アンサンブルスプレッドを活かしたガイダンスの精度向上
概要
Post-processing is an inevitable procedure to statistically correct systematic errors in a numerical model product. Karman filtering, a post-processing technique, often ameliorated the prediction skill by overlearning due to large random errors. This study conducted an experiment to determine the observation error variance from the ensemble spread, based on an idealised numerical model of Lorenz 96. The results suggested that the overlearning was mitigated in spite of synchronisation of systematic errors and the prediction skill was higher than in previous method.