Potential predictability of South America temperature and precipitation in CHFP and SHFP models

MARISOL OSMAN; CAROLINA VERA

Montevideo

Conferencia; WCRP Conference for Latin America and the Caribbean: Developing, linking and applying climate knowledge; 2014

World Climate Research Programme

The study documents the predictability of seasonal means of surface temperature and precipitation in South America during JJA and DJF. Ensembles of lead-1-month climate predictions of 17 models (each one with 10 members on average) from the WCRP/CHFP-SHFP Project are used. In particular, 5 different climate models which resolve the stratosphere (high-top models) and 12 models which do not (low-top models) are considered. An unbiased multi-model ensemble (MME) was constructed by pooling all models together. High-top and a low-top MME were also made. Predictability is computed as the ratio between the ensemble mean variance (or variance of signal) and the total variance, computed as the sum between signal variance and the ensemble spread (or noise variance). Temperature predictability in South America is higher at tropical latitudes in both seasons and drops at midlatitudes. However, there is a local maximum in DJF over central Argentina. Predictability for ENSO years shows the same spatial pattern but with higher values in most of the continent. In addition, high-top MME presents in DJF lower (higher) signal (noise) values over southern extratropical South America than low-top MME. Moreover, predictability is slightly higher over most of the domain with the low-top MME. On the other hand, in JJA predictability depicted by high-top MME is higher at tropical latitudes while at midlatitudes the opposite happens. In DJF, predictability is smaller for both El Niño and La Niña years than for all years and particularly at tropical latitudes. Noise (signal) changes are responsible for the differences at tropical latitudes in El Niño (La Niña) years. In contrast, only signal changes are relevant at midlatitudes. In JJA, predictability is higher (lower) in El Niño (La Niña) than in all years over central Argentina. South and north of this region, the relationship is the opposite. In all cases this behavior is due to signal changes. In DJF, maximum signal is found for precipitation at tropical South America and drops abruptly to near zero values at midlatitudes. The exception is southeastern South America (SESA) region that exhibits moderate predictability values. In JJA, high values of signal are restricted to both northern South America and central Chile, where also modest predictability is found. Precipitation predictability during ENSO years increases, especially in DJF. In DJF (JJA) high-top MME shows higher (lower) predictability than low-top MME at the tropics and in SESA. Finally, precipitation predictability in both El Niño and La Niña years is lower than in all years for both seasons, mostly due to signal changes. In summary, the study shows as expected higher predictability levels at tropical South America than at the extra-tropics. However, regions like SESA and central Chile can exhibit moderate levels of predictability depending on the variable and season. In general, predictability slightly increases in ENSO years than in other years, while the contribution to predictability increase due to a better resolving stratosphere is not that evident.