Annual Report
About CRC
Advanced Vehicle/Fuel/Lubricants
Atmospheric Impacts
Aviation
Emissions
Performance
Catalog of Reports (1942 to present)
Recent Requests for Proposal (RFP)
Joining a Potential Bidders List

Submitting Unsolicited Proposals
Partnering with CRC
Meeting Notices
CRC CRC
CRCAbout CRCCRCWorkshops & EventsCRCPublicationsCRCDoing Business with CRCCRCNewsCRCDialogCRCDialog CRC

SPEED ENHANCEMENTS (MPI) FOR CAMx PROBING TOOLS ON COMPUTER CLUSTERS

Project A-63

Leader:    A. M. Dunker

                  R. S. MacArthur

Scope and Objective

The objective of this project is to revise CAMx so that multiple computer processors can be used to increase the speed of simulations involving probing tools.

Current Status and Future Program 

Currently, the CAMx and CMAQ regional ozone/PM/visibility models are being applied to annual periods by states and several Regional Planning Organizations to address the regional haze requirements and PM2.5 and 8-hour ozone attainment issues.  Although there are some fundamental differences in their formulation, CMAQ and CAMx are both considered state-of-the science models.  CRC has funded the implementation of several probing tools in CAMx that extract additional information from the model to assist in understanding predictions and evaluating control strategies.  The three probing tools are the Decoupled Direct Method (DDM) for sensitivity analysis, Process Analysis (PA), and Source Apportionment (SA). 

With help from Sandia National Laboratory, EPA implemented the Message Passing Interface (MPI) capability in CMAQ so that the model could be run on computer clusters.  As a consequence, despite the fact that CAMx is more computationally efficient than CMAQ when run on a single computer, CMAQ simulations now require less time because the model can be run on multiple computer processors.  Thus, implementing MPI capability in CAMx is important to the continued use of the model.  This is especially true when using one or more probing tools for annual modeling, in which case a single simulation can take several months without MPI. 

The Midwest Ozone Group (MOG), a stakeholder group that consists of utilities, other industries and local agencies, provided a limited amount of initial funding to initiate the MPI implementation in CAMx.  The MOG seed funding has allowed the rewriting of the CAMx code into FORTRAN90, conversion of model data arrays to use dynamic allocation, and other code restructuring and modifications needed to allow the implementation of MPI.  The U.S. EPA Office of Air Quality Planning and Standards (OAQPS) provided additional funding of a comparable amount to implement the multi-processor capability into the core portion of CAMx that calculates pollutant concentrations. However, this funding is insufficient to implement MPI for the probing tools.

This project, built on the work funded by MOG and EPA, extended the multi-processor capability in CAMx to the probing tools. The modified code was tested for a regional simulation using each of the probing tools. The reduction in simulation time for the concentrations and the probing tools was determined, as compared to a simulation employing a single processor.  The advantages and disadvantages of using multi-processors was reported. Guidance was developed on the optimal number of processors to employ in simulations with probing tools.  The instructions for using multi-processors were documented in the User's Manual along with any required changes in input.

The committee approved the project plan and ENVIRON was selected to conduct the project beginning in June 2006.  Deliverables from the project include the revised CAMx code with multi-processor capability for the probing tools.  The User's Manual for CAMx will be updated to describe the new capability of the model and how best to use it.  The Final Report has been published and has been posted on the CRC website.

Back to the Atmospheric Impacts Index