Tristan da Cunha Island is one of the hot spots in the Atlantic Ocean. The discussion about its source have not reached consensus yet whether it is in shallow asthenosphere or deeper mantle, because of lack of the geophysical observations in the area. A marine magnetotelluric (MT) experiment was conducted together with seismological observations in the area in 2012–2013 by collaboration between Germany and Japan, in order to give further constraints on the physical state of the mantle beneath the area. A total of 26 seafloor stations were deployed around the Tristan da Cunha islands and available data were retrieved from 23 stations. The MT responses were estimated for those available sites. The detailed data processing will be presented by Chen et al. in this meeting. In this study, we report on the topographic effect on the observed MT responses. During the cruises for seafloor instruments deployment and recovery, detailed bathymetry data were collected around the stations by onboard multi-narrow beam echo sounding (MBES) system. We compiled the MBES data and ETOPO1 data to incorporate the local and regional topography. Then, we applied iterative topographic effect correction and one-dimensional (1-D) conductivity structure inversion. The MT responses of each station were simulated by three-dimensional (3-D) forward modeling. Preliminary results show the overall feature of the observed MT responses at some stations were qualitatively well explained by the seafloor topography included in the conductivity structure model over the 1-D mantle structure. An extreme example is the station near the Tristan da Cunha Island. The impedance phases varies ~300 degrees in shorter period range which is reconstructed by the 3-D forward modeling. Some implications on the lateral variation in the conductivity of the upper mantle will be discussed by demonstrating the residuals between the MT responses corrected for the topographic effect and the 1-D forward response.