Model errors offset by constant and time-variant optimal forcing vector approaches(termed COF and OFV, respectively)are analyzed within the framework of El Nio simulations. Applying the COF and OFV approaches to the well-known Zebiak–Cane model, we re-simulate the 1997 and 2004 El Nio events, both of which were poorly degraded by a certain amount of model error when the initial anomalies were generated by coupling the observed wind forcing to an ocean component. It is found that the Zebiak–Cane model with the COF approach roughly reproduced the 1997 El Nio, but the 2004 El Nio simulated by this approach defied an ENSO classification, i.e., it was hardly distinguishable as CP-El Nio or EP-El Nio. In both El Nio simulations, substituting the COF with the OFV improved the fit between the simulations and observations because the OFV better manages the time-variant errors in the model. Furthermore, the OFV approach effectively corrected the modeled El Nio events even when the observational data(and hence the computational time) were reduced.Such a cost-effective offset of model errors suggests a role for the OFV approach in complicated CGCMs. Model errors offset by constant and time-variant optimal forcing vector approaches (termed COF and OFV, respectively) are analyzed within the framework of El Nio simulations. Applying the COF and OFV approaches to the well-known Zebiak-Cane model, we re-simulate the 1997 and 2004 El Nio events, both of which were poorly degraded by a certain amount of model error when the initial anomalies were generated by coupling the observed wind forcing to an ocean component. It is found that the Zebiak- Cane model with the COF approach roughly reproduced the 1997 El Nio, but the 2004 El Nio simulated by this approach defied an ENSO classification, ie, it was hardly distinguishable as CP-El Nio or EP-El Ni In both El Nio simulations, substituting the COV with the OFV improved the fit between the simulations and observations because the OFV better manages the time-variant errors in the model. o events even when the observational data (and hence the computational time) were reduced .uch a cost-effective offset of model errors suggests a role for the OFV approach in complicated CGCMs.