J3.3
7-Day, 10-Day, 14-Day Forecasts: How far can we go?
Glenn Schwartz, NBC10 - Philadelphia, Bala Cynwyd, PA
The dominant forecast form from the 1970s through the 1990s was a 5-day. This remained constant even with significant increases in computer model accuracy. In recent years, more and more forecast outlets have expanded deterministic forecasts out to 7, 10, and even 14 days. This includes the National Weather Service (7 days), The Weather Channel (10 days), and Accu-weather (14 days). We have investigated the top 50 TV markets to see how many stations are now going beyond the traditional 5-day forecast.
The next question is whether the science allows forecasting that far out with any skill. NBC10 has conducted an extensive verification program over the past few years to see how far forecasts outlets can predict with at least SOME improvement over climatology. While the verification is solely for Philadelphia, it may provide an idea how far out we can predict with any skill at all. There is a limit, which varies depending on who is doing the forecasting. Some outlets showed no skill beyond day 7, even showing negative skill starting on day 8. This calls into question whether deterministic forecasts beyond 7 days have ANY scientific value at all.
In the multi-year study, only NBC10 was able to have even minimal skill out to day 10. As a result of the first year results, we changed from a 5-day to a 7-day forecast, with a "10-day trend" added on 11pm newscasts. Examples of some 10-day graphics and weather condition verification will be shown. We have found that it IS scientifically justified to provide GENERALIZED forecasts out to day 10, even with phenomena such as winter storms and hurricanes. Verification has also shown that certain forecast words in the extended period have much less skill than others.
The author will share some of the general forecast methods and sites available to help make forecasts beyond day 5, and encourage others to not only make extended forecasts, but verify them as well. A lot can be learned from verification, such as forecast weaknesses or biases.
The implication of the research and 27 years of TV experience is that mid-latitude TV stations SHOULD be forecasting out to 7-days, but not attempt deterministic forecasts beyond that. After NBC10 went to the 7-day format, one competitor expanded to a 10-day deterministic forecast within DAYS, and the other major competitor added a 6th and 7th day outlook to every forecast. It is clear that it is important to be first in forecast expansion, and for the meteorologists to lead the way (so they're not forced to go out farther than they're comfortable). If not, they might be forced into doing forecasts on TV or on the web site that are beyond the ability of the science.
Recorded presentationJoint Session 3, Communicating Uncertainty (Joint with 2nd Policy Symposium, 35th Conference on Broadcast Meteorology and 6th Communications Workshop)
Wednesday, 17 January 2007, 10:30 AM-11:15 AM, 205
Previous paper