. Species included in this category
are S02, S0^=, NOX, HN03, N03-, 03, CO, NH3, H202, aldehydes, and most
21
-------
of the volative organic carbons (VOC). Short-lived species such as 0(ID),
HO, H02, RO, R02, N03> N20s, PAN, and some other organic radicals are not
directly sensitive to initial and boundary conditions because their distri-
butions reach photochemical equilibrium very quickly. However, the short-
lived species are affected indirectly by initial and boundary conditions
because the concentrations of short-lived species are functions of long-
lived species. The results of the quantitative studies mentioned above
will help determine what existing data bases are satisfactory for testing
the ADM system, as well as provide information concerning necessary mea-
surements in future field programs.
It is clear that the initial conditions for short-lived species can be
specified simply at their photochemical equilibrium values once the initial
values of long-lived species are determined. For long-lived species, the
ideal initial conditions would be the observed values, just as in meteor-
ological models. In reality, this is not possible, because most of the
long-lived species are usually not measured in rural areas. Some species
such as HN03, NO^-, NH3, H202, aldehydes, and specific hydrocarbons are not
even measured routinely in urban centers. Therefore, the initial values of
long-lived species must be specified by a combination of extrapolation from
limited data and self-consistent model estimates. This is clearly an area
where new insights based on analyses with the model would be a necessary
bootstrap procedure for the proper initialization. The performance of this
procedure can and should be evaluated through additional field measurement
programs as will be discussed in Section 3 of this report.
For long-lived species, there is no satisfactory way to specify bound-
ary conditions other than by using measured values, which are not available
for most gases. The uncertainty introduced by specification of these spe-
cies on the lateral boundaries must be a part of the research program. Our
strategy would be to set the boundary as far away from the region of inter-
est as the computation cost will allow so that the impact of inaccurately-
specified boundary conditions can be minimized.
2.2.3 Emissions
An emissions inventory for a regional acid deposition model provides
information needed to investigate major questions on sources of acid depo-
sition. Emissions of the key aerosols and gases involved in acid formation
must be considered. Although S02 and NOX are the major acid precursors,
several other species significantly influence acid formation. These in-
clude VOC and CO, because of their roles as buffering agents in the oxi-
dation process; NH3, because of its role as a buffering agent in cloud
chemistry; and several reduced sulfur species, because of their potential
importance as naturally-emitted precursors to acid rain. In addition,
primary sulfate emissions could have significant impact on local areas.
It has been concluded (e.g., Galloway and Whelpdale, 1980a,b; Semb,
1978) that anthropogenic emissions of S02 exceed natural emissions of gas-
eous sulfur compounds by at least a factor of ten in eastern North America
and Europe. Although this is probably correct, there is large uncertainty
in the estimates of natural gaseous sulfur sources. The natural source of
NOX is probably dominated by lightning and soil biogenic activities, al-
22
-------
though some downward flux of NOX as well as 03 from the stratosphere is
expected. Quantitative estimates of their source strengths are available,
but large uncertainties remain. It is currently believed that anthropo-
genic NOX emissions exceed natural emissions by a large factor in eastern
North America. Natural sources of VOC, NH3, and CO are also poorly known.
Because of their uncertainties, the natural sources of these species should
not be included in the first phase of the ADM system development. Under
the National Acid Precipitation Assessment Plan (NAPAP), the National Oce-
anic and Atmospheric Administration (NOAA) has been assessing the natural
sources of gaseous sulfur, NOX, and NH3; these results should be incorpo-
rated when they are available.
The major man-made emissions data bases are described in Section 1 of
Chapter VI of (NCAR, 1983). Reasonably comprehensive data are available
for SOX, NOX, VOC, and CO. There is no NH3 emissions inventory for
North America, although a state-by-state survey is now being prepared (R.
C. Harriss, private communication; R. Husar, private communication). EPA
is currently sponsoring Brookhaven National Laboratory (BNL) for compila-
tion of source emissions inventories for acid deposition modeling research.
In the first phase of the ADM system development, we suggest including
only anthropogenic emissions of SOX, NOX, VOC, CO, and NH3. The emis-
sions data will need to be subdivided into the model grid. Large point
sources, however, need to be considered individually because of their im-
pact on chemical phenomena on a smaller scale (Lamb, 1982). Subgrid non-
linear effects due to point sources should be evaluated. Temporal varia-
tions (such as diurnal and weekday-weekend patterns) of emissions and their
effects on the transport and transformation of key species should a>so be
evaluated.
3. SYSTEM INTEGRATION AND VALIDATION
Figure 3.1, a flow diagram of the proposed acid deposition modeling
system, illustrates the flow of information from initiation of the calcu-
lation to application of the results. The meteorological data, meteorolo-
gical model and processor, chemical species emissions data, surface condi-
tions data, chemical initiation, and the core acid deposition model have
been discussed in the previous two sections of this report. We shall dis-
cuss here the overall structure of the model system and the post-processor
and sensitivity analysis components.
As shown by Figures 2.1, 2.3, and 3.1, we are recommending an inte-
grated modeling system with quite independent submodels. Given this sys-
tem, one can modify, update, and interchange any and all essential physical
component descriptions without undue reprogramming. Such an elaborate
structure is costly in computation time and storage, but, as the first com-
prehensive model of its type and with the clearly expressed desire by EPA
to incorporate new findings in a timely fashion, we feel this modular
structure is appropriate. When sufficient experience is obtained with this
system and its components, its developers should look to integration of
subsystems to achieve increased computational and storage efficiency.
Past experience with other large models at various national laboratories
suggests that most large computer codes can be speeded up by factors of 2
23
-------
c
o
to
O
(U
to
O
Q.
E
O
•i- (O
c
c <.
O O
«rt O «/>
O i-
a. Q- a»
t-H C
• s: o
00
3 O) 3
cms a.
•r- C E
U_ OJ O
E O
O T3
o c
O) (d
a:
la
o
a>
T3
O
(O
I
o
(/) -t-
o +J
•r- (O
CO 4->
s.
o
0)
o
ai
T5
o
to
o
a.
O 3 O -r-
i- O) O -r- +J
O C tO
a. o a.
o
•r- U
CO
-------
to 5 depending on the machine and the nature of the code. This effort can
be initiated near the end of the third year while validation and model an-
alysis are going on.
The post-processor is essentially a user-oriented interactive graphics
package for post-analysis of the ADM system. Past experience tells us that
the information flow from three-dimensional models can be simply overwhelm-
ing. Computer-assisted analysis is the only logical and feasible solu-
tion. Furthermore, one must recognize that not all future users of this
ADM system will have or desire to have direct access to the supercomputer
where this modeling system must reside. A post-processor designed for
smaller computers (minis or even supermicros) can nevertheless yield the
end products (e.g., species concentration contours, deposition patterns,
scenarios). The user should be able to reorganize the data interactively
and carry out simple analyses for scientific and regulatory purposes. This
"user-friendly" interface has not been traditional in scientific modeling
efforts. Recent advances in micro- and mini-computer systems have demon-
strated the great benefits of such an approach to system applications,
however. A recent example of such an application is the menu-driven
user-interactive chemical kinetics model at NCAR.
Sensitivity analysis is an essential component of the ADM system. As
discussed in (NCAR, 1983), there are several techniques in the literature,
but none has been applied to acid deposition modeling. Two possible ap-
proaches are direct parametric studies and Monte-Carlo-type statistical
analysis; they yield different types of information, and both should be
used. In direct parametric studies, individual or groups of input vari-
ables are modified according to data uncertainties and the model outputs
are compared to the reference case. These controlled numerical experiments
give very direct cause-effect results, but it is very difficult to use this
technique to cover the full range of possibilities. As an alternative, one
can use the statistical cases suggested in Section 2.1.2 as the statistical
sample and carry out Monte-Carlo-type uncertainty analysis (Stolarski et
al., 1978) on the input variables. This gives only a partial answer, how-
ever, since such a technique does not point to specific causal relations.
Therefore, a judicious application of both techniques is most desirable;
this has been shown to be of value in studying problems of stratospheric
ozone perturbations, for example.
Validation studies of theoretical models (in particular, computational
models) are frequently talked about and proposed but rarely completely car-
ried out. Complex computational models are built precisely because the
physical processes to be studied are much too complex for linear analysis.
In order to test the accuracy of the model, one must use the ADM system to
carry out a detailed analysis of a controlled scenario and then measure the
predicted physical quantities. Unfortunately, to our knowledge, this has
never been done; field experiments usually are designed in the absence of a
comprehensive model. Therefore, questions on the number of variables to be
measured, frequency of data collection, geographic coverage, instrument
sensitivity, and meteorological conditions are often independently ad-
dressed, causing incomplete testing of the model.
We recommend that, beginning in the third year of the project,
25
-------
planning efforts toward testing the ADM system be initiated using the model
system itself as the principal tool for analysis. Although many components
of the system would by that time have been tested in the laboratory or in
the field, the overall balance and performance of the ADM system still
requires testing to add another level of confidence in applications.
TabTe 3.1 summarizes the status of various components of the proposed
ADM system. Many of the meteorological components are well developed but
reside in a few operational models. In order to capitalize on the ongoing
efforts, this project should seek direct collaboration with one of these
efforts. Many components of the proposed acid deposition model are being
developed by research groups funded by EPA and others. Close monitoring of
these efforts should be established (see following section) to avoid dupli-
cation of effort and to facilitate technology transfers. Other components,
notably the trace species transport-transformation submodel, the gas-phase
and liquid-phase chemistry, and the post-process components, all require
dedicated new efforts. As was explained in Section 2.Z.I, although several
schemes on gas-phase chemistry have been proposed in the literature, they
need to be evaluated for suitable implementation in this system. In many
ways, this may prove to be more difficult than constructing the original
schemes.
4. MANAGEMENT PLAN
4.1 Manpower Needs
The manpower required to build the proposed ADM system will depend on
how the effort is partitioned between the prime contractor (i.e., the con-
tractor responsible for the system framework) and the various EPA contrac-
tors who will be building specific modules for delivery to the prime con-
tractor. Some of the modules could perhaps most readily be built by the
prime contractor, depending on its assembled talents and expertise, while
others might more appropriately be completed by EPA contractors already
well along on the relevant research.
Nonetheless, we have made some broad estimates of the manpower needed
to assemble the proposed ADM system. In so doing, we have assumed the fol-
lowing time table:
June 1, 1983: Project to begin.
Jan. 1, 1985: Progress report on the first generation model.
Jan. 1, 1987: Final report and documentation of the acid deposition
model.
During the first year of the project, it will be necessary to incorpo-
rate immediately all the "well-developed" (first column, Table 3.1) modules
of the proposed ADM system. This includes, most conspicuously, adapting a
contemporary, dynamic mesoscale model and its operating packages, such as
that described by Anthes and Warner (1978). These well-developed modules
would be used in the first-generation system. During the first year, it
will also be necessary to begin implementing the modules for which research
26
-------
Table 3.1
Status of Components of Proposed ADM System
Well developed; Research New
requires imple- in pro- efforts
mentation gress required
Dynamic Model
Meteorological data X
Objective analysis X
Initialization X
Numerical methods X
Boundary conditions X
Surface physics X
Planetary boundary layer X
Clouds and precipitation X
Radiation X
Processor X
Validation X
Synoptic climatology X
Acid Deposition Model
Emissions X
Surface characteristics X
Chemical (species) initialization X
Transport X
Cloud submodel X
Subgrid processes (point sources, plumes) X
Wet deposition X
Dry deposition X
Gas phase chemistry X
Liquid phase chemistry X
Radi ati on X
Heterogeneous processes X
Post-Processor
Post processor X
Sensitivity analysis X
Validation X
Statistical scenario X
27
-------
is in progress (second column). Finally, it will be important to begin
work on those modules which require new efforts (third column).
Our estimates of the first-year manpower requirements are (FTE =
full-time equivalent):
Incorporate existing modules 2.0 FTE
Complete research in progress
Dynamical model 2.5 FTE
Acid deposition model 6.5 FTE 9.0 FTE
Begin new research
Dynamical model 1.0 FTE
Acid deposition model 1.0 FTE
Post-processor 1.0 FTE 3.0 FTE
Estimated total first-year effort 14.0 FTE
Depending on the existing technical expertise of the prime contractor
staff and their preferred scheduling plans, some of the above tasks may
have a delayed start. For example, components under development by other
research organizations obviously must be included near the completion of
those projects. We expect that this initial distribution would be adjusted
in later years of the project. However, our past experience with similar
projects indicates an even level of manpower requirements over the duration
of the project.
4.2 Internal Management Structure
We find the management structure that we adopted during the present
project to be appropriate also for building the ADM system, if supplemented
with proper interfacing with other EPA contractors and subcontractors. The
overall management structure is shown in Figure 4.1. The project director
is aided by a Steering Committee which includes senior staff in meteorolo-
gy, chemistry, and model systems. The director (who should be experienced
in building large systems) oversees three groups representing the three key
disciplines. Each of these groups is headed by a senior scientist in that
specialty. Finally, these three groups draw on the institutional staff,
outside consultants, EPA staff, and EPA contractors.
The project director should be the overall scientific leader and man-
ager of the project and the official representative of the project to EPA,
EPA contractors, and other cooperating institutions. The project Steering
Committee should assist him in the scientific review and planning proces-
ses. Particular attention should be devoted to maintaining technical bal-
ance and quality. Furthermore, members of the Steering Committee should
serve as the senior project liaison with external groups as indicated in
28
-------
Figure 4.1
Proposed Management Structure for
ADM System Development
EPA Staff
EPA Contractors
Meteorology
Group
Senior Dynamic
Modeler,
Staff
Project Office
Director,
Staff
Chemi stry
Group
Senior Tropospheric
Chemist,
Staff
Steering Committee
Senior
Scientists
System Integration
and Validation Group
Senior Model
System Developer,
Staff
Staff, Consultants
29
-------
Section 4.3. The subgroups should be supervised by their respective senior
members with specific details designed to match the stages of program matu-
rity and levels of interactions desired by members of individual groups.
It is also recommended that the principal participants be housed together
in a centrally-located facility so as to encourage natural interdiscipli-
nary interactions and cooperation.
A detailed work schedule should be prepared within two months of the
start of the project. This would allow time to assess the progress and
plans of each of the groups preparing submodules for the ADM system.
Certain in-house studies should proceed during the first year of the
project. For example, the chemistry group should be selecting the minimal
set of reactions which will duplicate with reasonable precision the impor-
tant reactants for S02 and NOX oxidation to acids. This work is neces-
sary for the development of the initial version of the model system.
Similarly, early development of many meteorological submodules is also
recommended.
4.3 External Interactions
The anticipated acid deposition modeling program can provide a focus
for EPA-funded theoretical efforts in this field and provide important
guidance for related field measurements. As such, it must be structured
such that new and significant results from other researchers can also be
easily included in a timely fashion. Conversely, any findings of the
broadly-based research team building the ADM system should also be commu-
nicated to others. The modular structure of the proposed ADM system will
allow easy integration of new results, but regular and effective communi-
cation between the model framework contractor and the model development
contractors is a prerequisite. Accordingly, external communications
deserve attention.
First, a set of working groups is needed to cover the major technical
areas. A tentative list is: dynamic modeling; clouds, radiation, and pre-
cipitation; transport-transformation modeling; gas and liquid phase chemis-
try; heterogeneous processes and deposition; and sensitivity analysis and
data bases. The chairman of each working group should be a principal re-
searcher in the subject field. Each group would have about five working
scientists in the subject area as members. Each working group would be
paired with at least one member of the ADM Steering Committee—i.e, the
working groups would be monitored by the overall project director and by
the Steering Committee. This structure would assure continuous transfer of
information. As the project evolves, some of these working groups may be
reorganized, while others may be phased out and new ones added.
Another important role for the disciplinary working groups would be to
provide liaison with those external contractors who are developing ADM sys-
tem modules under EPA contracts. For the overall acid deposition model de-
velopment to succeed and to meet the desired schedule, a sequence of events
must occur according to plan and no hiatuses can be permitted. For exam-
ple, whoever is responsible for an individual model module must deliver it
on schedule, but punctuality alone is not enough. The contractor respon-
30
-------
sible for model system framework development must communicate with each
module development contractor before module delivery to assure compatible
goals, formats, and levels of sophistication. Situations in which a module
is delivered late or delivered with no prior communication between the par-
ties must be avoided. The goals and contents of each module must be agreed
upon or at least commonly understood well before module delivery. To this
end, one or more disciplinary working group members, Steering Committee
members, and/or the project director should be assigned to be a liaison be-
tween the ADM system contractor and the contractor developing each module
for the system.
To communicate with an even broader audience and to assimilate emerging
scientific knowledge, an annual modeling workshop should be held. Findings
of the working groups would be presented at this workshop for general peer
review. Because this workshop would be open to interested modelers and
other EPA-supported scientists, there would be opportunity for cross-
fertilization among modelers, experimentalists, and regulators. As stated
in (NCAR, 1983), we believe the Eulerian framework as proposed can best
meet the needs of the scientific community in general and EPA in particu-
lar, but much has been learned and will continue to be learned from other
types of models. Further interaction between modeler and experimentalist
is essential for the advancement of our understanding of acid deposition
processes. A model-oriented workshop would bring out details of successes
and failures of individual efforts, which are often lacking in the usual
publications and presentations.
Clearly, close collaboration with the academic community is another
essential component of external interactions. Much new information and
insight has been brought to the acid deposition modeling field by the more
discipline-oriented academic communities. Depending on the nature of the
organization which hosts the modeling project, long-term close collabora-
tive efforts with universities clearly can be and should be established.
"Hands-on" cooperation has proven to be an indispensable component of
multi-disciplinary research and development projects of this complexity. A
flexible management structure should be maintained such that it is possible
to invite visitors and send staff to participate actively in collaborative
research and development projects. These arrangements can be made on short
notice, as soon as the technical needs are apparent, and can last for ex-
tended periods. This flexibility will work to the advantage of the pro-
ject, and allow direct collaboration with others when their modules are
being implemented into the model system, for example.
This project must maintain frequent informal contacts and exchanges
with EPA staff on all levels. Each year, a mid-year informal oral report
on all aspects of the project—technical, financial, and managerial—should
be arranged. A formal written report should be prepared and delivered to
EPA within one month after the ending of each project year. Technical re-
ports and computer program documentations should be sent on a timely ba-
sis. Additional technical support for the EPA project office should also
be provided, within the resources of the project.
31
-------
4.4 Facilities for Model System Development
Essential facilities for developing a model system of the necessary
scope are identified as follows. The contractor must have modern digital
computer systems with adequate speed and peripheral services, or at least
have guaranteed access to such facilities. While computational power
(i.e., computer central processor speed and large memory) is essential,
input-output devices and facilities for tape handling, etc. are also
needed. It is probably also necessary to have midsized computers for
processing of input data and for post-processing of model-output fields.
Library facilities and clerical/editorial services must also be con-
sidered. Without these, the contractor would be involved in a constant
struggle to barely keep pace with scientific developments and with report
preparation.
Last, but not least, the host institution should be sufficiently large
and prestigious so as to attract and keep top senior scientists as well as
bright young staff. For a project of this scope and complexity, a well-
balanced mix of experience, dedication, and leadership is the only path to
success.
5. REFERENCES
Anthes, R. A., and T. T. Warner, 1978: Development of hydrodynamic models
suitable for air pollution and other mesometeorological studies.
Mon. Wea. Rev.. 106, 1045-1078.
Atkinson, R., A. C. Lloyd, and L. Winges, 1982: An updated chemical mech-
anism for hydrocarbon/NOx/S02 photooxidations suitable for inclusion
in atmospheric simulation models. Atmos. Environ., 16, 1341-1355.
Galloway, J. N., and D. M. Whelpdale, 1980a: An atmospheric sulfur budget
for eastern North America. Atmos. Environ., 14, 349-362.
Galloway, J. N., and D. M. Whelpdale, 1980b: An atmospheric sulfur budget
for eastern North America. Atmos. Environ., 14, 409-417.
Killus, J. P., and G. Z. Whitten, 1982: A new carbon-bond mechanism.
Final report, EPA Contract No. 68-02-3281, Systems Applications, Inc.,
San Rafael, California.
Lamb, R. G., 1982: A regional scale (1000 km) model of photochemical air
pollution. Part I: Theoretical formulation. Office of Research and
Development, Environmental Sciences Research Laboratory, U.S. Environ-
mental Protection Agency, Research Triangle Park, N. C. 27711, 227 pp.
(in press).
NCAR, 1983: Regional Acid Deposition: Models and Physical Processes.
Prepared under Interagency Agreement No. AD49F2A203 for the Environ-
mental Sciences Research Laboratory, U.S. Environmental Protection
Agency, Research Triangle Park, N. C. 27711, June 1983.
32
-------
Niemann, B. L., A. A. Hirata, and L. F. Smith, 1979: Application of a re-
gional transport model to the simulation of multi-scale sulfate epi-
sodes over the eastern United States and Canada. Presented at the WMO
Symposium on the Long-Range Transport of Pollutants and Its Relation
to General Circulation Including Stratospheric/Tropospheric Exchange
Processes, 1-5 October 1979, Sofia, Bulgaria.
Semb, A., 1978:
455-460.
Sulphur emissions in Europe. Atmos. Environ., 12,
Stolarski, R. S., D. M. Butler, and R. D. Rundel, 1978: Uncertainty pro-
pagation in a stratospheric model. 2. Monte Carlo analysis of impre-
cisions due to reaction rates. J. Geophys. Res., 83, 3074-3080.
33
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
. REPORT NO.
2.
3. RECIPIENT'S ACCESSION-NO.
. TITLE AND SUBTITLE
5. REPORT DATE
REGIONAL ACID DEPOSITION: DESIGN AND MANAGEMENT
PLAN FOR A COMPREHENSIVE MODELING SYSTEM
6. PERFORMING ORGANIZATION CODE
. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
The NCAR Acid Deposition Modeling Pro.iect
I. PERFORMING ORGANIZATION NAME AND ADDRESS
National Center for Atmospheric Research
P. 0. Box 3000
Boulder, Colorado 80307
1O. PROGRAM ELEMENT NO.
CCVN1A/01 Task 2295 (FY-84)
11. CONTRACT/GRANT NO.
Interagency Agreement No.
AD49F2A203
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory - RTP, NC
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park. NC 27711 "
13. TYPE OF REPORT AND PERIOD COVERED
Final - 7/1/82-5/31/83
14. SPONSORING AGENCY CODE
EPA/600/09
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This is a companion report to REGIONAL ACID DEPOSITION: MODELS AND PHYSICAL PROCES-
SES. This report presents a design and management plan for the development of an
Eulerian regional acid deposition model. It first reviews the fundamental physical
processes relevant to regional acid deposition and then describes the proposed model
system. The principal components (meteorology and chemistry) are discussed in some
detail with special emphasis on model initialization and subsystem validation. The
total system integration and full validation are presented separately. The manage-
ment plan section focuses on internal structure, external interactions, and special
facility needs. Strongly managed interdisciplinary interactions and intensive
"hands-on" working groups for external interactions are suggested.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
13. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
119. SECURITY CLASS (Tins Report)
UNCLASSIFIED
21.
- OF PAGES
20 SECURITY CLASS (This page/
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-7-,
-------
------- |