AUTOMOTIVE

The model gives scientists a way to quantify the uncertainty that surrounds the climatic phenomenon known as El Niño, which triggers severe weather changes in North and South America and Australia and endangers crops and wildlife, said Noel Cressie, professor of statistics and director of the university's Program in Spatial Statistics and Environmental Sciences. A statistical model created at Ohio State University predicts sea surface temperatures in the tropical Pacific Ocean, and gives scientists a way to quantify the uncertainty that surrounds the climatic phenomenon known as El Niño. The top image shows actual sea surface temperatures for November 2002, while the bottom image shows the temperatures predicted by the model. The color red indicates above average temperatures. While there are many methods for forecasting El Niño, this model is unique because it includes detailed measures of uncertainty, explained Cressie. Such measures are important for quantifying risk when, for example, farmers and insurance companies make decisions about the next harvest. "This is a Bayesian statistical model -- it represents a new paradigm in geophysical modeling, where all sources of uncertainties are accounted for in a melding of geophysical knowledge and statistical description," Cressie said. In this case, the strength of westerly winds in a particular location of the tropical Pacific is a key component. Although the research team discovered this relationship through their own exploratory spatial data analysis, this experimental finding fits with one dominant scientific paradigm that westerly wind bursts are an important factor in the pooling of warm surface waters in the eastern Pacific. Peruvians named the warming El Niño, or "The Christ Child," because it typically arrives in December during the years it occurs. Drawing on decades of data for wind, air pressure, and sea surface temperature, the model can forecast El Niño six months in advance -- long enough for farmers and commercial fishermen to plan for the coming season. The model provides a range of probable sea surface temperatures in each forecast. Over the last five years, only once did the actual temperature not fit within the model's predicted range. For the first time, a tool based on the model is available on the Web (http://www.stat.ohio-state.edu/~sses/collab_enso.php), so scientists and the public can view animations of El Niño forecasts from January 1985 through May 2004, and compare the forecasts to observed temperatures in the tropical Pacific. Normally, temperatures of surface waters in the western Pacific are 6 to 8 degrees Celsius (10 to 15 degrees Fahrenheit) warmer than in the east. But during an El Niño, the temperature differential reverses. The nutrient-poor warm water forces the fish that normally thrive off the west coast of South America to go elsewhere to find food. Birds that would feed on the fish die off, and the local fishing economy suffers. El Niño causes far-reaching weather events as well, including drought and heatwaves across Australia, torrential rainfall in Central and South America, and heavy winter snows and floods in the southern United States -- all of which affect water resources and food supply. Scientists don't fully understand the factors that cause an El Niño, and models of such large environmental systems are very complicated. But it's in just such a situation -- where there is a considerable amount of uncertainty involved -- that a statistical model that accounts for physical understanding can do very well, Cressie explained. The Web site where the model's results are presented grew out of a paper that Cressie and his colleagues -- Mark Berliner, professor of statistics at Ohio State, and Christopher K. Wikle, associate professor of statistics at the University of Missouri -- had previously published in the Journal of Climate in 2000.