J3.7
Altered data distributions conditioned on seasonal climate forecasts
Kelly T. Redmond, DRI, Reno, NV
Use of seasonal climate forecasts has been hampered by various barriers identified by Pulwarty and Redmond (1997). Prominent among these has been the lack of tools to help users to interpret the information for their own particular needs. Of particular interest is a distribution of a climate element conditioned on the forecast. Several attempts have been made at this over the years. However, most of the effort has been expended on the technical methodology alone rather than on an end-to-end approach that results in a practical tool with which a user can interact in a friendly way. In this work, another attempt is made, using Monte Carlo methods, fully designed with the expressed needs of users in mind. Terciles of an original distribution of a climate element are identified. A particular tercile is randomly chosen, with the frequency that the forecast indicates. Then, a member of the chosen tercile is (again, randomly) selected. This is repeated many times, and a distribution is built up. The entire process is then repeated, and a composite distribution is assembled. The stopping rule is that some property(ies) of the composite distribution change by less than a specified tolerance. The same process can be performed for the pre-forecast (or no-forecast) case, with all terciles equally likely. These two distributions can be compared, and various properties of the two distributions (unconditional and conditional) can be computed and presented to the user. The distributions themselves can also be graphed. A wide range of user capabilities can be addressed in this way. The user supplies the forecast probabilities, taken directly from the forecast product as issued, such as those supplied by CPC or the IRI (for international forecasts). Many, even most, simply want the implied change in the mean, others want major quartiles or quartile ranges, and others wish the full distribution. Information is presented in graphical and tabular form. It is essential that the tool is accessible via the web, and has features which facilitate use, such as explanations, tutorials, and examples.
Joint Session 3, Climate forecasting (Joint between 15th Conference on Probability and Statistics in the Atmospheric Sciences and 12th Conference on Applied Climatology)
Thursday, 11 May 2000, 8:40 AM-11:59 AM
Previous paper Next paper