Executive Summary

This study aims to develop new predictive content to address extreme climate events, focusing on seasonal forecasts based on extremes and physical quantities, as well as linking the characteristics of monthly and daily precipitation. To produce seasonal prediction information, we first analyzed the distributional properties of observational and climate prediction datasets and examined climatological biases—mean, range, and shape—embedded in dynamical models and grand ensemble data. Various bias-correction methods were applied to align ensemble distributions with the physical distributions observed, thereby improving the match between raw climate prediction data and observations. Nevertheless, the predictability of seasonal forecasts based on extremes and physical quantities yielded somewhat divergent outcomes. To integrate monthly and daily precipitation characteristics, correlations between monthly accumulated precipitation and both precipitation frequency and intensity were investigated. Regional differences were identified, with precipitation intensity generally showing a stronger influence on monthly precipitation variability than frequency. We further assessed the significance of the difference between the two correlations to identify regions with a high potential for extreme precipitation, and interpreted these area and seasons based on their characteristics precipitation regimes. These findings highlight the importance of bias correction and multi-scale precipitation analysis in enhancing the reliability of seasonal climate predictions, particularly under extreme climate conditions.

APEC (Asia-Pacific Economic Cooperation) Climate Center has been conducting comprehensive analyses of climate forecast models across a wide range of aspects. Analysis of various climate modes that represent seasonal scale climate variability can serve as an essential component for understanding climate prediction models, and they can also be used to interpret the sources of seasonal predictability and the physical processes linking remote teleconnections to regional climate variability. In this study, we evaluate the predictability of major oceanic and atmospheric climate modes simulated by APCC MME and investigate their characteristics. By doing so, we aim to derive approaches for interpreting model forecast information and ultimately enhance the usability of seasonal predictions. Additionally, we analyze the most recent hindcast period to better account for contemporary climate change, as well as the real-time forecast period available since 2012 to evaluate the model’s performance under operational forecasting conditions.

The oceanic climate mode was selected in the tropical Pacific, North Atlantic, and Indian ocean. For the tropical Pacific climate mode, we analyzed predictability by month and lead time using various ENSO indices. The prediction skill with respect to ENSO phase, spatial structure, and event evolution was also examined. For the North Atlantic, we analyzed the monthly and lead-time-dependent prediction skills of sea surface temperatures in the mid-and high-latitude regions, where the strong variability area in the SST tripole pattern, and assessed model biases in the climatological mean SST field. For the Indian Ocean, monthly and each lead-month correlations were analyzed to quantify the forecast skill of both the time-series evolution and spatial patterns of the IOD and IOB. For the atmospheric climate modes, we evaluated the prediction skills of the AO, NAO, NP, PNA, WP, and SOI indices, and analyzed their teleconnection responses to temperature and precipitation, as well as their relationships with ENSO. Based on these results, the MME seasonal prediction information is interpreted to identify factors that can influence temperature and precipitation over Korea, and the findings are applied to support seasonal prediction operations at KMA.

The APCC seasonal forecast is produced based on a Multi-Model Ensemble (MME) approach, utilizing data contributed by the world’s climate leading climate forecasting operational and research institutes. This year, key system upgrades were implemented, including expansion of participating models, extension of the hindcast climatology period, and provision of high-resolution verification data. Collaborative improvements and the addition of new models have strengthened prediction reliability, while enhanced data quality management was achieved through error identification and feedback sharing among participating institutions. As a result, noticeable improvements were observed in global temperature and precipitation hindcast skill, and the applicability of a new ENSO index accounting for climate change effects was explored. Additionally, automation of the APCC three-month outlook system improved operational efficiency and established a foundation for high-resolution climate prediction for Korea. These advancements and strengthened cooperation are expected to enhance the reliability of seasonal prediction and support effective international collaboration.

The Boreal Summer Intraseasonal Oscillation (BSISO), which originates over the equatorial Indian Ocean and propagates northeastward, is a key source of subseasonal variability affecting the Asian summer monsoon, convection, and large-scale atmospheric circulation. APCC provides real-time BSISO monitoring, prediction, and verification information each year from May to October. The operational system is run daily based on outgoing longwave radiation (OLR) and 850hPa wind fields, following procedures of data collection, quality control, monitoring and prediction production, and real-time system monitoring. Using five participating prediction models, the system produces and delivers various BSISO products, including phase diagram, time-series, reconstruction field, and impact anomaly. In 2025, APCC established its own BSISO input-data processing system, enabling direct use of ECMWF model forecasts within the BSISO operational framework. This improvement ensures the continued provision of ECMWF-based BSISO information to the Korea Meteorological Administration’s (KMA) monthly forecast briefing. Furthermore, in preparation for the future operationalization of APCC’s MME subseasonal prediction system, a framework was developed to utilize APCC’s raw prediction data as BSISO input fields once available, thereby strengthening both the potential for expanding participating models and the autonomous production and application of BSISO prediction information.

APCC’s in-house prediction model, SCoPS (Seamless Coupled Prediction System), is and atmosphere-ocean-sea ice coupled model with a horizontal resolution of approximately 80km and has been used as a participating model in the APCC MME seasonal prediction system since November 2017. Each month, SCoPS generates initial conditions using NCEP CFSR and Argo observations, applies atmospheric and oceanic initialization, and produces 6 months seasonal forecasts. The resulting forecasts undergo post-processing and verification and used both for the internal forecast briefing and as a participating model in the APCC MME seasonal prediction system. In 2025, the APCC MME hindcast period was updated from 1991-2010 to 1993-2016, and this new hindcast period was applied to the verification and forecast information produced for the SCoPS forecast briefing.

Since 2006, the WMO Lead Centre has served as a one-stop shop by collecting and standardizing Seasonal prediction data produced by Global Producing Centres (GPCs), applying various ensemble techniques, and providing seasonal prediction information. Forecast data are generated around the 15th of each month and made available through the WMO and the Korea Meteorological Administration websites. Currently, 15 institutions participate as GPCs, producing multi-model ensemble seasonal prediction information for nine variables using both deterministic and probabilistic methods. Whenever individual GPC models are upgraded, the standards processed by the WMO Lead Centre may change, requiring continuous data verification and program improvements. In 2025, four institutions upgraded their models, and the Lead Centre modified its programs to ensure stable service.

The Fire and Haze Early Warning system provides forecast information to Indonesia and Malaysia by calculating probabilistic risk levels based on regional rainfall predictions. Statistical downscaling and bias correction techniques are applied to tailor the forecast data to local conditions, with risks classified into four levels and visualized in map format. While previous methods provided risk information by region, the new high-resolution forecasts generate grid-based risk assessments. After a trial operation period, the high-resolution service will be officially launched, with errors identified during testing corrected to deliver regionally optimized, stable, and highly useful climate services.

Seasonal prediction information from the WMO Lead Centre is also used in regional climate forums to present outlooks. In 2025, the Lead Centre participated in ASEANCOF, organized by ASMC, and SASCOF, organized by the India Meteorological Department, to discuss climate services for Southeast Asia and South Asia. Furthermore, suggestions were made to expand the use of seasonal prediction information across diverse sectors such as agriculture, water resources, health, and disaster management.