| 講演要旨(和文) | | MT法は，電磁探査法の一種であり地下深部の比抵抗構造が導出可能である。しかしながら，取得した電場・磁場データにはノイズが混入し，地下の比抵抗構造の解析が困難となることがある。そこで，混入したノイズを除去し信号成分を抽出する必要がある。本研究では，周波数領域独立成分分析（FDICA）という，近年，混合した音声の分離に用いられている手法に基づく，新たなノイズ除去手法を開発した。FDICAの性能を評価するために，陸上MTデータに人工的なノイズ付加を行い，本手法のノイズ除去性能の評価を試みた。 |
|
|
| | 講演要旨(英文) | | Electric and magnetic field data are obtained by Magnetotelluric (MT) surveys, and often contain noises. If the noises are strong and coherent between electric and magnetic field, MT response functions derived by conventional method are far from true values. This can cause large errors in the inferred subsurface resistivity structure. In this study, we focused on Frequency Domain Independent Component Analysis (FDICA), which has been used for sound analysis. FDICA was applied for MT data at Kakioka Magnetic Observatory. As a result, compared with conventional method, apparent resistivity curves derived by FDICA were the almost same but the estimated error became smaller. To evaluate noise reduction performance of FDICA, two hypothetical noises were added to MT data at Kakioka Magnetic Observatory, then FDICA was applied for them. The efficiency was improved greatly by FDICA, better than the conventional method. |
|
|
|
| |
|