IBM
Skip to main content
 
Search IBM Research
     Home  |  Products & services  |  Support & downloads  |  My account
 Select a country
 IBM Research Home
Weather Modelling
Deep Thunder
 ·Details
 ·Results and Applications
 ·Frequently Asked Questions
 ·What the Press Says
Weather Data Visualization

Contact Us
More Information
 Tropical Weather Forecasting
 Optimization and benchmarking of weather codes
 Collaborative research with universities, government labs and industry
 
 


IBM Research
  Deep Thunder
IBM SP-2 for Weather Forecasting at the 1996 Olympic Games

Zaphiris D. Christidis 
IBM Thomas J. Watson Research Center 
Yorktown Heights, NY 10598 
zaphiri@watson.ibm.com

Introduction
The overall objective of this project is to provide accurate weather forecasts for four main venues at the 1996 Olympic Games in Atlanta. These forecasts will be made available to the media, the Olympic organizing committee, participants, and spectators via dedicated kiosks. In addition, this information will also be available on the IBM Olympic Games World Wide Web site.

In this effort, IBM is partnering with two governmental agencies: the Forecasting Systems Laboratory (FSL) in Boulder, CO, which is part of the National Oceanographic and Atmospheric Admininstration (NOAA); and the National Center for Environmental Prediction (NCEP), which is operated by the National Weather Service. IBM Research is heavily involved in the development of the numerical weather modeling code which will be run on an IBM RISC System/6000 Scalable POWERParallel (SP-2) computer in Atlanta, as well as in the development of advanced visualization capabilities to faciliate dissemination of the computed weather forecasts. IBM Worldwide Government Industry Solutions is providing overall coordination of the project.


The specific objective of the Olympic capability is to be able to produce forecasts of the weather over the next 6 to 12 hours for each of the four Atlanta venues. In particular, the goal is to complete a new forecast for two of the four sites every 3 hours; hence, each venue will have an updated 6-or 12-hour forecast every 6 hours.


To meet this very ambitious goal, we are working with FSL to develop a highly optimized regional weather forecasting model (known as RAMS for Regional Atmospheric Modeling System) that will fully exploit the parallel processing power of the IBM SP computer installed in Atlanta.


This SP machine has a total of 30 processors. A RISC System 6000/39H workstation with GXT1000-2 OpenGL graphics processor acts as 31st node of the SP via dedicated networking to enable rapid access to the model output for visualization and analysis.

The important technical aspect of this capability is that the RAMS regional model, which is responsible for the simulations of the weather at the Atlanta sites, will be coupled to a different global model which will be simulating the weather across a much larger geographical region, e.g. all of North America. This global weather model, known as Super-ETA, will be executed on a NWS machine in Eagan, MN, using a relatively coarse resolution (approximately 8 km between grid points in the numerical model) of the Earth's surface. The results of this global simulation will be periodically transmitted via T1 links to the SP2 in Atlanta, which will then use this data as input (specifically boundary conditions) to the subsequent execution on the SP2.


The SP-2 simulatons will use a much finer resolution (approximately 2 km) as well as more robust physical models in order to produce very accurate regional forecasts in the Atlanta area. The extensive computational power of the SP-2 will be needed to do such detailed simulations within the time constraints discussed above. In addition, the RAMS model will accept as observational data collected by the NWS in the Atlanta area.



zaphiri@watson.ibm.com

[ DX Home Page | Contact DX ]  



 
  
 

  

  About IBM  |  Privacy  |  Legal  |  Contact