854 Verifying the accuracy of experimental day-11 to day-14 weather forecasts

Wednesday, 9 January 2013
Exhibit Hall 3 (Austin Convention Center)
Harvey Stern, University of Melbourne, Melbourne, Vic., Australia; and N. E. Davidson
Manuscript (167.2 kB)

Handout (484.0 kB)

Introduction. In various papers, the present author and colleagues have explored how new technologies might be harnessed to integrate material from various sources on the web to generate new products. The specific purpose of the present paper is to provide an update on this earlier work, in particular that work dedicated to generating experimental day-to-day weather forecasts at the 'outer limit' of potential forecast capability, namely, for days 11-14.

Background. A "real time" trial of a methodology utilised to generate Day-1 to Day-7 forecasts, by mechanically integrating (that is, combining) judgmental (human) and automated predictions, has been ongoing since 20 August 2005. Since 20 August 2006, forecasts have also been generated for beyond Day-7 (out to Day-10). Since 18 January 2009, forecasts have also been generated out to Day-14. Accompanying the Day 1-14 forecasts are automatically generated monthly and seasonal climate outlooks.

Verification. The graphic illustrates the 12-month 'running' average correlation coefficient (Days 11-14) between forecast and observed precipitation probability, precipitation amount, minimum temperature and maximum temperature. The highest correlation coefficients (reflecting the most skilfil forecasts) were those for maximum temperature. The lowest correlation coefficients (reflecting the least skilfil forecasts) were those for precipitation amount. Correlation coefficients for the four elements averaged at about 0.10 during the first year of experimental forecasts. However, the forecasts deteriorated in association with the recent La NiƱa event, before once again averaging at about 0.10 during the most recent twelve months Aug-2011 to Jul-2012.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner