Scientists who try to forecast weather a month or more in advance may be ready for a breakthrough in their often frustrating enterprise. A new breed of fast computers running improved mathematical models of the ocean-atmosphere system promises to put long-range forecasting on a more solid basis. Growing understanding of ''boundary'' factors like soil moisture or sea-surface temperature as being atmospheric regulators may bring more predictability to forecasts of average temperatures, or even average precipitation, in various regions a season or more in advance.
Thus, for the first time in several decades of demonstrating only very modest skill in making such forecasts, scientists in the field see the possibility of making significant improvements over the next five years.
This feeling of cautious hope dominated the Climate Prediction Workshop held last week at the American Meteorological Society in Boston. Here Richard Somerville of the Scripps Institution of Oceanography, which has one of three US Experimental Climate Forecast Centers, summarized the situation.
He explained that short-range forecasting of weather a few days ahead is solidly based on a mature technology of machine forecasting which, in turn, is based on known physical laws. This technology has a track record dating back to the 1950s, which lends credibility to the forecasts. Long-range forecasting, by contrast, is an infant technology. Up to now, it has had no good global mathematical models of the atmosphere with which to work and no good track record.
Somerville added, however, that there has been much progress recently. There is a feeling within the scientific community - although it is not yet a consensus - that experimental forecasters are on the verge of something significant. He said the field is maturing rapidly.
Workshop chairman Alan D. Hecht, director of the congressionally mandated National Climate Program Office, repeated this point by noting that ''we probably couldn't have had a meeting like this up to the past five years.'' He added that ''the past five years have been a period of major change'' in this branch of science. He said he expects the next five to bring even more significant change.
In saying this, Hecht and other forecasting scientists do not expect the kind of detailed computer-based predictions now issued for short periods. The theoretical limit for such deterministic forecasts is two to three weeks. What they do hope to do is improve the prediction of averages, such as average temperatures, over the long term.
There do seem to be long-term trends imposed on the atmosphere by Earth's surface conditions. These are imposed by patterns of ice and snow, soil moisture , and sea-surface temperatures, which change slowly and thus may make such features as long-term average temperatures more predictable.
Specialists are only beginning to take account of these boundary conditions in a realistic way in computer modeling. But experiments using such data were reported at a conference on the subject held at Liege, Belgium, in May. These enabled computer models of the global ocean-atmospheric system to reproduce natural long-term weather patterns more realistically. It is this possibility of using more realistic computer models with worldwide data on sea-surface temperatures, soil moisture, and other boundary factors which buoys forecasters' expectations.
Donald L. Gilman, chief of long-range forecasting with the US National Weather Service, reminded the workshop that, today, ''this kind of forecasting is better called a technical art than a science.'' He noted that, among other basic needs, forecasters lack well-defined weather structures with which to work on these longer time scales. Short-term weather structure - fronts, air masses, storm systems - is hard to transfer to long-term prediction. It is difficult to isolate a useful phenomenon - an event with a definable beginning, middle, and end. Yet a large part of the progress in forecasting may come from doing this. That is why the new sense of progress in this work is so encouraging.
So far two such events, definable on time scales of months and beyond, have been identified, according to John M. Wallace of the University of Washington, who heads the second of the three Experimental Climate Forecast Centers. (The third center is at the NASA-Goddard Space Flight Center in Maryland.)
One such event is the type of extended drought that goes on summer after summer. The United States has had only one of these in this century - the 1930s Dust Bowl. Such drought is coupled to soil moisture, which is the mechanism by which the atmosphere ''remembers'' the drought conditions from year to year. Thus soil moisture data could be a predictor in such situations.
The other type of event, which has stirred much scientific interest, is the so-called El Nino and Southern Oscillation (ENSO). The oscillation is a coherent pattern of pressure, sea-surface temperature, and rainfall fluctuations long known in the tropical Pacific Ocean. When periods of abnormally warm sea surface - an El Nino - then occur, weather is affected around the globe. In particular, an ENSO can influence winter weather over North America. It accounted for the abnormally warm 1982-83 North American winter.
Wallace said the connection seems reliable enough to use in forecasting, although it is no cinch. For example, it is a guide to average conditions over certain North American areas. For temperatures over the area from Seattle eastward to Minneapolis and for precipitation in the Southeast, he said he thinks useful predictions could be made in, say, September for the following January and February.
Although it would never allow prediction of individual hurricanes, such as Diana, the ENSO phenomenon, which has emerged as the dominant global climate signal on time scales of a few months to several years, may also help predict general characteristics of hurricane seasons well in advance. For example, Neil Frank, director of the US National Hurricane Center at Miami, forecast a mild Atlantic hurricane season for 1983 because of the '82-'83 ENSO influence. In fact, the season was the quietest in over half a century. There were only four hurricanes, although one of them - Alicia - caused $2 billion damages around Galvaston, Texas. Also, in the Pacific, Typhoon tracks were substantially displaced from normal. Five typhoons struck French Polynesia, which rarely experiences them. An unusual northward moving typhoon crossed the Hawaiian Idlands. It was similar to a storm that had struck 25 years earlier, also during an ENSO event.
Thus it is that forecasting scientists see enough potential for improvement to want to get on with the job of developing it. But they are held back by poor funding. Hecht said that only about $1 million a year in federal money is available for the research. He called this 10 times too low to do a proper research job. ''We have something exciting here. Why don't we get off the dime?'' he asked rhetorically.
Hecht has a point. For decades, long-range forecasters have been working at the edge of what is possible. Now they are ready to move off the margin toward a higher degree of skill. Budget planners should wake up to this opportunity and put a bit more money into what now looks to be an opportunity to make a highly leveraged investment in research with a potentially big pay off.