We demonstrate how processing data from shallow water CSEM surveys using up-down separation can improve the performance of a global inversion scheme. Data from a receiver over a known prospect produces a markedly improved reproduction of the resistivity profile in a planelayer model employed for illustrative purposes. This improvement being particularly pronounced in the absence of a strong resistive anomaly, the results are directly applicable to finding starting models for more rigorous 3D inversion schemes as well as a to creating a reference model for Scanning survey interpretation.
The 1D inversion of marine CSEM data is an easy way to generate depth-resistivity profiles required by more advanced data processing and interpretation schemes. Examples are the generation of starting models for rigorous 3D inversion or the interpretation of Scanning (i.e. reconnaissance) survey data (Wahrmund et al., 2006). The inversion of marine CSEM data is inherently illconditioned, in particular in shallow water (water depth <500m), where the strong air wave dominates the measured electromagnetic field at large source-receiver offsets, thus masking the response from deeper resistors/hydrocarbon reservoirs (Roth and Maaø, 2007). Amundsen et al. (2006) introduced an effective method which attenuates the air wave and increases the sensitivity of marine CSEM methods by separating the measured wavefield into its upand downward traveling constituents. Here we present example results from inverting shallow water CSEM data acquired offshore Norway using a simple 1D inversion scheme that combines the sensitivity enhancement of updown wavefield separation with the global optimization capabilities of a simulated annealing (SA) search algorithm.
Air wave attenuation by up-down separation takes advantage of the fact that the information about the subsurface is contained in the upward traveling constituent of the wavefield in the seafloor just below the CSEM receiver, whereas the air wave is traveling downward. Assuming a primarily vertically traveling wavefield, as is the case for large source-receiver offsets, the separation can be applied on a receiver-by-receiver basis using a simple linear combination of the measured electric and magnetic field components. Here, the superscript (U) denotes “upward”, ?SF is the resistivity of the seafloor, ? is angular frequency, _ denotes the magnetic permeability. Similar expressions exist for the upward constituents of E yand H x, respectively. The seafloor resistivity needs to be known a priori, however Roth and Maaø (2007) showed that the decomposition relation (1) is well-behaved and tends to enhance the sensitivity to resistive subsurface structures even when the assumed resistivity is incorrect. We therefore propose to use the same seafloor resistivity in the up-down separation as in the top-most layer of the inversion model. This approach renders the problem more non-linear as compared to keeping the resistivity fixed a priori, which favors the use of a global inversion scheme such as SA over gradient-based methods.
The technique of simulated annealing (SA) was invented in the early-mid 1980’s, based on the Metropolis algorithm, and has since become a tool in most fields of computational optimization.