Relationship between error and ensemble spread in a regional ensemble forecast system for South America.

JUAN JOSÉ RUIZ, A. CELESTE SAULO Y EUGENIA KALNAY

Melbourne (Australia)

Congreso; : 9th International Conference on Southern Hemisphere Meteorology and Ocanography; 2009

American Meteorological Society

According to Kalnay (2003) ensemble forecasting has two major advantages, one is to provide a better estimation of the forecasted variables trough the use of the ensemble mean, and the other is the estimation of the forecast uncertainty through the ensemble spread. Though originally developed for medium range global forecasts, short range regional forecasts can also benefit from ensemble forecasting techniques (see for example Hou et al. 2001). Little is known about the performance of regional ensemble forecasts over South America in terms of error reduction and forecast uncertainty estimation (Silva Dias et al. 2006, and Ruiz et al. 2006). This work provides an assessment of the performance of a regional and a global ensemble system during the 2002-2003 warm season. Several works studied the relationship between error and ensemble spread: Houtekamer (1993) develops a theoretical model for this relationship, and Whitaker and Lounge (1998) analyze this relationship for a global ensemble over the Northern Hemisphere. In these works a linear correlation coefficient was used to measure the strength of the relationship between the two variables. Grimit and Mass (2007), use the theoretical model provided by Houtekamer (1993) to test other measures of strength of this relationship, most of them not based on a linear relationship between the two variables. In the present work we propose to study some aspects of the relationship between error and spread without restricting ourselves to the a linear framework.