229 7. Seismic inversion and characterization applied to geothermal energy Figure 7.5 Multi-well and multi-traces approach for a unique wavelet estimation 7.2.3 Construction of a low-frequency model Seismic information is limited by its bandwidth, both in low and high frequencies. The high-frequency limit determines the resolution, or the maximum precision of the final inversion results. But there are also low frequencies missing, including compaction trends and other regional changes, that need to be modeled through a low-frequency model. In practice, this model is simple, so that any complex features retrieved from seismic inversion can only come from seismic data. Interpreted horizons allow the construction of a structural model, defining different units, and correlation lines. Following the correlation lines the model is obtained by propagating the acoustic impedance values (and S-impedance and density for elastic inversion) from the wells in calibrated positions along the correlation surfaces defined during the creation of the structural grid. The extrapolation method used is inverse distance. Then, a low-pass filter is applied to eliminate high frequencies coming from the wells, limiting the model at the missing part from seismic data. The model can be used with several “intensity”, playing a strong role for each following proposal: 1. low-frequency model: An elastic model, in the low frequencies missing from the seismic, is used, either before (initial model) or after (by adding the missing frequencies). 2. initial model: The starting point of the optimization is an elastic model. 3. prior model: During optimization, impedance is compared to a prior model, from which it costs the algorithm to deviate (Tonellot et al., 1999). For (1), the frequency filter, for its higher limit, must be strict with no overlapping with the dataset. The frequency overlap is possible for (2), and even recommended
RkJQdWJsaXNoZXIy NjA3NzQ=