Bayesian inversion of marine controlled source electromagnetic data offshore Vancouver Island, Canada

Abstract

S U M M A R YThis paper applies nonlinear Bayesian inversion to marine controlled source electromagnetic (CSEM) data collected near two sites of the Integrated Ocean Drilling Program (IODP) Expedition 311 on the northern Cascadia Margin to investigate subseafloor resistivity structure related to gas hydrate deposits and cold vents. The Cascadia margin, off the west coast of Vancouver Island, Canada, has a large accretionary prism where sediments are under pressure due to convergent plate boundary tectonics. Gas hydrate deposits and cold vent structures have previously been investigated by various geophysical methods and seabed drilling. Here, we invert time-domain CSEM data collected at Sites U1328 and U1329 of IODP Expedition 311 using Bayesian methods to derive subsurface resistivity model parameters and uncertainties. The Bayesian information criterion is applied to determine the amount of structure (number of layers in a depth-dependent model) that can be resolved by the data. The parameter space is sampled with the Metropolis-Hastings algorithm in principal-component space, utilizing parallel tempering to ensure wider and efficient sampling and convergence. Nonlinear inversion allows analysis of uncertain acquisition parameters such as time delays between receiver and transmitter clocks as well as input electrical current amplitude. Marginalizing over these instrument parameters in the inversion accounts for their contribution to the geophysical model uncertainties. One-dimensional inversion of time-domain CSEM data collected at measurement sites along a survey line allows interpretation of the subsurface resistivity structure. The data sets can be generally explained by models with 1 to 3 layers. Inversion results at U1329, at the landward edge of the gas hydrate stability zone, indicate a sediment unconformity as well as potential cold vents which were previously unknown. The resistivities generally increase upslope due to sediment erosion along the slope. Inversion results at U1328 on the middle slope suggest several vent systems close to Bullseye vent in agreement with ongoing interdisciplinary observations.

Trans‐dimensional Bayesian inversion of controlled‐source electromagnetic data in the German North Sea

“…Model parameters are often dependent on each other and might only be resolved in combination (e.g., the product of resistivity and layer thickness; Edwards 1997). Other implementations of Bayesian algorithms for marine CSEM in fixed dimensions either constrain the subsurface resistivity layering using depths inferred from seismic data such as hydrocarbon reservoir depth (Chen et al 2007), severely constrain the prior parameter widths (Buland and Kolbjørnsen 2012), or use the Bayesian information criterion to estimate the most probable number of sub-seafloor layers, which is held fixed in the inversion (Gehrmann et al 2015). A widely applied technique is Occam's inversion (Constable et al 1987), which parametrises the model using a large number of interfaces at fixed depths such that layer thicknesses are below the resolution of the data, and introduces a regularization term minimizing the second depth derivative of resistivity to constrain the result to a minimum-structure model.…”

Effects of the sea floor topography on the 1D inversion of time‐domain marine controlled source electromagnetic data

“…Gehrmann et al . () and Moghadas et al . () investigated the non‐uniqueness problem by sampling the model parameter space.…”

“…Moreover, the models that can interpret the controlled source electromagnetic method (CSEM) data, are non-unique because of the diffusive nature of electromagnetic fields. Gehrmann et al (2016) and Moghadas et al (2015) investigated the non-uniqueness problem by sampling the model parameter space. In order to take into account the problem of the starting model of the Marquardt inversion, we have used different starting models with different layer thicknesses and resistivities.…”

“…Moghadas, Engels and Schwalenberg ; Gehrmann et al . ). Topography problems in time‐domain CSEM are also seldom discussed.…”

“…Therefore, 1D inversion schemes in time-domain CSEM are still widely studied and used (e.g. Moghadas, Engels and Schwalenberg 2015;Gehrmann et al 2016). Topography problems in time-domain CSEM are also seldom discussed.…”