ENSO Simulation and Prediction in a Hybrid Coupled Model with Data Assimilation

抄録

With a 3D Var assimilation scheme, several types of observations—sea surface temperatures (SST), sea level height anomalies (SLHA), and the upper ocean 400 meter depth-averaged heat content anomalies (HCA)—were assimilated into a hybrid coupled model of the tropical Pacific. The ocean analyses, and prediction skills of the SST anomalies (SSTA) from the assimilation of each type of observation, were presented for 1980-998. SST assimilation, besides improving the simulation of SSTA, also slightly improved the HCA and SLHA simulations in the equatorial Pacific, especially in the east. The ocean analyses with the assimilation of SLHA improved the simulations of SSTA, SLHA and HCA in the equatorial Pacific, while the assimilation of HCA improved the SLHA and HCA simulations.
For ENSO predictions, assimilating SST yielded the best prediction skills for the Niño3 region SSTA at lead times of 3 months or shorter, but severely degraded the predictions at longer lead times. The best Niño3 SSTA predictions for lead times longer than 3 months came from the initializations with the assimilation of HCA and SLHA data. Assimilating SLHA yielded prediction skills for the Niño3 SSTA almost as good as assimilating HCA, indicating considerable potential for improving ENSO predictions from altimetry data. In this study, a neural network (NN) approach was used to find the nonlinear statistical relations among model variables for the assimilation of HCA and SLHA. Using NN yielded better prediction skills than using multiple linear regression.