Forecasts of Tropical Atlantic SST Using a Statistical Ocean Model In The Operational Forecast Routine at CPTEC

Forecasts of Tropical Atlantic SST Anomalies Using a Statistical Ocean Model at CPTEC/INPE - Brazil

contributed by Luciano P. Pezzi¹, Carlos A. Repelli², Paulo Nobre³

Iracema F. A Cavalcanti¹ and Gilvan Sampaio¹

^{1CPTEC - Center for Weather Forecast and Climate Studies, Cachoeira Paulista, Brazil
^{2LDEO Lamont-Doherty Earth Observatory, Columbia Univ., Palisades, New York}}

The Center for Weather Forecast and Climate Studies (CPTEC) located in Brazil, has recently developed and implemented a modeling system to predict tropical Atlantic sea surface temperature anomalies (SSTA). This system named SIMOC (Statistical Modeling System) produces skillful forecasts (better than persistence) for lead times up to six months in advance (Repelli and Nobre, 1998).

The technique employed is multivariate statistics, canonical correlation analysis (CCA; Barnett and Preisendorfer, 1987). This technique identifies a sequence of predictor patterns which tend to evolve into subsequent patterns that can be used to form a forecast.

A full description of this model used to forecast tropical Atlantic SSTA can be found Repelli and Nobre (1998). The model uses data from the Comprehensive Oceanic and Atmospheric Data Set (COADS) for the 1945-1993 period and Optimum Interpolation SST (OI-SST) from 1994 until the present, provided by CPC/NCEP. The tropical Pacific and Atlantic SSTA between 20° S and 30° N are the predictor fields. The predictand is the SSTA over the same latitudes, but only over the tropical Atlantic. Forty-nine years (1945 to 1993) are used to build this model, which are divided in training period (1945 to 1978) and hindcast period (1979 to 1993). The results found in Repelli and Nobre (1988) suggest that the most skillful forecasts are for March-April-May using September to January data as predictors. This results can be detected in Fig. 1 (time series) and Fig. 2 (spatial values). These figures show correlation coefficients between the predicted SSTA and the observed SSTA during 15 years of hindcasts (independent period).

The forecasts for March-April-May 1998 are shown in Figs. 3a-d, using SSTs of October 1997 January 1998. Northern and southern tropical Atlantic dipole regions as defined by Moura and Shukla (1981) are important indicators of tropical Atlantic variability. The northern dipole is the area averaged SSTA from 60° W-20° W, 5° N-20^oN SSTA and the southern dipole is the area averaged SST from 20° W-0° , 15° S-0° . Table 1 shows these dipole indices for the October 1997 through January 1998 initial conditions. The SSTA of the northern tropical Atlantic are higher than the SSTA in the southern tropical Atlantic. This north-south asymmetry can be detected in the March-May predicted SSTA as well as the increasing amplitude of the anomalies with the later initial conditions.

	Oct 1997	Nov 1997	Dec 1997	Jan 1998
North	1.14	1.86	1.72	1.76
South	0.60	0.53	0.91	0.90

Tab 1 - March-April-May 1998 area averaged SST anomalies over the northern and southern branches of the dipole spanning initial conditions from October/97 to January/98

Barnett, T. P. and R. Preisendorfer, Origins and levels of monthly and seasonal forecast skill for United States surface air temperatures determined by canonical correlation analysis, Monthly Weather Review, 115,1825-1849,1987.

Moura, A. D., and J. Shukla, On the dynamics of droughts in Northeast of Brazil: observations, theory and numerical experiments with a general circulation model, J. Atmos. Sci.,38, 2653-2675,1981.

Repelli, C. A., and Nobre, P., Statistical Prediction of Sea Surface Temperature over the Tropical Atlantic. CPTEC, Internal Report, Jan. 1998. Cachoeira Paulista, São Paulo. Brazil.