III. DATA ANALYSIS AND INTERPRETATION 3.1. Rift Geometry Identification Based on recent analysis of modern and ancient rifts, many previous workers concluded that the basic structural unit of continental rifts is the halfgraben (e.g., Bosworth 1985; Rosendahl, 1987). Half-grabens generally are faultbounded at both sides, but the fault on the downdip side has much greater throw than that on the updip or flexural side, giving the basin its characteristic asymmetry. In the seismic sections, ideal examples across relatively simple halfgraben units in the Lake Tanganyika and Malawi Rift Zones are shown in Figure 3.1. Figure-3.1: The seismic section across halfgraben units in the Lake Tanganyika (Line 222) and Malawi (Line 817) Rift Zones (Lambiasse, 1988). From the existing 2D seismic sections located in study area, the rift geometry expression can be obviously recognized (Figure 3.2). The base of the rift in the area is bounded by the pre-tertiary basement horizon which is marked by III-1
the abrupt change in the seismic reflection pattern from chaotic or transparent to the layering reflection configuration. Over all the rift basin geometry showing a localized wedge-shape stacking pattern overlaid by a continuous strong reflector with a parallel stacking pattern at the top as the upper boundary. It is clearly that the rift in the study area has a quite good geometrical similarity with those shown in seismic lines across the Lake Tanganyika and Malawi Rift Zones. Figure-3.2: The N-S regional line (a) and SW-NE line (b) seismic sections showing rift geometry in the study area Most of the half-grabens in the area were showing a northward dipping, bounded by its border fault to the north and the flexural margin to the south, with approximately 10 km in wide. Those border fault segments identified along the several N-S seismic sections, apparently they showing an E-W lineament orientation in mapview (Figure-3.3a). In order to reconstruct the rift geometry model, then those border fault segment in each line should be correlated to draw the fault plan. However, as usual working with 2D data for fault segment correlation is not that simple task, since there will be many possibilities to connect one fault segment with others (Figure-3.3b). III-2
Figure-3.3: a. Fault segments trace in study area. b. The cartoon illustrates the classical problems of fault cuts correlation between a set of parallel 2d seismic lines (Freeman et al. 1990) In this case, a modern analog model will be very helpful to create a realistic fault framework interpretation in the study area. Based on a study of the continental rifts architecture in East Africa, Rosendahl (1987) has depicted an idealized half-graben as the intrinsic geometric unit of rift zones, and mentioned that the key to understand the rift morphology and structure is to learn the ways in which half-graben link together to make rift zones. The linkage between two halfgrabens is a structural high known as an accommodation zone or transfer zone. This author has recognized nine different case of half-graben linking arrangements which are grouped into three main families; Family-1: Overlapping, Opposing Half-Graben, Family-2: Non-Overlapping Opposing Half-Graben and Family-3; Similar Polarity Half-Graben (Figure-3.4.) Although this ideal case is unlikely will occur in nature, due to the anisotropy and heterogeneity of pre-rift structural fabrics, however this idealized model is a very good references for the rift geometry reconstruction. III-3
Figure-3.4: The plan view and hypothetical cross sections of an ideal half-graben and examples of linked half-graben families group (Rosendahl, 1987) Based on the existing 2d seismic data, and helped by the modern analog from the East Africa rift zone described above, the fault framework interpretation in study area is most likely matched with the family-3 model; similar polarity half-graben as it shown in figure-3.5. Figure-3.5: The plan view (a) and 3d view perspective (b) of fault frame work interpretation in study area III-4
Further more those fault frameworks modeling result was combined with the seismic interpretation of the base boundary of rift; Pre-Tertiary basement horizon in order to reconstruct the 3d rift geometry. 3.2. Rift Evolution Stages and Their Depositional Units Interval Division Distinct stages of rift evolution can be recognized (Figure-3.6.), each with characteristic linked depositional systems and distinctive expressions on seismic reflection profiles; rift initiation, rift climax, immediate post-rift and late post-rift stages (Prosser, 1993). Figure-3.6: The rift evolution stages and their distinctive expressions on seismic reflection profiles (Prosser, 1993). Based on the similarity with the seismic reflection characteristics above, the rift basin in the study area could be divided into three main depositional units as; rift initiation, rift climax and immediate post-rift (Figure-3.7). Figure-3.7: The seismic reflection profiles of rift depositional unit in study area. III-5
The main characteristics of rift-initiation on seismic section show an overall wedge-shaped geometry, the thin end of the wedge lies high on the hangingwall dip-slope and the internal reflector characteristics are generally hummocky and discontinuous. The rift climax is characterized by an increase in amount of aggradations, together with the development of divergent forms and mounded forms chaotic zones close to the footwall associated with footwallderived fans and talus. Meanwhile the immediate post-rift is marked by the end of divergent reflector packages, the succeeding reflectors may show a continuation of the preceding aggradations and may show a greater proportion of associated progradation. Further more those seismic reflection characteristics described above were used as the guidance parameter to divide the whole rift section into more detailed rift depositional units in the study area that later on will be used as boundaries for the modeling processes. 3.3. Lithologic Facies Recognition From Well and Seismic Attributes L46-1 well is considered to be the key well used for this study since it penetrated the complete synrift section and successfully discovered the hydrocarbon in the synrift play in the nearest area. The dominant lithologies penetrated in the synrift section consist of the fluvial conglomerate, coarse to fine sand with some shale intercalation at the lower part, this section interpreted to be the rift initiation unit. The next sediments package above is composed by finer grain clastic sediments dominated by the lacustrince shale with some coal intercalation and some medium to fine sand, this interval is considered to be the rift climax unit. The overlaying sediment above this unit is the intercalation between sand and shale deposited in the fluvio-deltaic environment which interpreted to be the immediate post-rift section. On top of it was overlaid by the deep marine shale and carbonate which considered as the post-rift package. The next process carried on this research is recognizing those lithologic facies from the seismic data. By tying the L46-1 well with its intersecting seismic line (Figure-3-8), in general the seismic characters of the synrift section in L46-1 well can be obviously recognized. The top of synrift package (red horizon) which marked by the facies changes from marine carbonate to the fluvio-deltaic shale in III-6
the well section can be picked in the boundary under the continuous strong reflection which most likely the marine carbonate even. Meanwhile the base of the synrift unit which marked by the lithology changes from metasediment to fluvial clastic facies can be inferred in seismic by the reflection pattern changes from chaotic-irregular pattern to something that more layered pattern bounded by a slight angular unconformity feature (magenta horizon). Figure-3.8: The L46-1 well seismic tie and lithologic-facies reconnaissance in seismic data. Furthermore, the quantitative relationship between the lithologic-facies and seismic data were achieved by cross-plotting the lithology information from well versus the extracted multi attributes data from seismic such as instantaneous frequency, cosine phase, envelope, and etc within the synrift section. After some trial and iterations, in this case the seismic amplitude envelope showing a very good correlation with the lithology (figure-3.9). The amplitude envelope attribute is the total instantaneous energy of the analytic signal (the complex trace), independent of phase. It is also known as Instantaneous Amplitude, Magnitude or Reflection Strength which can be useful for detecting sand bodies, channels, and hydrocarbon anomalies. Based on this cross-plot, the seismic amplitude envelope values range for each lithologic classes can isolated, and inferred that the smaller seismic III-7
amplitude envelope value correspond to the higher probability of having more sandy facies and the bigger seismic amplitude envelope value, correspond to the higher probability of having the more shaly or coaly facies. Later on this information will be used as a direct analogue for the simplified lithology prediction in geo-cellular modeling process in study area (figure-3.9). Figure-3.9: The amplitude envelope vs lithologic class probability cross-plot and the simplified lithology prediction 3.4. Geocellular Modeling In general the geo-cellular modeling can be divided into two main steps, the structural modeling and the property modeling. The interpretation results of the rift geometry and their depositional unit boundaries described in the previous section were used as the inputs for the structural modeling, aimed to generate the skeleton and the geometrical constrain for the geo-model. The structural modeling step were consists of several processes, i.e.; the fault modeling, 3D grid generation, and stratigraphic layering. Meanwhile the extracted seismic amplitude envelope was used as the input for the property modeling to populate and predict the lithologic-facies distribution. III-8
The geometric structural model parameters used in this research is shown in the figure 3-10. The top and base horizon of the synrift were used as the most upper and lower vertical modeling boundary, then it s divided into three more detailed zones of each synrift depositional units called the rift initiation, rift climax and immediate postrift zone. Further more the 3D grid was created for those zones with the 201 m x 205 m x5 m dimension. The size of modeled area which is located in the central area of interest is about 24,882 m in length, 17,087 m in width and 3,739 m of depth. This size will end up with about 7. 6 millions number of 3D grid cells. Figure-3.10: The 3D structural geometry modeling parameters. The next process carried out was to extract the seismic amplitude envelope attribute from 2D existing seismic line in study area, then those attribute values were resample by the 3D grid cells created in the previous step. The 3D grid cells intersected by those 2D seismic lines will be filled out by the attribute values of amplitude envelope (figure-3.11). Those filled grid cells were used as the main data control point for the property modeling processes to build the 3D litho-facies III-9
distribution model using the geostatistical approaches. Compared to the conventional interpolation-extrapolation techniques, the main advantages of the geostatistic method are it will take into the spatial relationship of the data sample in the calculation, quantify the uncertainty and it could provided many realization output which enable us to have multi-scenario models. There are many available algorithm, however the geostatistic algorithm applied in this research are the Object Modeling and the SGS techniques. Figure-3.11: The 3D grid cells filled by amplitude envelope values re-sampled along the existing seismic 2D lines. In theory, the Object Modeling is one of the geostatistic methods which have the capability to integrate the data and conceptual geological model. This method allows any discrete facies to be populated stochastically and honors the deterministic geological setting inputs parameter, such as the channel direction, channel geometry, sediment source direction and so on. So it will be very useful in the area where the data controls is limited, since we still could bring the conceptual geological input to take into account in the modeling calculation. For this purpose then those 3D grid cells filled by amplitude envelope values were III-10
classified using the value range in the previous step where each class will be correspond to its predicted litho-facies as it shown in the previous figure-3.10. Then each class will be coded to a discrete variable and will populated to the entire modeling grid using the deterministic geological setting inputs parameter constrain as shown in the figure-3.12. So in this case, the expected modeling result will be give a distribution pattern which follow the deterministic geological setting input and still honoring the data control. Figure-3.12: The geological setting inputs parameter of the object modeling in study area Another modeling technique called the SGS (Sequential Gaussian Simulation) was also applied in this research as alternative workflow. The SGS is one of the geostatistical stochastic conditional simulations to distribute the continuous variable. The different with the object modeling, the SGS modeling III-11
results will give many distribution pattern alternative outputs which determined by the spatial relationship of the data control. By using this workflow, it is expected that we can have a distribution model that are geologically realistic and honors the data control value and its statistical parameters. In the SGS workflow the first step carried out is the data variogram analysis. The aim is to find out the spatial relationship of the data, such as the major range (maximum distance of the spatially related data point), minor range, major direction (the direction of the most continuous data distribution trend), minor direction, nugget and sill. Those spatial relationship parameters will be taking into account in the calculation of the property modeling. In this research the seismic amplitude envelops is used as the continuous variable data input. The variogram analysis was carried out in each syn-rift depositional units, the results is shown in the following figure-3-13. Figure-3.13: The variogram analysis results in each syn-rift depositional units III-12