At the Climate Prediction Research Center I work on the development of the Korean Climate Model (K-CM) within the KMA project “Development of the National Climate Prediction System for Climate Crisis Response” (국가기후예측시스템 개발사업, RS-2025-02221669). As part of the coupled-model development team (WG1), my work covers model formulation, conservation and stability, tuning and performance evaluation of the K-CM.
Figure. Roadmap of the national climate prediction system project (2025–2031): a low-resolution coupled model in Phase 1 leads to middle- and high-resolution CGCM and Earth-system models in Phase 2, shared across four working groups.
K-CM is a coupled atmosphere–land–ocean–sea-ice model built on the Korean Integrated Model (KIM). The atmosphere runs on the KIM cubed-sphere grid, and ocean and sea ice use NEMO4/5–SI3. The model is being developed at low, middle and high resolutions.
In Year 1 (2025) the first low-resolution coupled versions capable of long-term integration were built (LRv0.1 → v0.2.1). The Year 2 (2026) goal is K-CM LRv1, the first relatively complete low-resolution version for long climate simulations, together with a middle-resolution prototype.
Figure. (a) Cubed-sphere grid (ne = 16 shown as an example). (b) Vertical levels of the truncated LR model (60) and the original KIM (91).
The KIM atmosphere and Noah land model are coupled through MCT to NEMO ocean and SI3 sea ice, with interfaces for WaveWatch III waves and CaMa-Flood river routing. The 30-year AMIP run of the first version showed a TOA–surface energy imbalance of about 2 W m⁻², traced to underestimated stratocumulus and energy loss in the physics. A 100-year coupled run was numerically stable but drifted slowly warm.
Figure. KIM-based coupled model framework.
Low resolution. Since 2025 the low-resolution coupled model has progressed from the first versions capable of long-term integration to versions with closed mass and energy budgets, improved clouds, and CMIP6-protocol coupled experiments comparable to observations, reanalyses and CMIP6 models. A version has been released for distribution, and tuning with perturbed-parameter ensembles continues toward K-CM LRv1.
Middle resolution. A middle-resolution version is being developed in parallel, together with CMIP long-term integrations (pre-industrial, historical and SSP scenarios) and an ESMValTool-based evaluation tool (KCMValTool) for routine performance checks.
Figure. Total cloud fraction in (a) K-CM LRv0.2, (b) K-CM LRv0.3.1 and (c) MODIS: one example of how the low-resolution model has improved during development.
Poster · Toward K-CM LRv1: Formulation and performance evaluation · ACM 2026
Poster · Accelerating K-CM model development using AI-based climate emulators · KCPS Workshop 2026
Poster · Development of a low-resolution climate model for the national climate prediction system · KMS 2025
Poster · Energy balance of the climate model of the national climate prediction system · KMS 2025 (Choi et al.)
Talk · Accelerating climate model development using AI weather/climate emulators · AGU 2025