This morning I heard a lecture by Rick Anthes, president emeritus of UCAR, former director of NCAR. His talk was entitled “Butterflies and Demons,” and the subject was predictability of weather and climate. He was a witness to, and participant in the development of numerical weather prediction in the form it exists today at weather centers worldwide. It was a particularly interesting and provocative talk.
Numerical weather prediction proved its worth in the forecasts of the track and severity of Hurricane Sandy. Without the forecasts, the property damage and loss of life would have been much worse than it was. One might compare the effect of Sandy to the Galveston flood of 1900 for which there was no warning and thousands of people lost their lives. Sandy was the only hurricane in history that made landfall on the Atlantic coast from the east. Dr. Anthes showed a slide with the tracks of every Atlantic coast hurricane since 1850. Most tracked up the coast, and those that went east into the Atlantic did not return. One might reasonably question the reliability of data extending back to the age of sail, but no statistical method based on previous experience could possibly have predicted that a hurricane would go northeast from the coast and then return westward to make landfall since it had never happened before.
At this point it is important to note that the national weather centers make more than a single numerical forecast. In addition to a main central forecast, they make a collection of forecasts, numbering in the hundreds at some weather centers, each differing slightly in some respect, usually in the initial conditions. They refer to such collections of simultaneous forecasts as “ensembles.” The spread among the ensemble members is expected to reflect uncertainty in the forecast. Dr. Anthes showed the ensemble produced by the European Center for Medium Range Weather Forecasting (ECMWF) of predicted tracks for Sandy. Nearly all of them exhibited the correct behavior. Perhaps five of the several hundred tracks predicted by the ensemble members led out into the Atlantic and did not return.
Dr. Anthes said that accurate forecasts such as the ones issued by ECMWF for the track of Sandy would have been impossible 20 years ago. He emphasized the fact that advances in science, in the form of improved numerical technique, data assimilation and understanding of rain and clouds, along with spacecraft as well as earthbound instruments and data processing techniques may well have saved thousands of lives and billions of dollars in property damage.
It was certainly good to see benefits to society that come from my corner of the world of scientific research. It’s the received wisdom in the world of hurricane forecasting that predictions of tracks have improved considerably over the years, while improvement in prediction of intensity has been much slower. Dr. Anthes’ graph showing improvement of skill in forecasting of hurricane tracks since 1980 didn’t strike me as being quite so impressive as other aspects of weather forecasting. If I read the graph correctly, the accuracy of present two-day storm track forecasts is about equivalent to the accuracy of one-day storm track predictions in 1980. By contrast, the graph shown by Dr. Anthes of global weather forecast skill showed that our 5-day forecasts today are as accurate as our 2-day forecasts were in 1995.
Dr. Anthes gave three talks at Oregon State during his visit, and this was the only one I was able to attend. The talk I heard was billed as the most technical of the three, and per his introduction, it wasn’t nearly as technical as the usual seminars in the series given by the Physics of Oceans and Atmospheres group. There were no equations, but much scientific insight and lots to think about.
Robert Miller
College of Earth, Ocean, and Atmospheric Sciences
Oregon State University
miller@coas.oregonstate.edu