Studies Of Predictability Problems For Zebiak-Cane ENSO Model

With an intermediate coupled ocean-atmosphere ENSO model of Zebiak and Cane (1987) (ZC model), the predictability problems for ENSO are explored by using a new approach of conditional nonlinear optimal perturbation (CNOP).The double-precision tangent linear and adjoint models of ZC model are derived by the approach of from code to code and are checked to be valid. After the test of the gradient of a prescribed cost function, we establish a nonlinear optimization system of ZC model.For annual cycle being basic state, CNOPs are calculated to investigate the optimal precursors of ENSO events. It is shown that CNOP (local CNOP) describes the initial anomaly patterns that evolve into El Ni?o (La Ni?a) events most probably. When the SSTA component of CNOP (local CNOP) consists of an east (+)-west (-) (east(-)-west(+)) dipole spanning the entire tropical Pacific basin, and the thermocline depth anomaly component exhibits a deepening of the whole equator Pacific, CNOP (local CNOP) will evolve into El Ni?o (La Ni?a) events most probably, which is consistent with observational results provided by TAO Project Office of NOAA/ PMEL. The spatial structure of the optimal precursors is remarkably insensitive to the initial time at which initial anomalies appear and to the duration over which anomalies develop. Nonlinearity plays an important role during the development of ENSO events. Nonlinear advections of SSTA enhance El Ni?o event, but suppress La Ni?a event. Nonlinearity in the parameterization of subsurface temperature also suppresses La Ni?a considerably. For these reasons, there exists obvious asymmetry of intensity and duration between El Ni?o and La Ni?a events.To investigate the maximal prediction error of ENSO events, CNOPs are also calculated by using ENSO events as basic state. It is indicated that during the transition period from La Ni?a to El Ni?o, initial errors of CNOP type possesses most important effect on the uncertainty of prediction results. Because of the growth of initial error, it intends to yield false alarms of El Ni?o prediction for a normal state...