SST Forecast and SST-US Precipitation Connections: A Diagnostic for North American Multi Model Ensemble

In this study, we evaluated the SST and US precipitation hindcasts from the North American Multi-Model Ensemble (NMME), focus on how coupled models could catch the SST variability in spatial modes, and the coupled co-variability between the SST with US precipitation. Based on the over 100 years SST observations, we isolated the four leading SST modes by Rotated EOF , i.e. Global warming trend, ENSO, AMO and PDO like patterns. We project the model simulated SST to those modes and then evaluated against with observations. The current NMME monthly and seasonal forecast could catch those variabilities in modes, measured by Pearson correlation, even after 5 or 7 months in lead time. The NMME show pretty good forecast skills over the baseline persistent forecast for ENSO and PDO modes, but no skill at AMO mode. The reason is the deficit in SST simulation at Atlantic extra-tropics. However, the current NMME models still cannot simulate the linkage between SST and US precipitation. Only half models could get the co-variability of ENSO and US precipitation at the leading pattern by SVD decomposition, compared with observations. The principal component regress (PCR) can get correct zonal structure of ENSO-precipitation pattern with different intensities, but diversity patterns in the connection between PDO and AMO modes. Those deficient contribute to the forecast error in the prediction of US precipitation. More detail results will be presented during conference.