EUROBRISA emerged from the 2001-2005
Ph. D work
Caio Coelho at the Department of Meteorology
the University of Reading
. The aim was to develop
statistical methods for improving the quality of probabilistic South American rainfall seasonal forecasts.
Caio Coelho was supervised by Dr. David B. Stephenson and Dr. Francisco J. Doblas-Reyes and also worked
in collaboration with Dr. Magdalena A. Balmaseda. They jointly developed a Bayesian framework known as
FORECAST ASSIMILATION for calibrating and combining predictions. This framework allows the integration
of empirical and coupled ocean-atmosphere model forecasts to produce a single well-calibrated probabilistic
The approach was developed and tested in three stages:
First, the approach was developed and applied to forecasts of an ENSO index (Nino-3.4) obtained from the
European Centre for Medium-Range Weather Forecasts (ECMWF) coupled model
(Coelho et al. 2004
). The same method was used by
Coelho et al. (2003)
to assess the skill of Nino-3.4
index forecasts produced by four versions of the ECMWF seasonal forecasting system.
Next, the method was extended for the calibration and combination of equatorial Pacific sea surface temperature
forecasts produced by seven DEMETER
coupled models (Stephenson et al. 2005
). The approach
is now able to treat multi-model gridded forecasts.
Finally, the Bayesian forecast assimilation approach was applied to spatial field multi-model forecasts of
rainfall over South America (Coelho et al. 2005a
, Coelho et al. 2005b
and Coelho et al. 2006a
Coelho et al. (2005a,b) and (2006a) applied forecast assimilation to South American rainfall seasonal forecasts produced by an empirical model and three DEMETER coupled models (ECMWF, Met Office and Meteo-France). These three models are now used at ECMWF to compose the European real-time multi-model seasonal forecasting system. Combined and calibrated South American rainfall forecasts obtained with forecast assimilation have been shown to have improved skill when compared to either empirical or coupled multi-model forecasts alone.
Coelho et al. 2006b
successfully applied forecast assimilation for multi-model seasonal forecast
downscaling of regional rainfall and river flows in South America.
These encouraging results stimulated the establishment of EUROBRISA, which will implement the Bayesian
forecast assimilation procedure at the Centre for Weather Prediction and Climate Studies (CPTEC). EUROBRISA
will use forecast assimilation to calibrate and combine real-time South American seasonal forecasts produced
by both the operational European multi-model seasonal forecasting system and the CPTEC coupled model.
EUROBRISA activities are scheduled to start in July 2006. The use of seasonal forecasts for non-profitable
governmental applications is planned within EUROBRISA. Additional information is available in the
, which was approved by
ECMWF council in June 2005.