Reservoir porosity controls the strategies for reservoir management. Porosity is the primary key to a reliable reservoir model. The most economic method of evaluating reservoir porosity on a foot-by-foot basis is from core and well log data analysis. Lateral reservoir porosity is estimated using geostatistical method from well log data or from the integration of well log data and seismic data. However, the petroleum industry needs more accurate, reliable methods to estimate porosity from seismic data. Neural network analysis is one of the latest technologies available to the petroleum industry.

In this study, I report results of an investigation of the use of neural network to predict lateral reservoir porosity. The approach is based on using average seismic acoustic impedances extracted from a 3D seismic volume to predict lateral average porosity for 13 reservoir geological layers. A neural network was trained using different subsets of well log data from 9 hydrocarbon wells and validated using the reminder of the wells. Data from the Unayzah reservoir in CNR field located in central basin of Saudi Arabia was used in this study.

Model-based post-stack seismic inversion was used to produce a seismic acoustic impedance volume. Average impedance maps were then created for 13 layers from the Unayzah reservoir interval in the CNR field.

Back-propagation neural network technique successfully estimated lateral reservoir porosity from seismic acoustic impedance and density attributes. The neural network performance using data from 6 wells (C, D, F, G, I, J), more or less distributed along the field axis, provided a better correlation and less scatter than other well training geometries in the testing phase. The A, B, and H wells were used for validation. Goodness of fit was 0.9985. The good neural network prediction in the testing phase reflects the neural network capability to estimate average reservoir porosities. Predicted lateral porosity maps incorporate heterogeneities introduced by the seismic data, and correlates with the seismic and geological interpretations.

Neural network results show that neural network method can be used to predict lateral reservoir porosities, provided neural network can be trained on the available wellbore data of that reservoir before application to seismic data.

---

The objective of this study was to describe depositional and diagenetic characteristics of the Oxfordian (Jurassic) Smackover Formation in the Womack Hill field, Alabama, as part of an integrated reservoir description program. In order to understand the distribution of reservoir units, this study utilized an integrated array of data from core lithological descriptions, borehole logs, core reports, thin section petrography, porosity and permeability measurements on core plugs, and mercury injection capillary pressure (MICP) measurements. These data made it possible to establish reliable measures of reservoir quality by comparing pore geometry with pore type; then determining which pore types correspond with highest porosity-permeability paired values. Pore aperture (throat) median sizes measured by mercury capillary pressures were tested for correspondence with porosity, pore type, permeability, and saturation in order to establish quality rankings for the reservoir units.

From this information, a method was developed to identify best, intermediate, and worst reservoir quality zones. These zones were then put in a stratigraphic context, allowing for a better understanding of the effects of pore categories, original depositional texture, and diagenetic influences on the distributions of the reservoir quality zones. Correlation of this information between wells allows for the three-dimensional mapping of these quality zones, or petrofacies units. The final results are maps predicting areas with optimal recovery potential and/or bypassed pay.

This study attempts to bridge the gap between petrological and petrophysical studies, merging the data into a comprehensive model. This model, instead of mapping facies based solely on lithology, or flow units based only upon permeability, depicts petrofacies, with each petrofacies having distinct porosity, permeability, and capillary pressure ranges. These characteristics are related to specific pore types and diagenetic processes. The resulting combinations are grouped into reservoir quality rock, baffle or poor quality reservoir material, and reservoir seals.

---

An important fraction of the reservoirs in the Outer Continental Shelf of Gulf of Mexico is comprised of thin-bedded deposits from channel-levee systems. These reservoirs are particularly difficult to describe. Not only is their architecture complex but the quality of the reservoir is determined by connection and length of beds below the resolution of usual reflection data. Improved characterization is needed to improve development and production of these reservoirs. This study presents an integrated approach to build a geologically consistent reservoir model, based on the 8 sand reservoir in Northern Green Canyon block 18. The underlying idea of the construction of this model is that reservoir quality is influenced more by the internal architecture than by the statistical values of petrophysical parameters.

Seismic interpretation and attribute extraction provided the reservoir geometry and stratigraphy. The structural framework and the limits of the reservoir have been determined, showing the preeminent role of salt and faults in the constitution of this reservoir.

Seismic attributes are calibrated to extract areal information on reservoir architecture. Gross thickness and net thickness maps have been estimated using geostatistical methods. Lateral variations in the quality of the 8 sand and the definition zones with different average properties were inferred from geostatistical results.
Lithofacies characterization from core showed that 3 facies could be used to describe the internal variability. The fine-scale heterogeneity is described in each zone from vertical facies distribution determined from wells.
A truncated Gaussian sequential simulation was performed to reflect both the regional trend and the internal variability on a 150*150*1 ft grid.

The major contribution of this work is to show the efficiency of this approach to describe complex reservoirs were the impact of internal variability is a major control of flow efficiency. This is especially valuable when the well information is scarce or not uniformly distributed. This model will be used for flow simulation and sensitivity analysis to improve the field description.

---

The majority of faults display a layer of crushed wear material ("fault gouge") between the fault blocks, which influences the strength and stability of faults. This thesis describes the results of a numerical model used to investigate the process of comminution in a sheared granular material. The model, based on the Discrete Element Method, simulates a layer of 2-D circular grains subjected to normal stress and sheared at constant velocity. An existing code was modified to allow grains to break when subjected to stress conditions that generate sufficient internal tensile stresses. A suite of five numerical runs was performed using the same initial system of grains with sizes randomly chosen from a pre-defined Gaussian distribution. A range of confining pressures was explored from 4.5 MPa to 27.0 MPa (in case of quartz grains with average diameter of 1 mm). The average effective friction coefficients of the five simulations were relatively unaffected by comminution and displayed a constant value of about 0.26. The amount of breakage was directly related to both the applied confining pressure and logarithm of the displacement along the fault. The particle size distribution evolved during the runs, but it was apparently determined only by the cumulative number of grain breakage events: two runs with the same number of breakage events had identical particle size distributions, even if they deformed to different extents under different stress conditions. These results suggest that the knowledge of both the local displacement and stress state on a fault can be used to infer the local particle size distribution of the gouge.

---

The Spraberry Formation is traditionally thought of as a deep-water turbidite sand formation in the center of the Midland Basin. At Happy Spraberry Field, Garza County, Texas, however, production is from a carbonate interval about 100 feet thick that has been correlated on seismic sections with the Leonardian aged, Lower Clear Fork Formation. The "Happy field" carbonates were deposited on the Eastern shelf of the Midland basin and consist of oolitic/peloidal grainstones and packstones, rudstones and floatstones, in situ Tubiphytes bindstones, and laminated to rippled, very-fine grained siltstones and sandstones. The highest reservoir "quality" facies are in the oolitic grainstones and packstones where grain-moldic and solution-enhanced intergranular porosity dominate. Other pore types present include incomplete grain moldic, vuggy, and solution-enhanced intramatrix.

The purpose of this study was to relate pore geometry measured by digital petrographic image analysis to petrophysical characteristics, and finally, to reservoir quality. Image analysis was utilized to obtain size, shape, frequency, and total abundance of pore categories. Pore geometry and percent porosity were obtained by capturing digital images from thin sections viewed under a petrographic microscope. The images were transferred to computer storage for processing with a commercial image analysis program trademarked as Image Pro Plus (Version 4.0).

A classification scheme was derived from the image processing enabling "pore facies" to be established. Pore facies were then compared to measured porosity and permeability from core analyses to determine relative "quality" of reservoir zones with different pore facies. Pore facies are defined on pore types, sizes, shapes, and abundances that occur in reproducible associations or patterns. These patterns were compared with porosity and permeability values from core analyses. Four pore facies were identified in the Happy field carbonates; they were examined for evidence of diagenetic change, depositional signatures, and fractures. Once the genetic categories were established for the four pore facies, the pore groups could be reexamined in stratigraphic context and placed in the stratigraphic section across Happy field. Finally, the combined porosity and permeability values characteristic of each pore facies was used to identify and rank good, intermediate, and poor flow units at field scale.

---

Time-dependent fluid-assisted mechanisms such as stress-induced dissolution and subcritical crack growth play an important role in porosity reduction and compaction of granular material in nature. Previous compaction creep experiments on loosely packed, well-rounded quartz sand at subcritical effective pressure show that the rates of compaction increase with presence of pore water, effective pressure (confining pressure minus pore pressure), and temperature due to operation of fluid-assisted mechanisms. We have investigated the role of cracking during creep compaction of quartz sand by monitoring acoustic emissions (AE). Experiments on water saturated St. Peter quartz sand packs (255 ± 60 mm grain size, initial porosity ~32%) and quartz powder made of disaggregated Arkansas novaculite grains (35 ± 12 mm grain size, initial porosity ~40%) were conducted at a pore fluid pressure of 12.5 MPa, at effective pressures, Pe, of 15 to 105 MPa, and at temperatures, T, of 24 to 225 o C. Volumetric strain rates were allowed to decrease to approximately 10 -8 s -1 before experiments were ended. Our experiments show decelerating creep after application of effective pressure, that volumetric strain, b, increases approximately linearly with log time, and that b rates are dependent on Pe and T. The ratio AE to b is constant with time at low T and Pe, but at high T and Pe the ratio decreases with time. We interpret this to indicate that, for higher Pe and T, there is a gradual transition in the dominant strain mechanism with time from critical crack growth to an acoustically quiet mechanism, such as subcritical crack growth or pressure solution.

---

Deformation and mineral alteration adjacent to a 2 km long segment of the Kern Canyon fault near Lake Isabella, California are studied to characterize the internal structure of the fault zone and to understand the development of fault structure and constitution and the mechanical and chemical processes responsible for them. The 140 km long Kern Canyon fault (KCF) is a fault of 15 km right-lateral separation exhumed from seismogenic depth that cuts batholithic and metamorphic rocks of the southern Sierra Nevada. The fault consists of at least three distinct phases: an early phase of lower-greenschist-grade ductile shear with an S-C' phyllonite, a subsequent, dominant phase of brittle faulting characterized by a through-going zone of cataclastic rock, and a late stage of minor faulting along discontinuous, thin, hematitic gouge zones. The S-C' fabric and subsidiary fault-slip data indicate that both the phyllonitic and cataclastic zones are approximately vertical and strike-slip; slip lineations within the hematitic gouge suggest oblique-slip. The phyllonite zone trends N20-40E and accommodated ~175 m of separation. The cataclastic zone cuts the phyllonite, trends N21E, and consists of foliated and non-foliated cataclasites; it accommodates the majority of displacement along the fault. Abundant veins and fluid-assisted alteration in the rock surrounding the fault zone attest to the presence of fluids of evolving chemistry during both ductile and brittle faulting. Mass balance calculations indicate quartz loss during phyllonite faulting and imply that the fault system was open and experienced a negative change in volume during phyllonite faulting. Mesoscale and microscale fracture intensities decrease with log distance from the foliated cataclasites and approach a relatively low level at approximately 500 m. The internal structure of the Kern Canyon fault is similar to other large displacement faults in that it consists of a broad zone of fractured and altered rock and a narrow zone of intense cataclasis.

---

Understanding the internal organization of the Lower Pleistocene 8 Sand reservoir in the Green Canyon 18 field, Gulf of Mexico, helps to increase knowledge of the geology and the petrophysical properties, and hence contribute to production management in the area. Interpretation of log data from 29 wells, core and production data served to detail as much as possible a geological model destined for a future reservoir simulation.

Core data showed that the main facies resulting from fine-grained turbidity currents is composed of alternating sand and shale layers, whose extension is assumed to be large. They correspond to levee and overbank deposits that are usually associated to channel systems. The high porosity values, coming from unconsolidated sediment, were associated to high horizontal permeability but generally low kv/kh ratio.

The location of channel deposits was not obvious but thickness maps suggested that two main systems, with a northwest-southeast direction, contributed to the 8 Sand formation deposition. These two systems were not active at the same time and one of them was probably eroded by overlying formations. Spatial relationships between them remained unclear. Shingled stacking of the channel deposits resulted from lateral migration of narrow, meandering leveed channels in the mid part of the turbidite system. Then salt tectonics tilted turbidite deposits and led to the actual structure of the reservoir. The sedimentary analysis allowed the discrimination of three facies A, B and E, with given porosity and permeability values, that corresponded to channel, levee and overbank deposits. They were used to populate the reservoir model. Well correlation helped figure out the extension of these facies.

---

The early-evolution of fault structure is inferred from analysis of detailed maps of portions of strike-slip faults with uniform displacements ranging from mm to decimeter in porous quartzose sandstone. Emphasis is on assessing the spatial relationship between the progressive addition of subsidiary fault segments (deformation bands) and earlier-formed linkage structures. The along-strike variability and distribution of fault structure are documented and used to assess the role of early fault geometry on the evolution of fault structure.

The study faults evolve in the initial stage by linkage of an early-formed array of isolated, en echelon fault segments with small relative spacing that step opposite to the sense of shear. The initial configuration of early fault segments favors the development of a geometrically distinct set of linkage structures denoted as Type 2 linkage structures. A simple Type 2 linkage structure consists of two, curved, overlapping and mutually abutting, synthetic extensions of the adjacent primary fault segments. Increasing displacement promotes a progressive increase in the internal extent of cataclastic deformation, structural complexity, and size of Type 2 linkage structures. Cumulative frequency curves of linkage structure dimensions indicate a progressive increase in the mean length, mean width, and mean ratio of length to width with increasing displacement. The along-strike distribution of deformation alternates from a section consisting of a single fault segment to a section consisting of a cluster of either multiple fault segments or a pod of cataclasis. A lacunarity analysis quantitatively demonstrates a progressive increase in the amount of deformation within clusters with increasing fault displacement. With increasing displacement subsidiary fault segments are preferentially added in close proximity to or within earlier-formed linkage structures and are not added adjacent to single fault segments. Accreted segments typically are arcuate and abut earlier segments at a high angle. Early linkage structures represent geometric irregularities (roughness) along the evolving fault that are interpreted to result in geometric stress concentrations that preferentially localize formation of new fault segments. This conceptual model explains evolution of a systematic variation of along-fault structure with increasing fault displacement without requiring the strain-hardening hypothesis commonly invoked by other workers.

---

Vertical cable seismic methods are becoming more relevant as we require high quality and high resolution seismic data in both the land and marine environment. Our goal in this thesis is to demonstrate the impacts of vertical cable surveying in these areas.

Vertical cable methods have been applied to the marine environment with encouraging results. Data quality is similar to that of traditional towed-streamer data, without the long, cumbersome towed-streamers which are difficult to maneuver in congested areas. The current marine vertical cable processing schemes tend to use primaries and receiver ghosts of primaries for imaging. Therefore, we demonstrate the ability of the current multiple attenuation algorithms developed by Ikelle (2001) to preserve either primaries or the receiver ghosts of primaries.

As we focus on land acquisition, we discover that vertical cable surveying can overcome many of the traditional problems of land seismics. In fact, our investigations lead us to believe that problems such as ground roll, guided waves and statics can be avoided almost entirely using vertical cable acquisition methods. Furthermore, land vertical surveying is naturally suited for multi-component acquisition and time-lapse surveying.

To fully analyze the applicability of vertical cable surveys in marine and land environments, we also investigate the problem of cable spacing and sampling within each cable. We compare the resolution of vertical cable data and horizontal data by calculating the maximum angular coverage of each acquisition geometry and measuring the occurrence of each angle within this coverage, such that the more occurrences relates to better relative resolution. From our investigations, we find that by using vertical cables of no more than 500 m in length at 500 m intervals, we can acquire higher resolution seismic data relative to horizontal surface methods for an image point, horizontal reflector or a dipping reflector.

The key tool used in these investigations is fully elastic finite-difference modeling. We choose this technique based on its ability to properly and accurately model the full wavefield through complex models, all the while preserving amplitudes and phase of reflected, diffracted and converted wavefields.

---