Year: 2019
Type: dissertation
Abstract: In this study, the optimal ensemble size and the factor affecting its determination for Sub-seasonal to Seasonal (S2S) prediction are explored. The results show that the prediction skill, which is quantified by the mean square skill score (MSSS), increases with increasing ensemble size in a perfect model with an unbiased and reliable ensemble spread to observed variance. In this idealized prediction system, 10 to 20 ensemble members lead to 90% to 95% of the skill improvement in the theoretical maximum (infinite ensemble size). This theoretical estimation is applied to the European Centre for Medium-Range Weather Forecasts (ECMWF) S2S real-time forecast that consists of 51 ensemble members. The MSSS of the 500-hPa geopotential height, which represents tropospheric skill, is in good agreement with the theoretical estimation, indicating that approximately 10 to 20 members are needed to obtain a skill improvement close to the theoretical maximum. However, the stratospheric skill, verified at the 50-hPa geopotential height, is substantially lower than the theoretical estimation. This result is particularly true in the tropical stratosphere, where only 20-30% skill improvement of the theoretical maximum can be obtained even when using all ensemble members. The substantial overestimation of the stratospheric skill is mainly due to model mean bias. The role of mean bias in the ensemble size effect is highly dependent on ensemble spread and natural variability. By removing the bias, overestimation in the stratosphere can be reduced so that it is possible for 10-20 members to be optimal. Extending the results to all the levels and seasons available in the S2S dataset indicates that the effect of mean bias is remarkable in the tropical and summer hemispheric upper stratosphere.
Author: 김상욱
Institution: Seoul National University
Language: English
Subfield: Atmospheric Science
Field: Earth and Planetary Sciences
Domain: Physical Sciences
SDG: Climate action
Open Access status: green
OpenAlex ID: W7170021305
