Velocity model determination by the SMART method, Part 2: Application SP3.8
Résumé
The SMART (Sequential Migration Aided Reflection Tomography) method, as explained in the first part of this paper, starts after a first set of traveltimes in the unmigrated prestack data has been picked and the inventarization of useful a priori knowledge related to these traveltimes has been made. Thereto a preparative phase is needed. First a global estimate of the subsurface structure is made. Hereto we use the standard stacking and poststack interpretation procedures which 'allow for getting insight in the degree of complexity of the subsurface. Next the traveltimes can be picked. When interpreting prestack data important qualitative structural information in difficult target zones (e-g. fault zones or salt structure flanks) can be obtained. Such an analysis guides the interpreter in selecting and picking the best traveltimes of primary events. Once the preparation is finished the SMART method can be applied for a detailed determination of a structural and velocity model in a very consistent way. It is emphasized that velocity variations in complex structures can be determined accurately by prestack traveltime inversion techniques. This phase has an iterative character. In order to update the velocity model after the first iteration additional traveltimes are needed. Next additional traveltimes are obtained by interpretation of the cube of migrated data which can be easier than in the time domain due to the focussing and positioning effect of the migration process. By tracing rays in the same velocity model as was used for mi.gration on the newly interpreted events, we will obtain additional traveltimes which will make the set of input data for the next iteration of tomography more complete. A new velocity model is calculated and the data are remigrated. In this paper we will demonstrate the feasibility of this approach using a 2D real data set. We executed a number of iterations of the SMART method and ended up with of the complex structure. a very satisfactory depth image THE DATA We used for this application a 2D dataset covering a salt structure. It consists of 300 shotrecords at a regular interval of 40m. The acquisition was done in a split spread. The half spread length is 1920 meters with 48 geophones. The data were delivered with a standard preprocessing (filtering, zero-phase deconvolution and muting). Because of some clearly visible groundroll, we applied a second filter in order to remove most of this in Figure 1. low frequency noise. A partial stack of the data is shown THE PREPARATIVE PHASE Analysis of complexity In order to get an idea of the degree of complexity of a subsurface, it is useful to construct several partial stacks with the same stacking velocity model. Because the stacking process is based on flattening of the hyperbola's in CMP's, through some NMO and DMO based correction, differences in between the partial stacks demonstrate the failure of the process. In areas with complex subsurface structures these hyperbola's aren't necessarily flat due to different raypaths left and right of the midpoint. In this dataset this phenomenon can be observed in a series of CMP's covering the saltdome (See Figure 2). Another way to get an idea of the complexity is to do a post stack depth migration by a layer stripping approach using the best partial stack. For these data the results are satisfactory for the sedimentary zones left and right of the dome, but are incorrect for the deep interfaces and the base of the salt. This is partially due to events that are lost during the stacking procedure. Other causes for this failure are: the uncertainty in picking the right interface that serves as the next velocity boundary and the difficult choice of the velocities which becomes more and more hazardous as the depth increases. The final result is unreliable and the resulting depth for the base of the salt depends largely on the choices made by the interpreter Clearly these data cannot be handled by standard processing techniques. Left and right of the salt dome and below it the nature of the trace gathers is too complex. A prestack imaging method using a velocity model computed by tomography seems adequate for solving the aforementioned problems. Data preparation for the SMART method The next step after the analysis of the complexity is the data preparation for the SMART method. Its goal is to prepare an initial set of traveltimes to be used in the first iteration. We split this phase in a number of consecutive sub-phases: • Creating a initial set of guides for the prestack interpretation. • Picking traveltimes. • Quality control of the traveltimes. • Selection of representative traveltimes and calculation of the associated weights. Creating a set of guides. Guides are indicators for the interpreter suggesting where to look in the prestack unmigrated data for a certain event. They are also warnings for complicated situations as multiples, triplications and situations were no reliable indications for the nature of an event is available. The geologic guides are qualitative (e.g. presence of a fault) or quantitative (e.g. the depth of horizon A is 2500m). The geophysical guides are for example the presence of multiples or diffractions. They are derived from the unstacked or stacked data. For this dataset the following data were used: a set of (partial) stacks, time-and depth-migrated stacks and the cube of preprocessed prestack data. It allowed us to determine the zones where picking traveltimes directly in the unmigrated data could lead to incorrect traveltime information for the tomography. These zones are indicated in Figure 1 (Za and Zb, a zone with triplications and a series of unexplained events. Picking the first set of traveltimes Using the guides the picking of the traveltimes can start. This is done in the cube of unmigrated data. There is no preference for picking in a specific trace gather. This depends of the available guide. When it is a geological one the common offset gathers are most suited. Using a geophysical one the interpretation is done in the shotgathers or the common midpoint gathers. Whatever direction is chosen, one has to end 1425
Origine : Fichiers produits par l'(les) auteur(s)