[M-GI33] Data assimilation: A fundamental approach in geosciences

[MGI33-12] Iterative ensemble variational methods and its application for the prediction of geomagnetic secular variation

*中野 慎也¹、南 拓人²、高橋 太³、松島 政貴⁴、清水 久芳⁵、藤 浩明⁶ (1.情報・システム研究機構　統計数理研究所、2.神戸大学大学院理学研究科、3.九州大学大学院理学研究院、4.東京工業大学理学院地球惑星科学系、5.東京大学地震研究所、6.京都大学大学院理学研究科地磁気世界資料解析センター)

キーワード：Data assimilation、Ensemble variational method、Geomagnetic field、Geomagnetic secular variation

The 4-dimensional ensemble variational method (4DEnVar) is a data assimilation method which is easy to implement as a post-process. The iterative version of the 4DEnVAR, which is sometimes referred to as the iterative ensemble smoother, is also regarded as a useful tool for nonlinear data assimilation problems. Since these methods are derived based on a linear approximation of a dynamical system model, the behavior of the algorithm under ths situations with large uncertainties is not trivial. In this study, the accuracy of the algorithm was evaluated after considering second and higher order terms of the Taylor expansion of a system model. The sufficient conditions for finding a local maximum of the objective function are then explored, and the behavior of the algorithm under the situation with nonlinearity is discussed. The applications for the prediction of geomagnetic secular variation based on the above approach will also be introduced.