[M-IS17] 歴史学×地球惑星科学

[MIS17-P10] Analysis on characteristics of surface-observed and model-based cloud cover for historical weather reconstruction using data assimilation

*王 小醒¹、市野 美夏¹ (1.情報・システム研究機構　データサイエンス共同利用基盤施設　人文学オープンデータ共同利用センター)

キーワード：データ同化、歴史文書、日記天候記録、雲量分布、再解析データ

Old diaries precisely recorded locations and dates of weather events during the historical period. Cloud cover amounts can be converted from these descriptive weather records; it remains objective, unaffected by recorder subjectivity. Previous studies (Wang et al., 2022, 2023) have confirmed the feasibility of assimilating cloud cover to reconstruct historical weather. Nevertheless, inconsistencies between surface-observed and model-simulated cloud cover may introduce additional errors in the assimilation process. This study examined the discrepancy between cloud cover converted from surface-observed “general weather conditions” and cloud cover from the reanalysis data over a five-year period starting from 1995. Results indicated that stations showing negative or low correlations in winter are mainly located on the Sea of Japan side. The error originates from the fact that model simulation could not reproduce the high frequency when cloud cover exceeds 90%. On the other hand, results in summer indicated less variability among locations, but the overall correlation coefficient was not high. Overall, it is difficult to reproduce the distribution characteristics of cloud cover observations by model simulation. The attributes in other seasons are currently under examination. The remaining questions are worth pondering: whether it is affected by model resolution, whether the parameter in the model is properly selected, and whether it is possible to directly establish a relationship between weather categories and model-simulated cloud cover. Despite these practical issues, significant differences between surface-observed and model-simulated cloud cover are confirmed in this study. These characteristics also show seasonal variability. Further investigation, including validation with other reanalysis data, will be elucidated during the presentation.