Executive Summary

Considering the limited human and computing resources in the domestic modeling community, strengthening our modeling capacity and continuously operating the testbed within the joint collaboration system (K-R2O process) are imperative to enhance the effectiveness of the climate prediction system. This report aims to strengthen the Korean Meteorological Administration (KMA) Global Seasonal Forecast System version 6 (GloSea6) and includes three tasks: (i) the development of land-surface model improvement technology for KMA GloSea6, ii) the evaluation of the possibility of applying the developed R&D technology to operations (testbed), and iii) the framework of verification system that suggests guidance for the models’s improvement.

To improve the prediction skill of KMA GloSea6, a groundwater algorithm and river routing model in land surface processes were investigated and developed. To enhance the land surface model JULES within GloSea6, the JULES-GrUB system was developed by integrating the GrUB algorithm. The GrUB algorithm replaces the existing TOPMODEL-based groundwater runoff scheme, offering improved applicability in ungauged basins. Performance evaluations were conducted for seven major basins, including the Mississippi, Amazon, Yangtze, Mekong, Murray-Darling, Gobi, and Tibetan regions. Comparisons with FluxCom and GLEAM benchmark datasets demonstrated that the GrUB algorithm significantly enhanced prediction accuracy, particularly in latent heat flux and soil moisture. The study utilized data from 2008, 2009, and 2010, with initialization on March 1, focusing primarily on spring. While the results confirmed the effectiveness of the GrUB-based model, further evaluation through long-term simulations and experiments across various seasons is necessary. This research highlights the potential of the GloSea6 JULES-GrUB system in improving hydrological and energy component predictions, contributing to enhanced reliability in climate modeling.

The characteristics of the current river routing model were also investigated, and its weaknesses were identified. It was found that GloSea6 uses a relatively simple river routing model, and the simulated river storage is overestimated compared to observations. To simulate accurate river flow and air-land-sea interactions, it is most desirable to couple a sophisticated and realistic river routing model. As a simple method, attempts were made to reduce the amount of freshwater flowing into the ocean by increasing the resolution of the existing river routing model. The GloSea6-TRIP system, which replaces the existing 1-degree resolution with 0.5-degree and 0.125-degree resolution for TRIP, was newly constructed. As the resolution of the river routing model changed, the setting coefficients for meandering and river flow velocity were also optimized. The new system clearly reduces errors in river flow and discharge compared to the existing operational system. Simulations with the optimized high-resolution river routing model showed reduced errors in ocean circulation and temperature, especially in the Pacific and Indian Oceans. These results are valuable for improving seasonal prediction due to more accurate air-sea interactions and for applied and policy research on water resources.

To more accurately predict the rapidly changing climate, it is essential to have a climate prediction model equipped with state-of-the-art technology. Proactive validation of cutting-edge technologies can help reduce potential risks in the operation of current climate prediction systems. The APEC Climate Center (APCC) is participating in the joint development of the KMA’s GloSea6 as part of the testbed, which serves as a bridge for the rapid acceleration of forecast technologies to forecast operations. In this study, research-to-operation (R2O) processes for climate prediction technologies were established, fulfilling the testbed's role.

First, a structured joint development system with stages and elements suited to our country’s circumstances was proposed. Key improvements in the joint development research environment were also suggested. On a small scale, these improvements establish a solid testbed foundation, and on a larger scale, they aim to activate joint development of KMA’s climate prediction system. In particular, regarding the testbed, experimental standards (semi-operational forecast experiments), research processes (application of new technology → semi-operational experiments → evaluation → suggestions for operational applicability), and evaluation perspectives (scientific effect and technical efficiency) were established, structuring a system to fulfill the testbed’s role. Additionally, a collaborative roadmap for model development reflecting the long-term demands of R2O was established through discussions based on the APCC-KMA-NIMS (National Institute of Meteorological Science) framework. The proposals made in this study were compiled into guidance documents and provided to NIMS, responsible for the forecast operation of the joint development of the climate prediction system. This systematic approach is expected to drive the growth of collaboration aimed at improving long-term forecasting in our country.

Second, by performing the role of a testbed, the feasibility of implementing new technologies in the operational forecast system was proposed. For developed prediction technologies from R&D to be applied to the operational forecast system, evaluation in an environment identical to the actual operational system is required. Therefore, the testbed evaluations were conducted under conditions identical to those of the operational system, assessing both scientific effectiveness and technical efficiency to determine operational stability. Technologies evaluated in this study included parameterization of sea ice, coupled initialization, and cumulus parameterization. In 2022 and 2023, the effects of parameters in sea ice were verified, and optimization strategies were proposed. Sensitivity experiments were conducted on parameters influencing winter temperatures in East Asia to verify their predictability. The correlation between Arctic sea ice and winter temperatures in East Asia was also improved. Operational efficiency and stability were confirmed, demonstrating the technical feasibility of applying these technologies in actual operational forecasts.

Additionally, the effects of coupled initialization were demonstrated. It was confirmed that coupled initialization improves climate factors influencing summer precipitation, and simulation performance related to winter tropical variability was also enhanced. From a technical perspective, stable integration was confirmed, and although the runtime may be slightly longer than that of the operational forecast, it was judged to be generally suitable for actual operations. In 2024, the effects of a newly developed convection scheme, the Unified Convection scheme (UNICON), on simulated East Asian summer monsoons, teleconnections, and extreme precipitation were analyzed. The advantages and disadvantages of UNICON were examined, and future development directions were proposed. Results from the testbed were shared with NIMS in the form of five technical reports. Notably, proposed technologies related to sea ice physics and coupled initialization were applied to the operational KMA GloSea6 in 2024.

The APCC’s testbed serves as a bridge between model developers and operational institutions, accelerating the virtuous cycle of “fundamental technology development → feasibility evaluation → application to the operational system.” It is becoming a hub for domestic collaboration that promotes the practical application of research results on climate models. Through this study, APCC has established an environment capable of quasi-operational forecast experiments and has built expertise by accumulating years of research know-how using GloSea6. In other words, the APCC’s testbed, with its comprehensive experimental environment and research capabilities, plays an irreplaceable role in the climate prediction modeling domain—one that no other entity can fulfill.

The verification framework, called CrEMA (Climate foRecast model Evaluation system by APCC), was developed to provide an objective assessment of the operational climate forecast system and diagnose structural problems and their causes within the model. CrEMA consists of two parts: skill assessment of various climate variables and diagnostics for major climate variability. The core of CrEMA is to construct metrics that effectively capture the multi-faceted characteristics of the climate system. The skill assessment metrics were designed to focus on quantitative measures of deterministic and probabilistic forecast skills through statistical scores. The diagnostic metrics were independently developed for each climate mode and variability, including the El Niño-Southern Oscillation (ENSO), Madden-Julian Oscillation (MJO), Boreal Summer Intraseasonal Oscillation (BSISO), Arctic Oscillation (AO), Sea Ice Extent (SIC), and East Asian Summer/Winter Monsoon (EASM/EAWM)—key components of East Asian climate variability. These metrics diagnose the dynamical modes and physical interactions within the model's climate system. Through the CrEMA diagnostic metrics, the performance of Arctic, EAWM, and BSISO predictions in the KMA GloSea6 system was comprehensively diagnosed, and suggestions for model improvement were provided.

According to the CrEMA Arctic diagnosis, the GloSea6 generally captures the vertical structure of the stratospheric polar vortex weakened by increased Eurasian snow cover with reasonable accuracy. However, it underestimates the intensity of the stratospheric polar vortex, particularly the downward wave propagation in the lower mid-latitude troposphere during winter. This deficiency impacts the performance of Arctic Oscillation (AO) predictions, which are notably significant only for 1-month lead times, and reduces the skill in simulating teleconnection related to AO. On the other hand, GloSea6 accurately simulates the interannual and intraseasonal variability of Barents-Kara sea ice, demonstrating a significant improvement over GloSea5. However, the model underestimates the surface heat flux due to the reduction in sea ice and exhibits excessive stability in the lower Arctic troposphere. This stability hampers the development of Arctic warming associated with the decline of Barents-Kara sea ice, resulting in a shallow and weak warming structure that prevents upper-level wave propagation. Consequently, this limits the model’s ability to simulate Arctic-Eurasian teleconnections. To improve GloSea6's Arctic performance, it is crucial to address the excessive stability in the lower Arctic atmosphere. Further improvements are also needed in the Arctic ocean-atmosphere and the troposphere-stratosphere interactions.

The East Asian monsoon, which significantly influences seasonal climate variability in East Asia, was also diagnosed. The diagnostic results of the East Asian winter monsoon show that the mean bias and the interannual variability of monsoon system components have improved, especially in the vertical temperature and geopotential height over East Asia. However, issues such as cold temperature biases and an overestimated East Asian trough still remain. The variability of the Siberian high, which is crucial for the East Asian winter monsoon, is also not adequately simulated. In tropical-mid-latitude teleconnections, delayed responses of East Asia to the Indian Ocean Dipole (IOD) have been reproduced with reduced mean bias in the tropical Indian Ocean. However, there is no significant change in the response to the ENSO. In Arctic-mid-latitude teleconnections, the responses of East Asia to the AO and Barents Oscillation (BO) have been reproduced adequately. The predictability of major modes of winter temperature variability over East Asia is not valid for more than two months, and the prediction of blocking patterns and intensity affecting cold surge frequency has slightly improved. Further improvements are needed in simulating the development and propagation of the Philippine Sea anticyclone and local atmosphere-sea interactions in the tropical western Pacific, which are crucial for Pacific-East Asia teleconnections.

GloSea6’s representation of the BSISO exhibits significant improvements, particularly in northward propagation over the Indian Ocean region, attributed to reduced biases in convection and easterly wind shear. The model demonstrates enhanced simulation of key physical processes, including moisture advection and barotropic vorticity, leading to better predictions of BSISO northward propagation. Notably, the model better captures East Asian meteorological patterns associated with BSISO phases, especially in predicting warm events linked to BSISO1 Phase 7 and BSISO2 Phase 1, and wet events connected to BSISO1 Phase 4. Nevertheless, limitations persist in the western Pacific region with persistent biases in outgoing longwave radiation (OLR) and zonal wind, significantly weakened intraseasonal variability beyond three-week lead times in both 30-60-day and 10-20-day modes, and limited representation of teleconnection impacts, particularly in simulating the barotropic structure associated with the Circumglobal Teleconnection (CGT) pattern. These findings suggest that while GloSea6 shows substantial improvement in BSISO simulation, further development should focus on convective activities in the western Pacific, maintenance of variability in predictions, and teleconnection processes.