Tab. 1

Experimental set up"

Tab. 2

Data used"

Fig. 1

Spatial distribution of correlation coefficient and RMSE/℃ between two experiments' SST initial condition anomalies and observation anomalies. (a) The correlation coefficient of experiment 1; (b) the correlation coefficient of experiment 2. The color filled area in (a) and (b) has passed the 99% significance test. (c) The difference of correlation coefficient between with and without ocean data assimilation (i.e., experiment 2 minus experiment 1). The warm color means improved, and the cold color means deteriorated. (d) RMSE of experiment 1; (e) RMSE of experiment 2. (f) The difference of RMSE between with and without ocean data assimilation (i.e., experiment 1 minus experiment 2). The warm color means improved, and the cold color means deteriorated"

Fig. 2

Spatial distribution of correlation coefficient and RMSE/m between two experiments' Z20 initial condition anomalies and observation anomalies"

Fig. 3

Spatial distribution of correlation coefficient and RMSE/℃ between two experiments' T100 initial condition anomalies and observation anomalies"

Fig. 4

Spatial distribution of correlation coefficient and RMSE/℃ between two experiments' T300 initial condition anomalies and observation anomalies"

Fig. 5

Time series of two experiments' variable anomalies in the initial conditions of the equatorial ocean regions. From the top to the bottom are the eastern equatorial Pacific region (a-c), the western equatorial Pacific region (d-f), the equatorial Indian Ocean region (g-i), and the equatorial Atlantic region (i-j). From the left are the time series of Z20 anomaly, T100 anomaly and T300 anomaly. The black solid line is the observation, and the blue (red) solid line, R and RMSE are the time series, correlation coefficient and root mean square error of experiment 1 (experiment 2), respectively"

Fig. 6

Time series of two experiments' global mean (60°S-60°N) ACC and RMSE prediction skills (i.e., 1-24 months lead) of SST, Z20, T100, and T300. The blue (red) solid line and dotted line are the correlation coefficient and RMSE of experiment 1 (experiment 2), respectively"

Fig. 7

Spatial distributions of two experiments' anomaly correlation coefficient (ACC) skills at 6 months lead. From the top to the bottom are SST, Z20, T100, and T300, respectively. From the left panels are the results of experiment 1 (a, d, g, j), the results of experiment 2 (b, e, h, k), and the differences (c, f, i, l) between with and without ocean data assimilation (i.e., b minus a). The warm color means improved, and the cold color means deteriorated"

Fig. 8

Spatial distributions of two experiments' anomaly correlation coefficient (ACC) skills at 12 months lead"

Fig. 9

Spatial distributions of two experiments' anomaly correlation coefficient (ACC) skills at 18 months lead"

Fig. 10

Spatial distributions of two experiments' anomaly correlation coefficient (ACC) skills at 24 months lead"

Fig. 11

Time series of two experiments' Nino3.4 ACC and RMSE prediction skills. The solid and dotted blue (red) lines are the correlation coefficient and RMSE of experiment 1 (experiment 2), respectively"

Fig. 12

Differences between two experiments' SST climatic mean states at different months lead. The left panel (a) is observation climatology, the middle panels (b, d, f, h) are SST climatic mean states of experiment 1, the right panels (c, e, g, i) are SST climatic mean states of experiment 2. From the top to the bottom are the results at 6, 12, 18, and 24 months lead over the period Jan 1985-Dec 2014"

Fig. 13

Differences between two experiments' SST climate mean states biases at different months lead. The left panel (a) is observation climatology, the middle panels (b, d, f, h) are SST climate mean states biases of experiment 1, the right panels (c, e, g, i) are SST climate mean states biases of experiment 2. From the top to the bottom are the results at 6, 12, 18, and 24 months lead over the period Jan 1985-Dec 2014"