PREDICTABILITY AND PREDICTION SKILL OF SOUTH AMERICAN CLIMATE

MARISOL OSMAN; CAROLINA VERA

Mar del Plata

Congreso; CONGREMET 2015; 2015

Centro Argentino de Meteorólogos

The study documents the predictability of seasonal means of temperature and precipitation in South America during winter and summer. The study is based on ensembles of lead-1-month climate predictions outputs for DJF and JJA of 15 coupled models from both CHFP and SHFP, coordinated by WCRP. Total variance (V), signal variance (S2) and noise variance (N2) were computed for the multi-model ensemble. Potential predictability is defined as the ratio between the signal variance and the total variance (V). In addition, prediction skill, defined as the mean Anomaly Correlation Coefficient (ACC) between models and observations was also computed. Predictability and skill associated with ENSO events were also studied by computing the same indexes only for ENSO years. Results show that in DJF, highest signal and noise values are found in northern and northeastern South America being signal higher than noise. In addition, modest values of signal and noise are found at extratropics where noise is maxima over southeastern South America (SESA). In JJA, signal and noise show the same structure than in DJF, but with lower values. Consequently, temperature predictability is highest at tropical latitudes, reaching values higher than 0.8 and drops to less than 0.5 at midlatitudes in both seasons. However, in DJF predictability at central Argentina is slightly higher than the rest of extratropics. As for temperature, precipitation signal is highest at northern South America and peaks in the western coast and in the equatorial Atlantic Ocean. At extratropics signal is closed to zero everywhere except SESA and southern Chile. Noise in DJF is higher than signal but remains with the same structure. As a consequence, predictability is highest at tropics, although is lower than 0.5, and at extratropics only SESA and southern Chile show nonzero values. In JJA, both signal and noise are highest at northern South America over the oceans. At extratropics, signal and noise are different of zero at central and southern Chile, southern Brazil and the adjacent ocean. Finally, predictability is higher in JJA than in DJF and is highest at northern South America, SESA, central Chile and southern South America. The analysis of prediction skill reveals that temperature ACC is significant at north and northwestern South America reaching 0.8 in the coast adjacent to tropical Pacific. However, at extratropics skill drops considerably and is significant only in central Argentina and southern Chile. In JJA, the situation is more dramatic as most of the land shows temperature ACC non-significant. On the other hand, the skill in forecasting precipitation is generally lower than that for temperature. In DJF, significant skill is found at tropics in northern South America and in the Oceans at the Equator while at extratropics only SESA shows significant correlation values. In JJA the skill is somewhat smaller than in DJF and migrates from continent to the ocean at tropics and extratropics. The fact that lower values of prediction skill than predictability were observed may suggest a possibility for models of improving their performance in the next years.