Figure 4 The anomaly correlation (top) and rms error (bottom) of the coupled model hindcasts verified against the observed monthly mean NN03 index for the periods 1971-80 (left) and 1971-94 (right), expressed as a function of lead time in months. Hindcasts were perfonned for February, May, August and November of each year. CNTL refers to the control experiment in which only the wind-forcing history is used to initialise the model. FCST refers to the hindcasts using both wind and the ALL DATA temperature analyses for initialisation. PERS refers to a NINO3 anomaly persistence hindcast.
