By Robert E. Livezey, Climate Prediction Center, NCEP/NWS/NOAA

The official seasonal forecasts for the United States for January through March 1999 (Fig. 1) made in mid-December 1998 by the National Weather Service's Climate Prediction Center (CPC) are excellent examples of a new approach to making these and longer-lead predictions. This approach is more physically-based than previous methods because it starts with an assessment of what aspects of climate variability are most important to the current situation, which in turn dictates to the forecaster the tools to bring to bear on the problem. From time to time this can lead to sharply focused, high-confidence predictions, so-called "forecasts of opportunity", which were not previously feasible (or defensible) with statistical techniques that attempted to treat all factors simultaneously and indirectly. This strategy was used to make the highly successful forecasts for the winter of 1997-98 possible.

The forecast process begins with a prediction of the state of the El Nino/Southern Oscillation (ENSO) - one of the most important factors controlling year to year changes in average winter conditions and also the most predictable. ENSO refers to multi-year shifts in sea surface temperatures (SSTs) in the central equatorial Pacific Ocean. When these waters are substantially warmer than normal, the condition is called El Nino, and when they are colder than normal, the condition is referred to as La Nina. El Nino and La Nina are important to wintertime North America because much of the energy which drives the global wind and weather systems is converted from the Sun's energy by extensive rain systems over the tropical Pacific. El Nino stokes up this engine while La Nina puts the brakes on it. In either case, the implications for the North Pacific subtropical jet stream and weather patterns north and east of the jet are dramatic. In a rough sense, La Nina's effects are the opposite of El Nino's - not only locally in the Pacific but also remotely over North America. For example, during January through March in the U.S., La Nina usually brings drier than normal conditions to the Southwest and Southeast where it is usually wet with an El Nino, and wetter than normal in the Pacific Northwest and the Ohio Valley where El Nino implies relatively dry conditions.

Last spring and early summer the 1997-98 El Nino was replaced by La Nina (Fig. 2), which subsequently intensified to at least moderate strength by the time CPC forecasters sat down in mid-December to make the forecasts for January through March 1999. Because of agreement between two statistical models and a highly sophisticated computer model of the coupled ocean/atmosphere, they were able to predict with a high degree of confidence the continuation of this moderately strong La Nina in the equatorial Pacific Ocean for at least several more months.

To account for the presence of the La Nina, the seasonal forecasters consulted a chart that shows estimates based on records back to the 1930s of how likely precipitation will be to fall amongst the wettest or driest thirds of historical observations given a moderate to strong La Nina. This chart (Fig. 3) formed the principal basis for forecast precipitation patterns and probabilities (Fig. 1), because all other factors had either small (compared to La Nina) or indeterminate effects. In contrast, at least two other factors had important implications for the temperature forecast.

The first of these factors is long-term trend that may be associated with global warming. While the existence of real upward trends in both wintertime temperature and precipitation have been demonstrated, only the former is thought to be strong enough to significantly distort an expected La Nina pattern. In the absence of temperature trends, the conventional scenario for a La Nina U.S. winter would be for warmer than normal in the southeast and south central regions, along with colder than normal in the Northwest and along the west coast and the western and central Canadian border. Strong warming trends (Fig. 4) over the last 30 or so years throughout the western United States might modify this picture substantially. CPC scientists believe this was part of the reason for more of the country experiencing warmer than normal conditions last winter than expected from El Nino alone.

Consequently, a chart (Fig. 5) corresponding to that for precipitation (Fig. 2) was developed for January through March U.S. temperatures to account for both the La Nina and observed trends since the mid-1960s. This information led to prediction of a much larger area of warmer than normal conditions encompassing all of the Southwest and extending to the west coast, and substantially reduced area and probabilities for colder than normal conditions in the Northwest and along the western U.S./Canadian border (Fig. 1). However, the forecasters reduced the probabilities of relatively warm conditions along the California coast compared to those inland, because of the expected persistence of cold coastal waters. This cold pool was the second factor used to modify the temperature forecasts.

Other factors also impact seasonal averages and some play their most important roles only at certain times of the year. Two very important factors for the cold half of the year are weather patterns over the Northern Hemisphere's two extratropical ocean basins independent of ENSO. These are characterized respectively by the North Pacific Oscillation (NPO) and the North Atlantic Oscillation (NAO). Often these weather systems are long-lived and accompanied by distinct SST patterns in their respective ocean basins. Either system will distort an expected La Nina scenario in known ways. Unfortunately, these systems sometimes undergo big swings lasting a few weeks - changes that are unpredictable more than a week or so in advance, but are significant enough to spoil a seasonal forecast. A good example was a short-lived shift in the NAO last winter, which not only contributed to the record New England ice storm in early January but also supplied enough precipitation in the eastern Ohio Valley to negate the El Nino effect for the season. For this year's forecast there were no indications that the forecast should be adjusted for either pattern.

Ultimately, CPC is striving to produce forecasts of not only seasonal averages but also aspects of within-season variability through use of computer models of the entire coupled global ocean/atmosphere system. The goal is to simultaneously account for ENSO, trends, long-term shifts in the NAO or PDO, and other factors such as soil-moisture feedbacks (important in the warm part of the year) in physically consistent ways. In the meantime, seasonal forecasts will be approached in a manner that takes full advantage of current insight into the phenomena relevant to the problem

For more information:

CPC's full suite of products and information is available on the Climate Prediction Center Web site.