Prediction skill assessment of large-scale variability influence in summer southeastern South America rainfall in multi-model CMIP decadal predictions

LEANDRO DÍAZ; CAROLINA VERA; RAMIRO SAURRAL

Barcelona

Workshop; CMIP6 Model Analysis Workshop; 2019

Barcelona Supercomputing Center (BSC)

The main motivation of the work is try to get useful predictions for austral summer SESA rainfall from CMIP5 decadal predictions. As it is known from previous works, prediction skill is very low in this dataset. The approach adopted is to study the covariability pattern of rainfall in that region and global SST anomalies, using singular value decomposition methodology. the pattern explains a large fraction of the variance in SESA region. So, we explore the prediction skill of this co-variability mode in CMIP5 decadal prediction ensemble. One of the challenges is also to deal with multi-member/model ensemble Information. Here, I will present results from four different methodologies, evaluating the skill using deterministic and probabilistic approaches. Some of the key findings is that some skill could be obtained until the second forecast year and that the ?superensemble? methodology has the best performance