Cellular Engineering

Mice with Sickle Cell Genes

Online Health Survey

Smog NOx Detector

Cleaning Up Dirty Silicon

Priority Service for the Internet

Climate Modeling Tools for Weather Forecasting

Imaging Liquids

Magic Sizes and Microscopic Crucibles

Climate Modeling Tools Improve Weather Forecasting

In October 1997, just as El Niņo was getting ready to soak California with one of the wettest winters ever, nearly 150 climate modeling experts from around the world convened at Berkeley Lab to discuss advances in using computers to study, model and predict weather patterns.

At the meeting, hosted by the Lab's National Energy Research Scientific Computing Division (NERSC), scientists demonstrated that their tools for predicting weather are getting more reliable and extending the range of forecasts.

This image reveals the abnormally warm Pacific Ocean surface temperatures (shown in red) associated with El Niņo. Computer scientists at NERSC are working with the Geophysical Fluid Dynamics Laboratory to adapt climate modeling codes to run on more powerful massively parallel processor supercomputers to more accurately predict climate changes.

Still, the need for better modeling tools was also made clear. More accurate models lead to more precise forecasts, which in turn can help save lives and resources as people have more time to prepare for severe storms. One way of doing this is by adapting the best current tools to run on the most powerful computers available.

To do this, the Department of Energy's Biological and Environmental Research Program recently announced that it will fund a joint climate research project between NERSC and the Geophysical Fluid Dynamics Laboratory (GFDL) to investigate how widely used climate models can be run effectively and efficiently on massively parallel processing supercomputers. Such computers, called MPP machines, are able to handle large jobs more quickly by dividing codes into smaller units and running them in parallel, thereby taking minutes or hours to process data that once required days or weeks to run. Using such machines, climate researchers can also include more data points in their simulations, thereby producing more detailed and accurate results.

NERSC will focus initially on the Modular Ocean Model developed at the GFDL, and investigate additional models later. As improvements are made in the codes, NERSC will work closely with the model developers at GFDL to incorporate the improvements into the models. For more climate modeling visualizations visit the GFDL website.

Next: Revealing the Secrets of Wet Stuff

Research Review Fall '98 Index | Berkeley Lab