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

JUAN RUIZ; CELESTE SAULO; EUGENIA KALNAY

Melbourne, Australia

Conferencia; 9th International Conference on Southern Hemisphere Meteorology and Oceanography; 2009

American Meteorological Society

<!-- /* Style Definitions */ p.MsoNormal, li.MsoNormal, div.MsoNormal {mso-style-parent:""; margin:0cm; margin-bottom:.0001pt; mso-pagination:widow-orphan; font-size:12.0pt; font-family:"Times New Roman"; mso-fareast-font-family:"Times New Roman"; mso-ansi-language:EN-US; mso-fareast-language:EN-US;} @page Section1 {size:612.0pt 792.0pt; margin:72.0pt 90.0pt 72.0pt 90.0pt; mso-header-margin:36.0pt; mso-footer-margin:36.0pt; mso-paper-source:0;} div.Section1 {page:Section1;} --> Abstract: A regional short range ensemble forecast system (SREF) has been tested over South America during the 2002-2003 summer season. A five pair breeding cycle based on the MRF (Medium Range Forecast) model was used to generate perturbed initial and boundary conditions in order to run a regional ensemble based on the WRF (Weather Research and Forecasting) model which has been initialized twice a day and integrated up to 48 hours lead time. The performance of the global and the regional ensembles over South America has been tested using reanalysis and observational data from the surface weather stations. The results show that both ensembles lead to an error reduction of approximately 10%, with reduction rates increasing with forecast lead time. In addition, it was found that the error of the regional forecast is also smaller than that of the global one when compared at the same spatial resolution. The error/ensemble-spread relationship was determined using a non linear approach where the sensitivity of the error probability distribution to the ensemble spread amount is examined for different error thresholds. The results show that the error/ensemble-spread relationship is strongly dependent on the variables considered and on the region under investigation. Also, it was found that both ensembles are more skillful in predicting larger errors than in predicting the smaller ones. The relationship between error and spread increases with the forecast length, being nearly negligible during the first 12 hours. The regional ensemble shows a slightly worse performance in forecasting the ensemble skill when compared to the global ensemble at the same spatial resolution.