EPA/600/R-99/030

Chapter 2

MODELS-3 ARCHITECTURE: A UNIFYING FRAMEWORK FOR ENVIRONMENTAL

MODELING AND ASSESSMENT

Joan Novak* and Sharon Leduc**

Atmospheric Modeling Division
National Exposure Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711

ABSTRACT

The Models-3 framework and Community Multiscale Air Quality (CMAQ) model were designed
and built to function together. The CMAQ science is documented in the other chapters and this
chapter is intended to provide science and technical details of the Models-3 framework which
integrates the CMAQ. The Models-3 User Manual is available and useful for those who want
operational details beyond the “big picture” provided in Chapter 2.

This chapter present the various components of the Models-3 framework (Dataset Manager,
Program Manager, Science Manager, Study Planner, Strategy Manager, Model Builder, Source
Code Manager, Framework Administrator, and Tool Manager). The Tool Manager provides
access to third party applications for emissions processing, visualization, and analysis. The
Framework Administrator provides the system administrator access to system lists and
administrative functions in Models-3. Each of the other components is provided for developing,
executing, and managing CMAQ applications.

The computer architecture, the client-server design, the object-oriented data base management
system, and the graphical user interface are also addressed in this chapter, along with the design
features of Models-3.

*
On assignment from the National Oceanic and Atmospheric Administration, U.S. Department of Commerce.
Corresponding author address: Joan Novak, MD-80, Research Triangle Park, NC 27711. Email:
[email protected]
**
On assignment from the National Oceanic and Atmospheric Administration, U.S. Department of Commerce.
EPA/600/R-99/030

2.0 MODELS-3 ARCHITECTURE: A UNIFYING FRAMEWORK FOR

ENVIRONMENTAL MODELING AND ASSESSMENT

2.1 Introduction

Models-3 is a flexible software system designed to simplify the development and use of
environmental assessment and decision support tools for a wide range of applications from
regulatory and policy analysis to understanding the interactions of atmospheric chemistry and
physics. The initial version of Models-3 contains a Community Multiscale Air Quality (CMAQ)
(Byun, et al., 1997, and other chapters of this document) modeling system with capabilities for
urban to regional scale air quality simulation of tropospheric ozone, acid deposition, visibility
and fine particulate. The Models-3 framework provides an interface between the user and
operational models, between the scientist and developing models, and between ever changing
hardware and software platforms.

The concept of an integrated modeling and analysis framework (Novak, et al., 1995; Dennis, et
al., 1996) was formulated in response to 1) problems and duplication of effort involved in
maintaining the separate data pre-processing and post-processing software systems required for
each model used to analyze different air quality pollutants or scales, 2) the difficulties States and
industry encountered using more complex modeling systems, and 3) the large expense required to
modify existing models to incorporate scientific advancements or to adapt the models to new
domains. Based on technology advances in the National High Performance Computing and
Communications program, Models-3 was designed to overcome these limitations and to serve as
a community framework for continual advancement and use of environmental assessment tools.
Capabilities and design features of the Models-3 framework provide more flexibility in data
handling, modeling, visualization and analysis.

2.2 Overview of the Models-3 Framework

The Models-3 framework contains components that assist 1) the model user with air quality
modeling studies and analysis of results, and 2) the model developer with creating, testing, and
performing comparative analysis of new versions of air quality models. The major design goal is
to simplify and integrate the development and use of complex environmental models, beginning
with air quality and deposition models. The function of each framework component is described
below.

2.2.1 Dataset Manager

Dataset Manager provides the user with the capability to register files for use with the modeling
and analysis programs within the Models-3 framework. The registration process involves entering
the location of the dataset (full path name) and metadata (information about the data such as
spatial - temporal extent and resolution, source of data, time convention, units, etc.) into the
Models-3 data base. Models-3 follows the Federal Geospatial Metadata Standard for metadata
content. The datasets may be located on any network-connected computer system known to the

2-1
 EPA/600/R-99/030

Models-3 system installed at the user’s site. Once a dataset is registered, the user can search for
the dataset based on its metadata information, file type, etc.

Dataset Manager allows the user to view the details of the selected dataset to ensure the correct
one has been selected for use with an application. Dataset registration eliminates the need for
the user to type the entire path name each time the dataset is used. Instead the user can highlight
the dataset from a list of candidates that satisfy the search criteria specified by the user.
Models-3 will automatically move selected data to the host where it is needed for a model
execution. Dataset Manager also provides standard capabilities such as deleting, copying,
archiving, and restoring files and metadata.

2.2.2 Program Manager

Program Manager allows the user to register, update, and search for executable programs and/or
scripts to make them available for use in defining studies within the Study Planner component.
During program registration, the user enters characteristics of the program into the framework
including descriptive information on program function, input requirements, output specifications,
runtime environment variables, target architecture and operating system. Once the program or
script is registered, this executable can be used in Study Planner to sequence a series of
executions which may depend on previous executions for input data. The user can access and
execute programs that are not registered. However, registering programs ensures that all
mandatory inputs have been specified and automatic naming and registration of output files is
activated, to facilitate tracking output from numerous program executions. Recommended model
configurations for standard domains are preregistered in the system, eliminating the need for the
typical user to deal with the details of program registration.

2.2.3 Study Planner

Study Planner allows the user to define a study and control the execution of its associated
processes. A study is a collection of plans and properties that are necessary to describe and
perform one or more environmental modeling analyses. A plan is a collection of information
defining dataset and program interdependencies and the sequence of execution of one or more
programs. Study Planner gathers much of its information from the Program Manager and Dataset
Manager registration data. The relationship between a program (node) and its required and
optional datasets (links) is user-defined through the process of constructing and annotating a
graphical diagram with simple drag-and-click mouse operations. Once a plan is constructed and
its graphical diagram fully annotated with desired input datasets and options, the plan can be
executed. User specified program options are entered by editing program environment variables.

Studies and associated plans are named entities that are saved in the system data base. Therefore,
a typical user can start with an existing study plan provided by the model developer and simply
change the dataset annotations by selecting, through a file browser, appropriate datasets needed
for execution. The Study Planner provides capabilities to create new studies, copy and modify
existing studies, and delete existing studies.

2-2
EPA/600/R-99/030

2.2.4 Strategy Manager

Strategy Manager provides the capability to estimate point-, area-, and mobile-source emissions
for future years and to determine the relative effectiveness of specified control scenarios. The
user may adjust pollutant growth factors and emissions control factors to perform "what if"
analyses for entire EPA regions, states, counties or for user-defined study areas. By applying
estimated yearly emission growth factors from the Emissions Growth and Assessment System,
control efficiency, rule effectiveness, and rule penetration factors to the EPA 1990 base year
emissions inventory, the Strategy Manager provides estimates of future year (1991 - 2010)
emissions for carbon monoxide, nitrogen oxide, particulate matter up to 10 microns, sulfur
dioxide, and volatile organic compounds. Strategy Manager is based on EPA’s Multiple
Projection System. An input data processor is included to process the National Emissions Trend
(NET) inventory (URL http://www.epa.gov/oar/oaqps/) data format.

2.2.5 Tool Manager

Tool Manager provides access to third-party applications (tools) that are registered with the
Models-3 framework. Third-party applications that are currently available include Models-3
Emission Processing and Projection System (MEPPS), Vis5D, Text Editor, Package for Analysis
and Visualization of Environmental Data (PAVE), SAS®, Arc/Info®, IBM Visualization Data
Explorer™, DXDriver, and VisDriver. Tool Manager allows the user to add tools to the system
that will help with the user’s work. A file converter is included to convert between ASCII, SAS,
and I/O API files.

2.2.5.1 Emissions Processing

The majority of the emissions processing needed to support air quality model applications can be
prepared through minor modifications to pre-specified emissions plans in the Study Planner.
However, for those users who need total flexibility in specifying details for emissions
preparation, the Models-3 Emissions and Projection System (MEPPS) (Benjey and Moghari,
1996) is available for interactive processing of emissions. MEPPS provides capabilities to input
and perform quality control on emission inventory data, and reformat and subset data for the
user-specified modeling domain. The main emissions processing is performed by a significantly
revised version of the Geocoded Emission Modeling and Projection (GEMAP) (Wilkinson and
Emigh, 1994) system which requires the user to have Arc/Info® and SAS® licenses for
operation. This Emissions Processor (EMPRO) transforms the annual county based emissions
inventories into spatial and temporal resolutions that are consistent with the target model
application, i.e. typically hourly gridded emissions for a selected domain. The processor also
estimates mobile and biogenic emissions and performs chemical speciation consistent with the
chemical mechanism in the target model. Mobile emission factors are calculated using Mobile
5a; biogenic emissions are calculated using the Biogenic Emissions Inventory System (BEIS2).
Finally, according to a user-specified criteria, an output processor merges the emissions files to
provide the 2-D gridded files and a point source stack emissions file required for the model
execution. The user is offered numerous choices for interactively changing parameters and

2-3
 EPA/600/R-99/030

reviewing results in all of the emissions components. The global science specifications created by
the user in Science Manager are accessible from within MEPPS to ensure consistency when
desired.

2.2.5.2 Visualization and Analysis Tools

Several freely available visualization tools are accessible from the Models-3 framework as well
as commercial visualization and statistics packages, such as IBM Visualization Data Explorer™
and the SAS®. A graphical user interface component, VisDriver, provides an interface for
finding and selecting Models-3 data files on local or remote hosts and launches visualization
applications with Vis5D or PAVE. VisDriver performs requisite data conversions from the
internal Models-3 data format to the data format required by these third party visualization
packages. VisDriver also provides a data export capability to such formats as Advanced
Visualization System (AVS™), Flow Analysis Software Toolkit (FAST™), and an ASCII
spreadsheet. On-line help is available to assist the user with these visualization tools
(http://www.epa.gov/asmdnerl/models3/vistutor/user/user-guide-00.htm/).

PAVE

The Package for Analysis and Visualization for Environmental data (PAVE), developed at the
North Carolina Supercomputing Center of MCNC, allows the user to visualize 2-D multi-variate,
gridded environmental data with smooth tile plots,3-D mesh plots, time series line and bar plots,
vertical cross sections, wind vectors, and scatter plots. More information can be found at
http://www.iceis.mcnc.org/EDSS/pave_doc/Pave.html.

Vis5D

Vis5D, a public domain package developed at the University of Wisconsin Space Science and
Engineering Center, provides interactive visualization of large five-dimensional gridded datasets
- the data are real numbers at each point of a grid that spans three space dimensions, one time
dimension, and a dimension for enumerating multiple physical variables. Vis5D provides
isosurfaces, contour line slices, colored slices, volume rendering of data in 3D, and rotation and
animation of the 3D image in real time. There is also a feature for wind/trajectory tracing. More
information can be found at http://www.ssec.wisc.edu/~billh/vis.html.

IBM Visualization Data Explorer

IBM Visualization Data Explorer™, a commercial visualization package, has been used to create
custom visualizations to support diagnostic evaluation of the air quality models with visualization
of multiple/nested grids, terrain following grids, grid cell time aggregate statistics such as
maximum ozone concentration and hours of non-compliance, comparison of measured aircraft
data with model predictions along the path of flight, and detailed chemical analysis of the
integrated reaction rates predicted over space and time in the air quality models. Other standard

2-4
EPA/600/R-99/030

visualization capabilities for scalar and vector quantities, 2-D and 3-D graphs have also been
integrated through the DXDriver interface.

2.2.6 Science Manager

Critical model specifications typically found hardwired in air quality model codes, such as details
of the horizontal coordinate system, specification of horizontal grid dimensions, vertical layers,
chemical mechanism specification (reactions and rate constants), etc. are treated as globally
shared information in the Models-3 framework. Use of this global information enables more
consistency throughout the system (i.e. emission, meteorological, and chemistry-transport
components). Detailed specifications for these key science components are entered by the user
only once using Science Manager graphical user interfaces. These specifications are saved as
named entities in an object-oriented database accessible by all model components. A typical user
would access and modify previously defined Science Manager components to define a new
modeling domain.

The functionality of Science Manager, however, is targeted to the model developer to facilitate
experimentation with new model components. For example, extensive research is done on
improving photochemical mechanism specifications. One component of Science Manager
enables the model developer to either edit an existing chemical mechanism or import a new set
of chemical reactions, specify chemical species, molecular weights, etc. Both the Regional Acid
Deposition Model, Version 2 (RADM-2) and the Carbon Bond-4 mechanisms are available in
the Models-3/CMAQ framework and the State Air Pollution Research Center (SAPRC) chemical
mechanism is being modified to fit within this paradigm to facilitate comparative studies with
these chemical mechanisms. As long as the chemical species in any modified mechanisms are
present in source emission profiles associated with Source Classification Codes (SCC), the
chemical mechanism specification will propagate to the emission processing subsystem to enable
generation of emission species consistent with the newly defined chemical mechanism.

2.2.7 Model Builder

The Models-3 framework facilitates the interchange of science components within a model,
customization of the chemistry mechanism, and changes in the modeling domain, horizontal grid
resolution, and vertical layering without the need for reprogramming. Modifying an existing
model or building a new model is a two-step process. First, the model developer must use
Science Manager to specify the horizontal coordinate system, horizontal grid dimensions, vertical
layers, chemical mechanism, and select science modules for each class (i.e. advection, diffusion ,
cloud process, etc.) of science process to be included in the new model. Module selections are
contained in a named configuration file. After the needed science components are specified, the
model developer uses Model Builder to select from available science specifications and
configuration files to create a single custom model executable.

2-5
 EPA/600/R-99/030

2.2.8 Source Code Manager

CMAQ source code is distributed with the Models-3 framework to provide the user flexibility to
create a model for a user specified domain or to allow a user to build a new executable with a
selected chemistry mechanism. Model Builder uses Source Code Manager to access appropriate
source code to transform the user’s selections into a working executable. Source Code Manager
allows a user to retrieve a version of a source code file, modify it, and return it to the code
archive after the change has been tested. Source code should only be modified by knowledgeable
model developers. Source Code Manager is based on Concurrent Versions System
(CVS)/Revision Control System (RCS) public domain software for code configuration
management and tracks the history and version numbers for all source code changes. These
version numbers become part of the history information associated with each output file.

2.2.9 Framework Administrator

The Framework Administrator subsystem menu allows the Models-3 system administrator to
manipulate system lists and to access administrative functions of other Models-3 subsystems.
System lists include users, hosts, device types, site IDs, file format types, compiler names,
operating system names, platform names, and time zone names. Through this subsystem, only the
authorized system administrator can set access controls for users, add or delete entries from the
system lists, and maintain the integrity of controlled files.

2.3 Models-3 System Architecture

The Models-3 framework is designed as a three-tier client-server architecture with an object-

oriented data base management system (OODBMS) ,ObjectStore™, to store persistent data (i.e.
studies, horizontal grid definitions, metadata containing pointers to physical dataset locations ,
etc.). When the user brings up the Models-3 graphical user interface on user’s desktop system,
the user becomes a "client" requesting services from a "server". Several Models-3 servers act as
"clients" requesting services from other Models-3 "servers". The client and server can be on the
same machine or on separate machines giving the system scalability by either adding more client
workstations with only slight impact on performance, or moving to faster and larger server
machines. Each of the subsystem functions described in Section 2.2 is implemented as a server
and can be can run on different machines. The Models-3 servers were initially developed on a
Sun workstation using the Solaris 2.5.1 operating system (Models-3 Version 2.1). Models-3 is
being ported to a Sun Solaris 2.6 operating system (Models-3 Version 2.2). The next server to be
released for Models-3 will be SGI (late 1998). Models-3 Version 3 (summer 1999) will allow
server to be a PC with a Windows NT® operating system.

This client-server architecture in conjunction with commercial cross-platform communications

interface - ORBIX™ by Iona Technologies, Ltd. which is compliant with the Common Object
Request Broker Architecture (CORBA) standard - and an OODBMS containing pointers to
datasets distributed over the network enables transparent use of multiple computing platforms
and access to data across the network. Figure 2-1 illustrates a possible configuration for multiple

2-6
EPA/600/R-99/030

installations for Models-3 software. Each independent site must have at least one server to
support the local OODBMS and Models-3 sessions. Each site can run autonomously. A two-
server configuration is recommended at each installation: one to support emissions processing
activities, and the second to support other client-server and model computations. Of course,
additional servers will improve performance for larger applications and the current version of
Models-3 also supports batch submissions of model executions to a supercomputer with job
status reporting back to the framework through the standard Network Queuing Service (NQS).
In the future, we envision maintaining a master database to foster sharing of data and to minimize
duplication of effort.

State Offices HPC

Internet
 Loc al DB
Desk top
PC(s)
PC N T Server
UN IX Se rve r(s)
Centr al DB

Public Access Master

Metadata
& G lobal Info.
Object Orien ted DBMS
Federal A gencies
EPA
 Loc al DB

         
Desk top Local DB Desk top
Workstatio ns UN IX Se rve rs
PC(s) PC(s)
PC N T Server
UN IX Se rve rs

Figure 2-1. Seamless Computing and Data Management

from PC to Scalable Parallel Computer

2-7
 EPA/600/R-99/030

To adapt to ever changing hardware and software, the framework uses a layered interface
design that isolates critical system components, thus minimizing the impact of hardware and
software upgrades. Each layer, as seen in Figure 2, is connected to its adjacent layer(s) by a well
specified interface thus easing the task of replacement if a more advanced component becomes
available. Another key design choice is the use of an input/output application programming
interface (I/O API) for the data access layer. This concept is implemented by using a standard
FORTRAN or C callable library for all input/output in the air quality model codes. Thus I/O
efficiencies can be improved by replacing this centralized library or new data types can be
handled by adding new routines to the library.

The I/O API is layered on top of the netCDF standard data file format developed at the National
Center for Atmospheric Research. This data format uses the underlying eXternal Data
Representation (XDR), which is IEEE (Institute of Electrical and Electronics Engineers, Inc.)
compliant and, therefore, enables cross-platform transfer of data without conversion. The
Models-3 implementation conforms to the netCDF specification, but sets additional conventions
for information stored in the file header. NetCDF is a self documenting file format that contains
complete specifications (parameter names, units, etc) for the file contents. All netCDF records
can be accessed via direct access methods to minimize data access times. When registering
conforming programs, the Models-3 framework automatically writes execution history
information to the file header record of each output file and to the OODBMS metadata
associated with the output file. The history information, which establishes traceability of output
datasets to the model that generates them, contains the date/time of model execution, the version
numbers of the science process modules contained in the model executable, Science Manager
specification names, host computing environment, and other pertinent information. Science
Manager’s global shared data objects and the netCDF format are the key aspects of building
“plug compatible” science modules that conform to Models-3 guidelines.

US E R IN T ER F A CE

Ma n a ge m e nt L a y er:
D a ta M a n ag e r Sc ie n ce M a na g e r
US E R P ro gra m Ma n a ge r Mo d e l B u ilde r
IN T E RF A C E S tu dy M a na g e r So u rce C o de M a na g e r
S trate g y Ma n a ge r Fra m ew o rk A d m inis tra tor
T o ol M a n ag e r
E n v iron m e nt L a y er: O S, S y ste m “P e rs on a lity ”

C o m pu ta tion a l L ay e r: Pro g ram s : m o d e ls, a n a ly s is , vis u a liza tio n,...

D a ta Ac c e ss L a y er: I/O A pp lic a tio n P ro g ra m m in g In te rfac e

D a ta Stru c tu re /R ep re s en ta tion : ne tC DF , X D R

D a ta Sto ra ge : File s ys te m s & d a ta b a s es

P h y s ic a l D ev ic e La y e r: Dis k s , n e tw ork s , prin te rs , m a ch in e s

Figure 2-2. Architectural Layering: Flexibility for Future Change

2-8
EPA/600/R-99/030

2.4. Schedule and Future Plans

The Models-3 framework provides a unifying foundation for continual community evolvement of
environmental modeling and assessment tools with possible extensions beyond the current air
quality implementation. The use of the Models-3 framework for multimedia exposure and risk
assessment is possible. A simple integration of the Hydrologic Simulation Program FORTRAN
(HSPF) and the Chesapeake Bay Water Quality Model (CBWQM) in the Models-3 framework
explored multimedia model linkage capabilities. Maintaining Models-3 framework to use
emerging computing capabilities and transferring that capability to science researchers and
environmental decision makers will continue to be the focus of the Models-3 effort.

2.5 References

Benjey, W.G., and N.M. Moghari. Functionality of an integrated emission preprocessing system
for air quality modeling: the Models-3 emission processor. In The Emissions Inventory:
Programs & Progress. VIP-56. Proceedings of a Specialty Conference, Research Triangle Park,
NC, October 11-13, 1995. US Environmental Protection Agency, Research Triangle Park, NC,
and Air & Waste Management Association, Pittsburgh, 463-471 (1996).

Byun, D., J. Young, J. Gipson, J. Godowitch, F. Binkowski, S. Roselle, B. Benjey, J. Pleim, J.

Ching, J. Novak, C. Coats, T. Odman, A. Hanna, K. Alapaty, R. Mathur, J. McHenry, U.
Shankar, S. Fine, A. Xiu, and C. Jang. Description of the Models-3 Community Multiscale Air
Quality (CMAQ) model. Proceedings of the American Meterological Society 78th Annual
Meeting, January 11-16, 1998, Phoenix, AZ.

Dennis, R.L., D.W. Byun, J.H. Novak, K.L. Galluppi, C.J. Coats, M.A. Vouk. The Next
Generation of Integrated Air Quality Modeling: EPA's Models-3. Atmospheric Environment, Vol.
30, No.12, pp. 1925 - 1938, (1996).

Emission Inventory Improvement Program. URL http://www.epa.gov/oar/oaqps/eiip/

Novak, J.H., R.L. Dennis, D.W. Byun, J.E. Pleim, K.J. Galluppi, C.J. Coats, S. Chall, M.A. Vouk.
EPA Third-generation Air Quality Modeling System, Volume 1: Concept. EPA 600/R95/084,
National Exposure Research Laboratory, Research Triangle Park, NC, 55 pp. (1995).

Wilkinson, J.G., and Emigh, R.A. The Geocoded Emissions Modeling and Projections System
(GEMAP): Advanced Training Workshop. Prepared for Environmental Protection Agency,
Office of Research and Development, by Alpine Geophysics, Boulder, CO (1994).

This chapter is taken from Science Algorithms of the EPA Models-3 Community
Multiscale Air Quality (CMAQ) Modeling System, edited by D. W. Byun and J. K. S.
Ching, 1999.

2-9