[A-CC20] 雪氷学

[ACC20-P09] 雪氷圏大気陸面結合系におけるアンサンブル予報誤差共分散と誤差相関の構造

*鈴木 和良¹、Milija Zupanski²、Dusanka Zupanski³、沖 大幹⁴ (1.国立研究開発法人海洋研究開発機構、2.Colorado State University、3.Zupanski Consulting LLC、4.東京大学生産技術研究所)

キーワード：アンサンブルデータ同化、積雪モデル、降雪、単一観測実験

Coupled numerical models address interactions between processes in the atmosphere, ocean, land surface, biosphere, chemistry, cryosphere, and hydrology. Including the interactions between such processes can potentially extend the predictability and eventually help in reducing the uncertainty of the prediction. Coupled data assimilation is a branch of data assimilation that deals with coupled modeling systems. In this article the fundamentals of coupled data assimilation are first described through a mathematical example of a model including two coupled components. Then, through a series of single observation experiments, we analyze the forecast error covariance and correlation structures using the Maximum Likelihood Ensemble Filter (MLEF) data assimilation system with coupled atmosphere-land surface Weather Research and Forecasting (WRF) model . The atmospheric WRF component has been coupled with two land surface models: Noah and Noah-MP. Two observation locations with different precipitation regimes have been considered. Through this study, we found that error covariance and correlation were dependent on both location and land surface scheme. Snow precipitation likely caused more complex structures in error covariances and correlations compared to the precipitation-free site. The employment of a more realistic snow model was found to reduce the error covariance and error correlation between the atmosphere and the soil in the coupled system. We also have demonstrated, for the first time in a data assimilation study, that correlation structures can be useful in understanding the physical meaning of the forecast error covariance and as a basis for selecting the most important forecast error covariance components for the coupled data assimilation system. Overall, the complexity and structure of ensemble-based forecast error covariance appears to be meaningful, which is encouraging for the future applications of coupled atmosphere – land surface data assimilation.