Executive Summary

APCC has contributed to the long-range forecast operated in KMA for the last several years (since 2017). We have continuously improved technical affairs and decision making process necessary for long-range forecast. In this year, we have tried to improve long-range forecast through the objective process. The objective process in this year includes the selection of monitoring factors for seasonal forecast; therefore, we choose the three most significant monitoring factors for the boreal winter temperature forecast.

To improve the information related to the ENSO in 3-month long range forecast, impact of the tropical precipitation which act as a oceanic ENSO forcing is investigated. The leading EOF mode of the precipitation over the tropical Pacific and Indian ocean is a ENSO related pattern. The seasonal forecast model also reproduce the similar pattern as the observed mode. However, in the model precipitation, the ENSO related leading mode takes larger ratio than the observation, also model could not simulate precipitation pattern related to the intra-seasonal variation. Within the winter season, relationship between ENSO and tropical precipitation has monthly variation, and the atmospheric responses to the tropical precipitation is changed depending on the precipitation region and month. The model precipitation has highly related to the ENSO, and the wave train from the tropical Indo-Pacific to the mid-latitude East Asia is simulated dominantly throughout the winter season. The model could not show simulated variations over the Eurasia, which are important impacts during late winter season in the observation. The results of this study suggest that adding both the analyzed result associated to the ENSO and precipitation from the observation and information from the other prediction predictors to the model forecast information is necessary in the late winter month forecast.

As one of the important monitoring factors for boreal winter temperature prediction, we analyzed the availability of the index describing austral Eurasian snow cover. The index has its own increasing trend so that we consider it after detrending. For the early winter, the seasonal prediction model can reasonably represent the temperature variability induced by the surface albedo effect. However, for the late winter, the model cannot describe the tropospheric-stratospheric interaction during the hindcast period. Therefore, we use the detrended index for a longer observation periods instead of considering its response from the seasonal model prediction.

We analyzed the usage of Arctic predictors for predicting monthly temperature during winter, mainly about Arctic sea ice and Arctic vertical distribution. First, we analyzed whether various autumn Arctic sea ice precusors are still practical and accurate even in consideration of climate change. Next, in order to replace the less predictable sea ice precusors, the possibility of utilizing the Arctic vertical temperature distribution, which were presented based on the last year’s study, was explored. Then, the predictive performance of the Arctic vertical temperature distribution event was analyzed in the operational seasonal prediction model, GloSea6. Through this, we examined whether the model can simulate the Arctic vertical temperature distribution events and the characteristics of the model related to it. Finally, based on the research results, we suggested a plan to utilize the predictive information related to the Arctic at a 3-month outlook.

To improve the accuracy and efficiency of 1-month long range forecasts, we have objectified the methodology that forecasters use significantly when looking at 1-month long range forecasts. We examined the potential for improved predictability using these methods. Drawing inspiration from the method of utilizing the outcomes of the medium range prediction model for 1-month forecasts, we conducted experiments on various ensemble selection methods. By using this ensemble selection method, it was shown that if the ensemble is selected at the +2 week point, the superiority of predictability continues until the +3 week prediction. The effect of ensemble selection was particularly evident in the improvement of temperature probability prediction results in the mid-latitude inland region. Through a comparison with the experiment of randomly selecting ensembles, we showed that optimal ensemble selection has a more significant effect on improving predictability than merely doubling the number of ensembles.

In the practice of weekly mean forecast for 1 month in the APCC, it is often observed that consistency (same anomaly at the same location and same target time) between consecutively issued recent forecast anomalies were considered as a signal of credibility of the forecast. Sometimes, extrapolation of tendency of forecast anomaly were attempted due to reletively small amplitude of ensemble mean forecast after 2-weeks lead time. An attempt has been made to quantify the consistent tendency and usefulness of its extrapolation. It is found that extrapolating recent tendency of the forecast anomaly does not guarantee the improvement of the forecast quality as less than 50% of cases showing consistent tendency can contribute to reducing forecast error.

The Madden-Julian Oscillation (MJO), a prominent intraseasonal oscillation during the winter season in equatorial regions, manifests differently in the temperature of East Asia as it propagates across the major convective centers from the western Indian Ocean to the tropical Pacific. First, we investigated the influential phase and associated dynamic mechanisms that exhibit a significant correlation with observed winter temperatures in South Korea. The large-scale convection occurring near tropical maritime continents can induce a propagating wave pattern extending from Southeast Asia to the northeast over the course of a week, creating a teleconnection pattern that affects temperature in South Korea. Meanwhile, we assessed the simulated performance of phase/temperature skill scores, tropical convection activity, and associated atmospheric patterns. In particular, we selectively identified optimal forecast information by considering model-simulated characteristics of MJO amplitude and the duration of tropical forcing from the +3week forecast.