Velocity model of a typical geothermal reservoir. There is a fractured zone with very low velocity in the center of the model at about 1000 m depth. 

Context in source publication

Context 1

... computed the synthetic acoustic data with a finite difference code based on a typical reservoir model ( Figure 1). The synthetic datasets consists of 67 shots every 30 m, starting at 1000 m in-line coordinate. The number and position of the geophones was kept constant, starting from 500 m in-line coordinate with 10 m spacing. Each of the 100 traces per shot consists of 1600 time samples with 1 ms sampling time. The zone of interest consists of the slow velocity fractured zone in the center of the velocity model at about 1000 m depth. To investigate the feasibility of imaging the small fractured zone as well as the complex structure of the layers, we computed the synthetics with the fractured zone and without the fractured zone, respectively. Figure 2 shows that the differences are significant, but they are hardly noticeable in the shot gathers only. Since we are purely interested in the structures below 500 m we restrict the seismic processing from this depth downward. For imaging we applied the pre- stack wave-equation migration without any additional processing. The image in Figure 3 shows the clearly defined interface between the two layers including all structural features. The image of the fractured zone is limited to the top of the zone itself. Due to the large velocity contrast between the layers most of the seismic energy is refracted and only little energy is reflected from the fractured zone back to the surface. However, pre-stack migration yields more information in the common image gather, which shows the diffraction from the top of the fractured zone. Hence, the fractured zone can be inferred from diffractions in the common image gather. Detailed information about the seismic surveys in the Coso Geothermal Field can be found in Pullammanappallil et al. (2001) and Bevc. et al. (2002). Initial velocity models were generated based on the picking of first breaks on pre-stack data using seismic turning ray tracing and inverting the data to obtain parameterized velocity tomograms (Pereyra, 2000). The seismic data were preprocessed and statically shifted to a floating datum, applying static corrections and pre-stack energy enhancement (Figure 4). The deeper part of the velocity model was determined with Gamma-Scans and migration velocity analysis of the common image gathers. Based on previous studies line 109 in the Coso Geothermal Field shows a deep reflector at about 16’000 ft, which may be interpreted as the local brittle-ductile zone (Pullammanappallil et al., 2001; Unruh et al., 2001). Hence, we focused our investigation on this depth range only. We applied Kirchhoff pre- stack migration on the preprocessed data assuming two-dimensional geometry. The results show the deep reflector at about 16’000 ft depth extending laterally for about 12000 ft (Figure 5). Since the common mid points of line 109 show small variable cross-line components, we decided to process the data in 2.5-D with Kirchhoff pre-stack migration. Compared to the 2-D migration the stacked image resolves the deep reflector better, which indicates that the subsurface shows significant three-dimensional structures (Figure 6). Line 109 crosses line 110 at 11000 ft in-line coordinate, which allows a 3-D migration for a small image cube of 4000 ft in-line length and 4500 ft cross-line length. However, the velocity model was assumed to remain constant, because the data fold was too small for a three-dimensional velocity analysis. The 3-D migrated image in Figure 7 shows that the deep reflector extends horizontally and recedes in the cross-line direction. More reflectors, which could not be resolved by 2-D migration and 2.5-D migration, become visible. Seismic attributes are inferred physical quantities from seismic data and are commonly used in hydrocarbon exploration and reservoir monitoring. The goal is to find attributes, which are characteristic for a particular area. Hence, we investigated several commonly used attributes for the reflector at 16’000 ft in the Coso Geothermal field to characterize this reflector. Near-offset stacks and far-offset stacks are preferred tools to characterize geological areas. We computed the near-offset stack with 10’000 ft maximum offset and the far-offset stack with 10’000 ft to 24’000 ft offset (Figure 8). In the near-offset stack the deep reflector is not visible at all, whereas the far-offset stack resolves it clearly. This may be due to the physical characteristics of the reflector, but it may also be the result of the seismic processing in this challenging environment. Figure 9 shows the energy at line 109 in the Coso Geothermal Field in the common image gather. The energy was computed between 15’000 ft and 18’000 ft by taking into account the lateral coherency of the reflections. Most of the energy is concentrated at far-offsets between 15’000 ft and 22’000 ft. Since almost all the reflected energy is located in the far-offset data, we computed the energy of the far- offset data only (Figure 10). The computed energy illuminates the deep reflector only, which makes the combination of far-offset data combined with the energy a perfect attribute to locate similar reflectors. Seismic attributes are especially useful in three- dimensional surveys, because they allow a better visualization of geological structures. Figure 11 shows the energy attribute for a small image cube at the intersection of line 109 and line 110 in the Coso Geothermal Field. Compared to the initial migrated image the energy attribute illuminates the three- dimensional character of the reflector and the heterogeneity of the subsurface much more clearly. The frequency power spectrum was computed from 15’000 ft to 18’000 ft depth. The spectrum shows distinct single frequency peaks in the in-line direction (Figure 12). Compared to the source frequency content from 4 Hz to 30 Hz, the frequency content of the reflector is smaller and lies between 4 Hz and 20 Hz. For further investigation of the characteristic peaks of the frequency power spectrum the power spectrum of the reflection response must be computed. To extract the reflectivity power spectrum we applied the procedure: 1) compute the frequency spectrum, 2) divide the frequency spectrum by the source spectrum to obtain the Green’s function, 3) compute the attenuation, 4) compute the Quality factor Q, 5) compensate for the intrinsic attenuation between surface and reflector. We obtained the source spectrum by stacking several direct waves close to the source. The Quality factor Q was computed with linear regression from 8 Hz to 20 Hz with the spectra from 10’000 ft in-line coordinate to 24’000 ft in-line coordinate. We found that the Quality factor Q = 55 fits the data best between the surface and the reflector at 16’000 ft. The reflectivity power spectrum shows similar features like the frequency power spectrum, but the frequency band-with is increased and the frequency peaks are slightly shifted to higher frequencies (Figure 13). These changes become even more apparent, when the spectra are stacked over the in- line direction (Figure 14). After analyzing the attributes we can make the hypothesis that the deep reflection is caused by a horizontal geological structure, which causes distinct frequency peaks in the power spectrum. The simplest model for such a reflector is a single horizontal layer with thickness d . Such a layer produces reflection peaks for λ /d = 1/4, 3/4, 5/4, ..., where λ is the seismic wavelength within the layer. After analyzing the reflectivity power spectrum in the in-line direction, we can conclude that there are single peaks only at each location. Hence, we can assume that the peaks represent the first maximum reflection and therefore the ratio of wavelength and layer thickness is one quarter at each location. Table 1 shows the resulting reflector thickness between 260 ft and 450 ft for the three dominant ...