关键词: 海洋资料同化, 集合卡尔曼滤波, 气候预测

Abstract:

Ensemble Kalman filter (EnKF) is a widely used ocean data assimilation scheme. It uses the members' state to represent background error covariance of the model and uses the observation error covariance to calculate the Kalman gain matrix; and then it adds the information of observation to the model's initial conditions. Data assimilation can effectively improve climate prediction skill since seasonal-to-interannual climate prediction is largely influenced by initial conditions. Based on the NUIST-CFS1.0, previously SINTEX-F, that used coupled sea-surface temperature (SST)-nudging initialization method, we employ EnKF to assimilate SST, gridded altimeter satellite sea-level anomalies, and in-situ temperature and salinity profiles at the end of each month. We assess the differences in the initial fields and climate (including the ENSO) prediction skills at lead times of up to 24 months with and without the ocean data assimilation. The results show that the initial conditions are improved largely with the ocean data assimilation, and the improvement of the initial fields is getting better with the increase of depth. The global mean prediction skills of SST and subsurface temperature with EnKF are improved at 1-24 months lead, and the improved prediction also becomes more obvious in deep layers. However, the prediction skill of the ENSO is reduced after the EnKF assimilation. This might be due to the existence of model errors, the predicted initial field with EnKF is gradually restored from a state close to the observed state to match the model dynamics again, which displays a larger cold SST bias in the west and warm bias in the central-eastern equatorial Pacific, compared to the original NUIST-CFS1.0.

Key words: ocean data assimilation, ensemble Kalmen filter, climate prediction