CAMx Training for CAPIA Introduction to CAMx

1
CAMx Training for CAPIAIntroduction to CAMx

• Greg Yarwood Ph.D.
• ENVIRON
• Novato, California 94945
• gyarwood_at_environcorp.com
• April 15-19, 2002

2
Outline

• Introduction
• Ozone photochemistry
• Photochemical grid model concept
• Where is CAMx used?
• CAMx features for regional modeling
• Computer requirements and run times

3
CAMx Overview

• 3-D Photochemical Grid Model
• emissions, chemistry, dispersion, removal of
  gaseous and aerosol air pollution
• geographic scales from individual point sources
  (lt 1 km) to regional (gt1000 km)
• State-of-the-Science
• new coding (1995)
• computationally and memory efficient
• easy to use
• modular framework, many options available
• publicly available (www.camx.com)

4
Sources and Sinks of Tropospheric Ozone

• Sources
• smog chemistry involving VOCs and NOx
• global methane and CO oxidation
• stratosphere
• Sinks
• chemical reactions, e.g., NO, alkene titration
• deposition

5
Central Role of Hydroxyl Radical (OH)
6
Photochemical Grid Model Concept
7
Ozone and Precursor Relationships EKMA Diagram
VOC Sensitive
NOx Sensitive
8
Grid Cell Processes
9
Coupling Between Grid Cells
10
CAMx 3-D Visualization
11
El Paso (EPA)Los Angeles (CRC)Central CA
(BAAQMD)Idaho (IDEQ)
Major U.S. CAMx Applications
OTAG (EPA)Kansas/St. Louis (KDHE/MDNR) Texas
(TNRCC)Florida (FLDEP)Midwest (LADCO)
12
CAMx Worldwide
Europe UK Greece France Spain Germany
United States
China
Japan
Taiwan
Mexico
Chile
Australia
South Africa
13
CAMx Grid Nesting

• Efficient regional modeling using coarse grid
• High resolution for key source/receptor areas
• Capabilities
• built-in two-way nesting
• information flows from fine to coarse and coarse
  to fine grids
• variable meshing factors (2, 3, 4..)
• multiple levels (e.g., 36/12/4/1.33 km grids)
• configured for up to 10 nested grids
• vertical nesting (no longer recommended)

14
(No Transcript)
15
CAMx Setup -- Domain Definition (continued)
16
CAMx Coordinate Systems

• Coordinate systems to match many met models
• Avoid interpolating met data
• Horizontal
• Lambert Conformal (MM5)
• Polar Stereographic (RAMS)
• UTM
• Geodetic latitude/longitude
• Vertical
• Physical height above ground
• time/space varying, can match pressure
  coordinates

17
Mass Consistency and Conservation

• CAMx formulated to conserve mass by solving
  continuity equation for mass, not mixing ratio
• UAM conserved mass by holding pressure constant
• RADM/SAQM models dont conserve mass
• Mass consistency
• ability to preserve the atmospheric density field
• test using unit initial/boundary conditions
• CAMx version 3.1 usually within 1
• very difficult, for many reasons

18
CAMx Chemistry Options

• Two main options for gas-phase chemical mechanism
• CB4 (Carbon Bond4)
• early 1990s
• widely supported, history of successful
  application
• lumped structure approach
• SAPRC99
• late 1990s
• more detailed and modern than CB4
• limited support and application
• VOC chemistry much more reactive than CB4
• run times 60 longer

19
CAMx 3.1 Mechanism Options

• Mechanism 3 is usually used

20
CAMx Plume-in-Grid

• Plume-in-Grid (PiG) concept developed to avoid
  rapid dispersion of point source emissions in
  coarse grids
• CAMx PiG
• Puff model embedded within CAMx
• Treats initial dispersion and chemistry for
  selected large NOx point sources
• Most useful within coarse regional modeling grids
• You select which point sources should use PiG
• PiG requires compromises to allow two different
  models to work together

21
(No Transcript)
22
Other Features in CAMx Version 3.1

• Parallel Processing
• shared memory architecture (e.g., dual-processor
  PCs)
• Flexi-nesting
• add/remove fine grids at model restarts
• interpolate information from parent grids
• Probing Tools
• OSAT ozone source attribution
• DDM sensitivity analysis
• Process Analysis diagnostic tools

23
CAMx Flexi-Nesting

• Ability to add/delete nested-grids during a
  simulation
• model spin-up
• episode focus during longer runs
• Provide as much or little fine grid information
  as you have available
• Any fine-grid inputs that are unavailable are
  interpolated from the next coarser grid
  automatically
• Example application
• Study grid resolution around major point source
• Compare with sub-grid scale plume algorithms

24
OTAG Test Case Domain
25
CAMx Run Times

• OTAG test case
• nested grid modeling for Eastern US and Canada
• 36-km coarse grid is 64 x 63 x 5
• 12-km fine grid is 137 x 110 x 7
• 126,000 grid cells
• Linux PC with 1 GHz Pentium III
• 40 minutes per episode day
• 200 MB disk per episode day
• 85 MB memory needed

26
CAMx Run Times - Impact of Nested Grids

• OTAG test case coarse grid only
• 36-km coarse grid is 64 x 63 x 5
• 21,000 grid cells
• Linux PC with 1 GHz Pentium III
• 5 minutes per episode day
• 90 MB disk per episode day
• 25 MB memory needed
• Nested grids have strong impact on run time
• fine grids have large number of cells
• fine grids take shorter timesteps

27
CAMx Parallel Processing

• Uses Open-MP (OMP) compiler directives
• Shared memory architectures
• Implemented for chemistry and transport
  algorithms
• Tested on
• dual-processor Linux box with Portland Group
  compiler
• SGI Origin 16-processor server (Weining Zhao,
  TNRCC)
• Sun 4-processor server with KAI guide (Winston
  Hao, NYSDEC) - Sun compiler did not support OMP
• Identical results for single and multi-processors

28
CAMx Run Times with Multi-Processors
29
CAMx - Speedup per CPU
30
CAMx Multi-Processor Performance

• Speedup of about 1.6 with dual processors
• Scales well for 4 processors
• When to use?
• Annual runs
• Large grids
• Need the answer NOW!

31
ENVIRON Linux Workstation

• Dual Athlon MP 1800, 1 GB DDR-RAM
• Tyan motherboard with onboard SCSI
• 7200 rpm IDE disk for OS
• RedHat 7.1 or 7.2
• 4 x 120 GB 5400 rpm disks onATA100 controller
• lower speed cooler, more reliable
• several failures of 7200 rpm data drives
• Sony DDS4 tape backup using onboard SCSI