Assessing the Korean Integrated Model (KIM) for MJO Prediction Performance

The Madden-Julian oscillation (MJO) is the dominant mode of tropical atmospheric
variability on sub-seasonal time scale, characterized by a periodic eastward propagation of
large-scale atmospheric disturbances, including enhanced convection, cloudiness, and relative
atmospheric circulation. Understanding and accurately predicting the MJO is crucial due to its
significant impact on global weather systems. Accurate MJO forecasts have the potential to
improve the predictability of weather patterns weeks in advance, providing valuable insights
for various socio-economic sectors. The Korea Institute of Atmospheric Prediction System is
actively working to extend the forecast range to 30 days using the Korean Integrated Model
(KIM), which is developed by KIAPS and serving as the operational global numerical weather
forecast model for medium-range forecasting by the Korean Meteorological Administration.
This study aims to evaluate the performance of the KIM in simulating the MJO by utilizing
extended-range hindcast datasets, with the goal of identifying the characteristics of KIM in
MJO prediction. The performance of the MJO simulation is assessed using the Real-time
Multivariate MJO (RMM) index. Results indicate that as the prediction lead time increases, the
MJO amplitude in the ensemble prediction mean decreases, while the phase gradually
accelerates. Additionally, KIM datasets were divided into two groups based on the performance
of individual initializations, and MJO performance was compared as a function of prediction
lead time. The analysis revealed that the bivariate correlation (BCORR) of the high-skill group
decreases gradually, whereas the BCORR of the low-skill group declines more sharply as
prediction lead time increases. Furthermore, the study investigated the impact of
teleconnections in the Norther Hemisphere during each MJO phase across both groups. These
findings provide foundational data that can contribute to the improvement of MJO prediction
in future iterations of the KIM model.