Model Forecast Error Correction Based on the Local Dynamical Analog Method: An Example Application to the ENSO Forecast by an Intermediate Coupled Model

Numerical forecasts always have associated errors. Analog correction methods combine numerical simulations with statistical analyses to reduce model forecast errors. However, identifying appropriate analogs remains a challenging task. Here, we use the Local Dynamical Analog (LDA) method to locate analogs and correct model forecast errors. As an example, an El Niño–Southern Oscillation (ENSO) intermediate coupled model forecast error correction experiment confirms that the LDA method locates high quality analogs of states of interest and improves the model forecast performance, which is due to the initial and evolution information included in the LDA method. In addition, the LDA method can be applied using a scalar time series, which reduces the complexity of the dynamical system. The LDA method is a promising method to locate dynamic analogs and can be applied to existing numerical models and forecast results. Earth‐science models are important tools in the analysis of physical processes and in forecasts of future conditions. However, numerical models always contain forecast errors. Model forecast error in historical data may appear again. Thus, the historical model forecast error can be used to correct the forecast results of focused states, which can reduce the model forecast error without building the new numerical model. The key question is how to locate suitable historical model forecast errors for the focused states. In this paper, we use the Local Dynamical Analog (LDA) method to locate the model forecast error and firstly correct the model forecast results. In the ENSO prediction experiment by an intermediate coupled model, the LDA is proved the advantage over other analog‐locate methods to find analogs and improve the whole forecast skill and the ENSO event forecast. The improvement from the LDA method in the root squared mean error skill is significant, and the forecast intensity of ENSO events is closer to observation than that of the uncorrected forecast. Analogs based on historical forecast can be used to estimate model forecast error and correct model forecastsThe Local Dynamical Analog method can locate high quality analogs based on the initial and evolutional informationAnalog correction using the Local Dynamical Analog improves ENSO forecast skill, significantly in root‐mean‐squared error skill