Network-magnetotelluric method and its first results in central and eastern Hokkaido, NE Japan

A new field observation technique based on the magnetotelluric (MT) method has been developed to determine deep and large-scale 3-D electrical conductivity distributions in the Earth. The method is named 'Network-MT', and employs a commercial telephone network, to measure voltage differences with long dipole lengths ranging from 10 to several tens of kilometres. This observation configuration enables us to obtain the telluric field distribution with nearly continuous coverage over a target region. Response functions are estimated between the respective voltage differences and the horizontal magnetic fields at a reference point. Owing to the long electrode spacing, the observed responses are relatively free from the effects of small-scale near-surface heterogeneity with a scalelength shorter than the typical electrode spacing. Therefore, physically meaningful direct comparison between the observations and model responses is feasible even if the fine-scale features of near-surface heterogeneity are ignored. This extensively reduces the difficulty, especially in 3-D MT interpretation. The first Network-MT experiment was performed in central and eastern Hokkaido, NE Japan, in 1989. It took about five months to complete all of the measurements, and used 209 dipoles to cover the target area of 200(EW) x 200(NS) km(2). The long electrode spacing enabled us to obtain the voltage differences with a high signal-to-noise ratio. For 175 dipoles, the squared multiple coherency between the voltage difference and the horizontal magnetic field at Memambetsu Geomagnetic Observatory was determined to be more than 0.9 in the period from 10(2) to 10(4) s. 193 MT impedances were computed in tensor form by linear combination of the response functions. The estimated impedances generally possessed smooth period dependence throughout the period range. No drastic spatial change was observed in the characteristics of the tensors for neighbouring sites, and some regional trend could be detected in the spatial distribution. Thus, we confirmed the merit of the Network-MT method, that its responses are little affected by small-scale near-surface structures. The regional feature of the response implied a significant influence of the coast effect, and was well correlated with the regional geological setting in Hokkaido. Conventional Groom-Bailey tensor decomposition analysis revealed that the target region is not regionally one- or two-dimensional. Therefore, we developed a 3-D forward modelling scheme specially designed for the Network-MT experiment, and tried to reproduce the Network-MT responses directly. In the 3-D model, a realistic land-sea distribution was considered. The resistivity of sea water was fixed to be 0.25 Ohm m and, as a first trial of 3-D modelling, the resistivity of the land was assumed to be uniform and its value was determined to be 300 Ohm In by a simple one-parameter inversion. Overall agreements between the observations and the best-fit model responses indicated the importance of the 3-D coast effect in the target region. However, there remained significant discrepancies, especially in the phase of the responses, which provide a clue to determining a regional deep 3-D structure.