| Summary: | One important factor that affects the performance of statistical downscaling methods is the selection of appropriate parameters. However, no research on the optimization of downscaling parameters has been conducted in South Korea to date, and existing parameter selection methods are dependent on studies conducted in other regions. Moreover, several large-scale predictors have been used to predict abnormal phenomena such as droughts, but in the field of downscaling, parameter optimization methods that are suitable for drought conditions have not yet been developed. In this study, by using the K-nearest analog methodology, suitable daily precipitation downscaling parameters for normal and drought periods were derived. The predictor variables, predictor domain, analog date size, time dependence parameters, and parameter sensitivity values that are representative of South Korea were presented quantitatively. The predictor variables, predictor domain, and analog date size were sensitive to the downscaling performance in that order, but the time dependency did not affect the downscaling process. Regarding calibration, the downscaling results obtained based on the drought parameters returned smaller root mean square errors of 1.3–28.4% at approximately 70% of the stations compared to those of the results derived based on normal parameters, confirming that drought parameter-based downscaling methods are reasonable. However, as a result of the validation process, the drought parameter stability was lower than the normal parameter stability. In the future, further studies are needed to improve the stability of drought parameters.
|
|---|