Enhancing ENSO Prediction Skill by Combining Model‐Analog and Linear Inverse Models (MA‐LIM)

To enhance El Niño–Southern Oscillation (ENSO) forecast skill, we devise a model analog (MA)‐linear inverse model (LIM) by nudging sea surface temperature and sea surface height anomalies forecasted by the LIM into the MA. The performances of the LIM, MA, and MA‐LIM are compared to general circulation model simulations and observations. At short (long) lead month τ, the LIM (MA) predicts the Niño 3.4 SST anomalies better than the MA (LIM). On the other hand, the MA‐LIM shows the best performance at all τ. At τ=6, the MA performs better than the LIM in the eastern equatorial Pacific and Indian Oceans but worse in other regions. The MA‐LIM substantially remedies the undesirable aspects of the MA. The success of the MA‐LIM appears to come from the use of more accurate initial conditions than the MA and an ad hoc implementation of seasonal cycle and nonlinearities into the LIM through nudging to the MA. Plain Language Summary The El Niño–Southern Oscillation (ENSO) is the most important tropical atmosphere‐ocean phenomena driving changes of weather and climate over the globe. Two empirical forecast methods—the linear inverse model (LIM) and model analog (MA)—are known to have nearly the same ENSO forecast skill as general circulation models. To further enhance the ENSO forecast skill, we develop a model analog‐linear inverse model (MA‐LIM) by combining these two empirical methods. We found that the MA‐LIM performs much better than individual LIM and MA. Thus, the MA‐LIM can be effectively used to improve ENSO predictions. Key Points To enhance ENSO prediction, we combine two forecast methods by nudging a linear inverse model (LIM) with a model analog (MA) Overall, the MA‐LIM forecasts tropical sea surface temperature anomalies better than the LIM and MA at all lead months This success is due to the use of more accurate initial condition than MA and an implementation of seasonal cycle and nonlinearity into LIM