EXPERIMENTAL LONG-LEAD FORECAST BULLETIN
This Bulletin, issued quarterly by the Center for Ocean-Land-Atmosphere Studies (COLA), is intended as an opportunity to present experimental forecasts with a long lead time. In the process we hope to achieve an exchange of ideas on long lead forecasting, which would serve as a mechanism to stimulate research in this area. The forecasts are mainly intended for the surface climate in the United States, such as a forecast of winter U.S. surface temperature 6 months or a year in advance. Forecasts for other aspects of climate, such as ENSO, the hurricane season severity, or surface climate in other parts of the world are also welcome.
Long range forecasting, with or without a lead time, is in an embryonic stage of development. Some recent progress has been made with dynamical approaches to such forecasts, using either simple or complex, coupled ocean-atmosphere or uncoupled atmospheric GCMs. A comparable level of achievement has also been attainable using empirical methods.
It is our goal that the next several decades will bring a major improvement in understanding and in forecast skill. In the process of this progressive advancement in knowledge and ability, a Bulletin such as this can help focus new ideas and disseminate information that will keep interested researchers abreast of the lastest developments. It will also encourage collaboration among researchers and forecasters in this field. The real-time forecasts that appear in this Bulletin will serve as examples of what can be done now, and add some reality and suspense to what otherwise could only be an academic research discussion.
The contents of this Bulletin are not routine; i.e. some of the forecasts presented are expected to vary widely in their methodology, lead time and specific predicted quantity from issue to issue.
This Bulletin is distributed to researchers as well as those with more applied interests. Persons wishing to contribute to the Experimental Long-Lead Forecast Bulletin should contact Ben Kirtman (editor) at 301-595-7000 (fax 301-595-9793), write to COLA, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705 or e-mail to firstname.lastname@example.org.
Researchers from research and academic centers, who are interesed in contributing to this Bulletin are invited to do so. A variety of forecasting techinques would add value and interest to the Bulletin. Dynamical approaches are particularly encouraged. Prospective participants should read the following guidelines to determine if their contribution meets a minimum standard of credibility.
GUIDELINES FOR CONTRIBUTIONS
A. Time Scale, Lead, and Real-Time Nature of Forecasts
For the purpose of this Bulletin, we define the forecast averaging time scale as at least one month in duration, and extending to as much as 3 or more months. Thus, daily weather variability is to be essentially excluded, and short-term climate variability emphasized.
The long-lead aspect means that the forecast target period will not begin at the time of the forecast, but at a future time at which numerical weather prediction skill would, on average, be minimal. Two weeks of lead time appears to be a reasonable minimum for satisfying this condition. Leads of much longer duration are heartily encouraged, if warranted.
Forecasts are to be of real-time nature, i.e., the target period must not yet have occurred. This is in contrast to hindcasts, for which observed data for the target period are available at the time of the forecast.
In order that the forecasts be acceptable, they must ideally satisfy the following criteria:
(1) Reproducibility: The method must be describable in terms such that others could generate an identical forecast by following the method description---at least in principle.
(2) Skill: A track record is required. The skill of the method, as evaluated in a manner that accounts for artificial skill must be available, and should be better than that of random forecasts beyond reasonable doubt.
(3) Documentation: The method must have been presented in at least a non-refereed publication, in the current general context.
(4) Presentation Standards: The forecast should be cast in unambiguous terms. When using anomalies and/or classes the base period must be stated.
C. Narrative Summary of Method, and Its Estimated Skill
Accompanying a long-lead forecast, there should appear a brief description or explanation of the method employed, which may be either ongoing or new. While a few specific formulations or other technicalities are appropriate, references to refereed or informal publications should be made in most instances. Estimates of forecast skill must be reported, using tools that minimize artificial skill (e.g., cross-validation, or independent forecast testing). Contributions are due by the 13th of the months of March, June, September, or December; first-time contributions should be sent earlier, as major revisions may be needed.