The State-of-the-art Remapping Process for the Korean Integrated Model (KIM)

This study introduces the remapping process of the state-of-the-art Numerical Weather Prediction (NWP) model, the Korean Integrated Model (KIM) system, developed by the Korea Institute of Atmospheric Prediction Systems (KIAPS). The KIM employs the quasi-uniform and pole singularity-free spherical geometry, a cubed-sphere grid, essentially requiring the grid transformation in the pre- and post-process. The remapping process of the KIM consists of two parts: (1) figuring out reference points from source grid points and weights at each reference point and (2) calculating the values at destination points by multiplying the values at reference points and appropriate weights. The KIM uses the linked-cell mapping algorithm to search the reference points around the destination point. The KIM provides various interpolation methods including the alternative-bilinear algorithm (Kim et al., 2019), which can obtain the weights from irregular grid points without assuming a rectangular shape. We are developing an advanced pre- and post-process that downscales cubed-sphere grids to a higher resolution using neural network techniques as an ongoing project. We also plan to share our results that improved the efficiency of the remapping process, which requires a massive resource compared to the simplicity of its calculation. For example, only the address and weights of reference grid points required for the interpolation are stored in a remap matrix (pruning) as an efficient storage managing strategy. Moreover, the advantage of parallelization, evaluated in the Korean 5th super-computer, will also be presented. Through parallelizing the remapping process, the speed improves up to about twelve times, and a more stable calculation speed is expected.