17:15 〜 18:30
[SEM12-P03] On synthetic marine magnetotelluric inversion study: reproduction of error in data
キーワード：海底マグネトテルリック探査、人工データインバージョン
Inversion studies using synthesized data is useful to test a hypothesis that one would like to investigate. Synthesized data should contain errors and they should as realistic as possible to maximize the reliability of the hypothesis test. And therefore, it is important to know the characteristic of the real data. In this study, I investigated a marine MT data set collected in the northwestern Pacific through Normal Oceanic Mantle project (Baba et al., 2017). The data set consists of MT impedance tensors at 24 sites in the period range between ~50 to ~100,000 seconds, which were obtained by processing approximately one-year-long time series data using BIRRP program (Chave and Thomson, 2004). The errors of the MT impedance are generally smaller for the major off-diagonal elements than for the minor diagonal elements. Also, they are smaller in the middle range of the period and become larger to the shorter and longer period ends, which is related to number of samples available to estimate the impedance. Plotting the relative error of the MT impedance to the magnitude of the MT impedance for all tensor elements, all periods, and all sites, I found an empirical linear relation with a smallest limit. This relation can be used to synthesize the error of data. I run one-dimensional inversion for two synthetic data which have constant relative error and error based on the empirical relation, respectively. The comparison of uncertainties in the obtained models gives information how synthetic inversion with the too much simplified assumption for the data error overestimates the sensitivity to the model.