CIMA   09099
CENTRO DE INVESTIGACIONES DEL MAR Y LA ATMOSFERA
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
Probabilistic forecasts during the monsoon season: the analogs technique as a tool for precipitation prediction over southeastern-south America.
Autor/es:
ALDECO, L.; J. RUIZ; C. SAULO
Lugar:
Noumea
Reunión:
Congreso; X International Conference on Southern Hemisphere Meteorology and Oceanography; 2012
Institución organizadora:
American Meteorological Society
Resumen:
Precipitation forecasts quality becomes critical for many applications and, in the context of Monsoons, it is of main interest to determine, for example monsoon onset, demise and/or break phases. This type of forecasts is related with temporal scales ranging from synoptic to sub-seasonal time scales. At these temporal scales, precipitation forecasts have very low skills. Still there are some alternatives that may be applied to generate precipitation forecasts at medium ranges that exhibit some skill and have potential to be further improved. In this work an implementation of the analogs technique to generate probabilistic precipitation forecasts (PPF) for short and medium range is evaluated. This implementation is based on the MRF-NCEP reforecast system, covering 29-year period, from 1979 to 2007. This reforecast runs with relatively coarse resolution (2.5°), and provides up to 15-days forecasts with 15 ensemble members, and guarantees that the same model has been used to generate all the forecasts. Data availability is critical for implementing the analogs technique, so, at this initial stage, instead of verifying the monsoon core region; we selected a sub-domain over Southeastern South America (SESA) where the amount of precipitation observations is reasonable. Daily observed accumulated precipitation over this region is used both for training and verifying PPF at different time ranges between 1979 and 2007 summer seasons. The Brier Skill Score and its components have been used to evaluate the performance of these forecasts. Results show that it is feasible to generate reliable PPF up to 9 days in advance over the region of study. Future work will use other data bases like CMORPH in order to assess if similar results can be obtained over SESA and then adapt the technique to the detection of monsoon onset over South America.