The CPTEC global model bias in the Amazon region: results of long term simulations and predictive skill in seasonal forecasting.
Helio
Camargo, CPTEC/INPE, helio@cptec.inpe.br
(Presenting)
Pedro
Leite
Silva Dias, CPTEC/INPE - IAG/USP, pldsdias@master.iag.usp.br
José
A.
Marengo, CPTEC/INPE, marengo@cptec.inpe.br
Anete
Fernandes, CPTEC/INPE, anete@cptec.inpe.br
Lincoln
Muniz, CPTEC/INPE, lincoln@cptec.inpe.br
Nuri
Calbete, CPTEC/INPE, nuri@cptec.inpe.br
Christopher
Castro, CPTEC/INPE, castro@cptec.inpe.br
Ana Paula
Maletba, CPTEC/INPE, malerba@cptec.inpe.br
Ana Cláudia
Araújo
Preste, CPTEC/INPE, anaprest@cptec.inpe.br
Operational Seasonal forecasts have been issued by the Center for Weather Forecasting and Climate Studies (CPTEC) based on an Atmospheric General Circulation Model (AGCM - spectral resolution T62 and 28 vertical levels) since 1995. These runs are performed in two different modes: the first mode assumes that the Sea Surface Temperature anomalies (SSTA) is persisted during the 8 month integration; the second mode is based on the SSTA produced by the NCEP Equatorial Pacific coupled model and/or CPTEC's Tropical Atlantic Canonical Correlation model and persisted SST anomalias elsewhere. The evaluation of the CPTEC AGCM reference skill is based on a 50 year simulation forced by the observed SSTA (1952-2001) with 9 members of the ensemble (which differ by the choice of different initial conditions).. Simple statistical analysis for the Amazon was performed. Results show that even though acceptable and statistically significant linear correlation values were found for northern Amazon for March-April-May (MAM) concerning on the long term simulation, interanual variability shows positive and negative bias, comparing to Xie-Arkin observed precipitation. Considering the last few operational forecasts, results show a positive biased signal for northern Amazon. In the southern portion of the region although the last forecasts captured the observed negative anomalies, no statistically significant correlation values were found for the rainy season (December-January-February - DJF). An analysis of the last 7 years performance of the operational forecasting is also presented