NMSU branding

New Mexico State University

New Mexico State University

News Center

Physical Science Laboratory scientist presents artificial intelligence research

Donald Bustamante, a project manager at New Mexico State University's Physical Science Laboratory (PSL), presented two papers on artificial intelligence applications in weather forecasting at the Artificial Intelligence Research in Environmental Sciences '98 conference in Victoria, British Columbia.

Bustamante has developed an artificial intelligence-based atmospheric temperature profile retrieval system, which provides the temperatures of the atmosphere at various heights above ground. The data are then used as inputs to meteorological forecast models. His research is funded by the U.S. Army.

Traditionally, meteorologists have obtained this information with weather balloons, but economics and the need to obtain data from inaccessible areas have created a need for alternative methods. Data from weather satellites have been used to obtain the temperature profiles, but the current conventional procedure requires significant computational capabilities. Bustamante's artificial intelligence approach, based on neural networks, has resulted in a system which can be run more quickly and on smaller computers. Bustamante said the artificial intelligence approach provides results that are superior to conventional techniques.

Bustamante's methodology uses electronic neural networks that are based on the biological networks of the human brain. "It learns by example, which is different from the traditional computation approach in which you have to use logarithms," he said. "To use an analogy, if you wanted to teach a child what a truck is, you could use a lengthy conversation to describe it or you could go to a street corner and point to a truck, and show how it's different from a car."

Bustamante and his co-author, Robert A. Pearson of the University of New South Wales, Australian Defence Force Academy, presented two papers at the AIRIES '98 conference. A paper titled "Effects of Cloud Contamination on Neural Network Retrieved Temperature Profiles" described the performance of the system under cloudy conditions and methods of improving results. The second paper, "Improving Error Structure in Temperature Profile Retrievals from Satellite Observations," presented a variety of approaches to improving atmospheric temperature profiles.

Pearson recently completed a nine-month sabbatical from the University of New South Wales during which he worked with Bustamante at PSL researching the use of neural network applications in satellite meteorology.

Participation in AIRIES '98, held Oct. 21-23, 1998, was by invitation and limited to 60 participants. AIRIES is international and interdisciplinary, involving researchers from environmental sciences and artificial intelligence fields.