[M-GI24] Data assimilation: A fundamental approach in geosciences

[MGI24-P05] Ocean Data Assimilation Focusing on Integral Quantities Characterizing Observation Profiles

*杉浦 望実¹ (1.国立研究開発法人　海洋研究開発機構)

キーワード：４次元変分法、海洋データ同化、反復積分、シグネチャ

In our implementation of a global long-term ocean state estimation (aka, ESTOC), the representation of heat and material balance on a climatological scale is of importance. For this reason, as data assimilation method, we have employed a strong-constraint four-dimensional variational approach with a long assimilation window on several years scale. However, the traditional method of assimilating individual observations at each spatial point may not adequately estimate the macroscopic heat and material balance under conditions where the model's representation error is significant. Therefore, we are attempting to fundamentally revise the conventional observation operator, shifting from assimilation at the level of individual observation points to assimilation at the level of observation profiles. We expect that this can be achieved by representing the vertical profiles of observations and the model in terms of iterated integrals (or signature). We have already implemented such a mechanism in our data assimilation system.
In this presentation, we report the results of a prototype assimilation experiment that assimilates observation profiles as the unit of assimilation. The assimilation window spans ten years, and only Argo profiles are assimilated. For comparison, we also conducted a control experiment using the conventional method of assimilating individual observation points. We will describe the outline of the data assimilation method and results, including the diagnoses of steric sea level and water mass distribution.