GEOLOGICAL INTERPRET AT ION OF WELL TEST ANALYSIS A
Shared by: gty33410
Journal of Petroleum Geology, vol.26 (1), January 2003, pp 00-00 1 GEOLOGICAL INTERPRETATION OF WELL TEST ANALYSIS: A CASE STUDY FROM A FLUVIAL RESERVOIR IN THE GULF OF THAILAND S.Y. Zheng1*, P.W.M. Corbett1 and A. Emery2 One problem with the inversion of transient well test data is that it can yield a non- unique solution. The uncertainty resulting from this type of approach can only be resolved by considering information from another source such as geology. Geological information will help to define the interpretation model which will ensure the correct analysis of the well test data. The results of well test analyses are of little value to reservoir characterisation and modelling unless they can be explained from a geological point of view. This last step is what we refer to here as geological interpretation. Other sources of information which can help with well test analyses come from seismic surveys and petrophysics. Modern well test interpretation therefore consists of two major steps: analysis of the well test data; and interpretation of the results. In detail, this should include the following: (1) definition of an interpretation model — this requires the integration of geological, seismic and petrophysical data with transient pressure data; (2) analysis of the well test data based on the interpretation model defined; (3) geological interpretation of the results, which is necessary in order to explain or give meaning to the results. In this paper, we present a case study from a fluvial gas reservoir in the Gulf of Thailand which demonstrates these procedures. In the context of a defined geological environment, a transient pressure test has been fully analysed. Newly-developed software based on the finite element method has been used to forward model the test scenarios. This allowed the results of seismic and petrophysical analyses to be integrated into the well test model. This case study illustrates the integrated use of geological, petrophysical, well test and seismic attribute data in defining a reservoir model which respects both the reservoir geometry at some distance from the well location and also the reservoir’s heterogeneity. We focus on a particular well in the Pattani Basin at which conventional well test analyses have been conducted. By considering the results of these analyses, forward modelling was carried out in which the drainage area was “cut” out of the structural map defined by 1 Department of Petroleum Engineering, Heriot-Watt University, Edinburgh, EH14 4AS. 2 Western Geophysical, London. *Author for correspondence: email@example.com 2 Geological interpretation of well test analysis seismic interpretation; also, the formation’s internal heterogeneity was modelled according to well logs and petrophysical analyses. Finally, analytical and simulation results were compared with the transient pressure data. We conclude that the integration of geological, seismic, petrophysical and well test data greatly reduced uncertainties in well test interpretation. The consistency of the results and the fact that they satisfied all the relevant disciplines meant that much more confidence could be given to their interpretation. INTRODUCTION A general problem in conventional well test analysis is the non-uniqueness of the solution due to the inverse nature of the problem. This can lead to uncertainties in the characterization of a reservoir, particularly with respect to its geometry and properties such as layering. Geological data can help to reduce this kind of uncertainty by ensuring that a correct interpretation model is selected. The importance of combining geological data with well test analyses to improve reservoir characterisation was not fully recognised until the early 1990s (Zheng, 1997), but after about a decade of development, tools such as computer software for this purpose are now available to the oil industry. The role of well testing has therefore been extended from production testing to reservoir characterisation, reservoir management and reservoir modelling (Horne, 1995; Zheng, 1997). The integration of geology and well testing requires an improved understanding of these measurements particularly with respect to their scale and limit. In general, a well test measures a reservoir volume at a length scale of about 500m. From this, the effective properties of the reservoir such as its permeability can be derived, and can be compared to upscaled core-derived measurements (Zheng et al., 2000). Furthermore, the reservoir geometry (or the “geological shape”) derived by inversion of the transient pressure data tends to be symmetric. Together, these yield a reservoir model which has a symmetric geometry and uniform properties. This, of course, can be matched by an analytical solution if it is available. However, this kind of model can be very different from one derived from a geological and petrophysical approach (Zheng, 1997). In order to define the meaning of the derived parameters from such an analysis, a geological interpretation is required. Another aspect of the integration relates to the difference between the dynamic performance and static description of the reservoir. For example, a change in formation heterogeneity at a borehole scale (e.g. due to a high permeability streak of limited extent) will change the test performance (i.e. the pressure derivative curve), although the description at a reservoir scale will remain the same (Corbett et al., 1996). A consistent interpretation satisfying both the analytical solution and the geology can only be derived when the reservoir is homogeneous or nearly homogeneous. In the case of a heterogeneous system such as a fluvial reservoir, the analytical solution alone cannot clearly explain the model scenarios. Also, an analytical solution may be very hard (or impossible) to achieve, in which case a numerical approach is required. The numerical method of modelling well test data in a heterogeneous reservoir is straightforward. A numerical model is constructed based on a geological description and on the well test analysis. The dynamic reservoir response, expressed in the form of a pressure derivative curve, is derived by well test simulation. Analytical solutions (e.g. those in published literature) can easily be generated in this way. In addition, geological and petrophysical inputs together with those from other relevant disciplines can be integrated into the reservoir model, so that a range of responses can be generated to represent scenarios which are close to reality. This approach can also be used to validate an existing well test interpretation. By forward modelling the interpretation results, data from geophysics (seismics), geology and petrophysics can be integrated into the well test model to derive an interpretation with improved confidence (Zheng et al., 2000). S.Y.Zheng et al. 3 Fig. 1. Location map of the study area in the Pattani Basin, Gulf of Thailand. CASE STUDY: GEOLOGICAL AND PETROPHYSICAL DESCRIPTION We use well test data from an appraisal well, “well E”, located in the gas-producing Pattani Basin, Gulf of Thailand (Fig.1). Reservoir rocks are Tertiary fluvial channel sandstones (Lian and Bradley, 1986; Duval and Gouadain, 1994). Sandbodies in the thick Miocene succession are discontinuous, and most of the fields are affected by normal faulting. A major challenge in reservoir characterisation is to distinguish between reservoir boundaries produced by channel margins and those caused by faulting. Since these complex reservoirs are seldom penetrated by more than one well, it is important to make good use of techniques which can enable us to visualise the reservoir morphology at some distance from the well location. Both well testing and seismic data can provide the necessary information (Brown, 1991; Beydoun and Guerillot, 1993) and when properly integrated with geological data from core and wireline logs, should give enhanced confidence to reservoir models. The cored interval at well E, which corresponds almost exactly with the tested interval, represents a 44ft thick sandstone body (Fig. 2). The basal 2.5ft of the sandbody consists of 4 Geological interpretation of well test analysis Fig. 2. Well E, Paltani Basin: composite logs (left) show the GR response and density profile through the 40ft thick perforated interval. Core description (centre) and permeability distributions (right: plug and probe) for the sandbody tested were together used to define four flow units. a carbonate-cemented, conglomeratic lag deposit with mudstone clasts, which is overlain by a 7.5ft thick, medium-grained fining-up sandstone with some cross-bedding. Another, thinner, coarse lag is followed by a cross-bedded sequence which fines up only slightly to medium-fine sand grade, some 20ft above the base of the sandbody. Above, another thin lag is succeeded by a coarser (mainly coarse sand) unit which is cross-bedded; this coarsens up, with pebbles in places and sometimes reaching very coarse sand, to the top of the sandbody. At the top of the sequence, some ripple sets were recorded as thin calcite- cemented intervals. Based on these observations, the lower part of the section (to 20ft) is considered to represent meandering fluvial channel deposits. Supporting evidence includes the coarse lag representing the channel base, the fining-upward nature of the overlying units, and the typical internal structures. Above 20ft, the evidence points to a less sinuous, more braided, type of fluvial system. Permeability was measured on conventional core plugs, spaced at 2ft intervals, and also using a probe permeameter with readings taken every 0.1ft (Fig. 2). In general, the probe was of more use in providing a detailed picture of permeability variations in the S.Y.Zheng et al. 5 Fig. 3. Geological interpretation of the tested fluvial reservoir sands at well E. Braided channel sands overlie meandering channel sands. core. The probe data-profiles generally follow grain-size trends, i.e. the coarser sands give higher permeabilities and vice-versa although the coarse lag at the base of the sand body is very tight because it is carbonate-cemented. However, the probe data picked out subtle permeability variations and allowed us to define four flow units, one more than we would have done had we used the core description alone. Flow unit 2, the second one up from the base, was picked based on permeabilities rather than sedimentology. The extent to which the different flow units (especially the lower three “meandering” units) represent channel stacking or slight shifts in channel orientation etc., is not known. The geological interpretation (Fig. 3) comprises cross-bedded coarse sandstones in sets 0.5-1ft thick which coarsen up slightly, suggesting small bars developed in the coalescing channels of a braided river. This is consistent with the regional structural interpretation: braided streams generally develop on higher gradient surfaces, suggesting a component of tectonic tilting. Overall then, the sandbody seems to represent a composite channel form, where initial incision and deposition of a coarse lag was followed by a meandering channel, which was in turn succeeded by a braided channel system, before abandonment and deposition of overbank silts. The sequence may represent an incised valley fill. 3-D SEISMIC INTERPRETATION Data Set A 3D seismic dataset covering an area of 65.64 sq.m (a subset of a much larger dataset) was made available to the project. However, this study deals only with the area around well E together with 7.72 sq.m. of seismic coverage. This particular well was selected 6 Geological interpretation of well test analysis Fig. 4. West-east inline seismic section through the well location. The bold black line is the well path; the green and yellow lines are the “top” and “base” sand picks, respectively. Dotted lines are faults. Note the high amplitudes of the reservoir horizon adjacent to the major fault (centre right). because the operator had indicated that the channel could easily be recognised on seismic profiles; and it was the only well in which cored and DST (Drill Stem Test) intervals coincided. The 3-D survey was acquired in 1993. The acquisition configuration consisted of a single vessel, single source and dual streamers. Inline acquisition direction was east-west, and inline bin spacing was 41ft. Cross-line, an unusually large subsurface line spacing of 246 ft was used, with an interpolation step included in the processing to reduce this to 123 ft. Data quality is in general good with two caveats: (a) wave-front distortion, velocity shadow effects, etc. due to faulting – this caused reflection continuity to break down in the vicinity of the faults; and (b) the large cross-line spacing which, especially in combination with the above distortion effects, served to reduce resolution and interpretability. Point (b) is expanded below. Reflector continuity was reasonable, the dominant frequency at the reservoir level being approximately 30 Hz. Processing followed a fairly standard flow. The data were “zero-phased” in the designation step so interpreted picks should be on the peak or trough for “thick-bedded” geology, i.e. where the top and base of the bed can be resolved. Analysis and Interpretation The gas sand tested in well E is 44ft thick. Theoretically, one cannot discriminate between the top and base of beds which are thinner than a resolvable limit, which in this S.Y.Zheng et al. 7 Fig. 5. Time-structure map for the “top” sand pick. Highs are shown in blue, lows in red. The well path is marked with surface location (solid symbol), reservoir intersection and TD (crosses). Faults are shown; fault to the NW is projected (dashed line) to intersect with the fault to the South forming a “closed” reservoir “block”. Fig. 6. Amplitude map for the “top sand” pick. Reds and yellows are high amplitudes; greens are low amplitudes, and blue is phase changes. The well path is marked, including surface location (solid) and reservoir intersection (circle). 8 Geological interpretation of well test analysis Fig. 7. Interpretation of the seismic amplitude map in terms of sandbody distribution and faulting. The heavy stipple represents possible fluvial channels (high amplitudes); the light stipple represents possible crevasse splays (moderate amplitudes); floodplain muds and silts (low amplitudes) are also shown (white areas). There is uncertainty about the fault closure near the well. case is approximately 160 ft (although this is rather pessimistic: Hardage et al., 1994). The tuning thickness is the critical thickness at which reflections from the top and base of a bed constructively interfere to boost the amplitude. Some interpreted horizons were supplied in digital form by the operator and these initially tied the well, although they required re-picking to “snap” them to the correct part of the phase. The “top” and “base” events were picked around the well. Faulting was interpreted afresh. Interpretation followed the conventional approach using in-lines, cross- lines, user-defined traverses and time slices to tie-down fault positions and fault-horizon relationships. Although we agree with Brown (1991) about the value of time slices in interpretation, their utility in this case was reduced because the cross-line spacing was three times the in-line spacing, causing the display to be “squashed” cross-line; and also because the time slices were blurred which was again thought to be due to the large cross- line spacing, which on migration would produce smearing in 3D. Reduction in event amplitudes adjacent to fault planes was one factor which did help in picking faults. Fig. 4 shows an in-line seismic section (oriented east-west) through the well location; Fig. 5 is a time-structure map (“top sand”). The well encountered the reservoir sand on a fault terrace between two major normal faults (displacement 500ft), which trend north- south and are down-thrown to the east. The reservoir horizon dips 10° west, and the depth to the top reservoir at the well location is 9,802ft (measured depth, MD) or 7,091ft (true vertical depth sub-sea, TVDSS). Fig. 5 also shows a series of minor faults with various orientations, including some which are normal to the trend of the major faults and serve to segment the fault terrace. S.Y.Zheng et al. 9 Fig. 6 is a map of the amplitude of the event picked as “top sand”. There is an elongate zone of high amplitudes around the well location, but this does not exactly correspond to the morphology of the structural high shown in Fig. 5. Indeed it is more “banana” shaped, and following the high amplitudes away from the fault terrace itself, both to east and west, one gets an impression of a sinuous feature cutting across the structural grain, approximately from the top right (NE of the area) to the bottom left (SW). If one mentally “strips out” the effect of amplitude dimming adjacent to fault planes, this effect is enhanced. There are other high-amplitude zones, especially to the east. This suggests, by correlation with the core interpretation, that the elongate, high-amplitude zones represent sand-prone fluvial channel belts; that low-amplitude “background” areas are potentially alluvial plain muds; and that intermediate amplitudes possibly represent thin sandy/silty crevasse splay deposits. Several authors have described the seismic mapping of fluvial channels from other parts of the Gulf of Thailand (Lian and Bradley, 1986; Duval and Gouadain, 1994; Trevena et al., 1995). The map in Fig. 7, is an interpretation of the seismic amplitude map and shows that the reservoir appears to be compartmentalised by a combination of structural features (faulting) and stratigraphic (channel belt) elements. Uncertainties in the seismic interpretation of this area include the degree to which the margins of high-amplitude areas represent the edges of channel belts. Also, several of the faults mapped are subtle and low-throw, and some interpreters may not have mapped them; furthermore, the potential effects of sub- seismic faults has not been addressed. The input from well testing is therefore important in helping to tie down the reservoir structure. The seismic interpretation shows a major fault 110m east of the well location (Figs.5 and 7), the western side of the reservoir being bounded by a low-throw fault and / or the margin of a channel belt; to north and south, it is probably bounded by minor faults. The reservoir area defined in this way covers approximately 160 acres. Below, we bring together the seismic and DST interpretations, together with the core description and permeability data. WELL TEST ANALYSIS DST-1 (well E) Description In well E, a drill-stem-test (“DST-1”) was conducted over the interval between 9,802.0 and 9,846.0 ft (MD). A 3.3/8" tubing-conveyed perforation (TCP) was performed under- balanced; the perforation density was six shots per foot (SPF). The objectives of the test were to derive information on zone deliverability and reservoir performance, to sample fluids (gas composition, condensate gravity), and to determine the reservoir limit from reservoir depletion. DST-1 comprised 31 minutes of initial flow (draw-down) followed by an initial shut-in period (first “build-up”) lasting about four times the preceding flow time. Main flow for 24 hours then took place while maintaining critical flow using a 64/64" choke. The well was then shut-in again before a final pressure build-up which lasted 36 hours or about one-and-a-half times the preceding flow time. Fig. 8 shows the test overview. Well Test Analysis In Fig. 8, the reservoir depletion was determined from a comparison of the final pressures at the end of the first and second build-ups. Test analysis was completed following a conventional procedure, i.e. (a) definition of the time regions and flow regimes for model 10 Geological interpretation of well test analysis Fig. 8. DST-1 well test overview, comprising an initial drawdown, an initial pressure build-up, a main draw-down and a final pressure build-up. Reservoir depletion can be determined by comparing the end pressures from the initial and final build-ups. Fig. 9. Well test history match which considers only the transient pressure data. The derived reservoir model is a simple “tank” with uniform properties in which a well is centrally located. The pressure history match is very good, but no other data (geological, seismic, well logs, petrophysical) have been integrated. S.Y.Zheng et al. 11 selection; (b) analyses of the identified flow regimes; and (c) flow regime and test history matching using the defined reservoir model and the derived corresponding parameters. Since the theme of this paper emphasises the integration of different types of data, the details of the well test analysis are not discussed in detail although the results are shown in Fig. 9. This figure shows that the test history match is very good. The reservoir model derived is an elongate “box” which is 4,800ft long, 680ft wide, and 34 ft thick. The reservoir permeability (k) is 550mD with a skin factor (S) of 8.1. This homogeneous reservoir model includes no information about formation layering or property contrasts between layers; it consists of a uniform box or “tank” with two parallel no-flow boundaries and a well located centrally between them. Note, however, that a good history match does not represent a unique solution for the test. The reservoir model is based on three flow regimes which were identified from the test analysis: radial flow in the near well-bore region; linear flow due to the parallel no- flow boundaries; and pseudo- steady-state flow resulting from reservoir depletion. This yields a reservoir whose 2D geometry is a 4,800ft x 680ft rectangle with a drainage area of 75 acres. The distance between the two parallel boundaries is 680ft. The volume of gas initially in place, based on the cumulative production and the reservoir depletion, is calculated as 1.69 billion standard cubic feet (Zheng, 1997). WELL TEST FORWARD MODELLING AND SIMULATION In order to resolve the uncertainty resulting from the conventional well test analysis, well test forward modelling was conducted. Since the test results cannot explain the reservoir’s heterogeneity as indicated by the seismic, geological, well log and petrophysical analyses, well test modelling aimed to integrate these data in order to construct a more realistic model which honours all the data. Then, by comparing the simulated well test pressure response against the DST data, and matching the flow regimes identified in the primary test analysis, a more confident test interpretation can be achieved. Forward Modelling In the first step, a reservoir model was constructed by “cutting” the drainage area out of the seismic map. Reservoir boundaries were defined along the faults defined by the seismic interpretation (Fig. 7), where the fault to the NW of the reservoir has been extended southwards. This is because the well test showed clear reservoir depletion within the testing time scale, and the tested reservoir must therefore be a “closed” system. As shown in the top-structure map in Fig. 10, the reservoir comprises a triangular northern section and a rectangular southern section. The colours suggest that these two sections are different regions within the modelled reservoir; however, both regions were assigned the same properties, so the reservoir is areally isotropic. The reservoir’s vertical heterogeneity was modelled according to the well-log and petrophysical measurements in Fig. 2. Instead of giving the reservoir uniform properties, a two-layered reservoir model was constructed; both layers had identical porosities but different permeabilities. The overlying 30ft thick braided-channel sandbody was given a permeability of 400mD, while the underlying 4ft thick meandering-channel sandbody had a permeability of 1,000mD. A production well was located according to the seismic amplitude map in Fig.6, and the reservoir model is shown in Fig. 11. In order to capture the pressure response in the near-well region with a high-pressure gradient, as well as the pressure changes due to the reservoir boundaries, the finite element method was used to construct the reservoir model and the reservoir grid. Fig. 12 shows the model in 2-D; the reservoir drainage shape is shown on the left, the finite element grid 12 Geological interpretation of well test analysis Fig. 10. Combining the results of seismic and well test analyses, a reservoir model has been defined with fault-defined boundaries in which the uncertainty about fault closure has been constructed from well test analysis. The reservoir area is composed of yellow and pink regions (see text for details), but these areas can be unified because the reservoir is considered to have areally isotropic properties. Fig. 11. Revised 3-D reservoir model, which integrates seismic, well test, well log and petrophysical data. This model has taken the realistic reservoir’s drainage area and the formation vertical heterogeneity (layering) into account. S.Y.Zheng et al. 13 Fig. 12. Constructed 2D reservoir model and reservoir grid based on seismic data and the results of well test analysis. (a) reservoir model; (b) finite element grid around the well; and (c) combined reservoir model and grid. Fig. 13. Transient pressure distributions generated from well-test flow simulations. Radial flow (a) followed by linear flow (b, c) due to the parallel faults which define the reservoir’s boundaries. There is then a transition from linear flow to full reservoir depletion (d, e): pseudo- steady-state flow states are clearly captured. Fig. 14. Flow regime match corresponding to the pressure distributions in Fig.13. The dotted points are DST pressures and associated derivatives with well-bore storage, while solid lines are analytical solutions with zero well-bore storage. Vertical axis, i.e. Delta m(p), is calculated pseudo-pressure change. Radial flow, linear flow due to parallel faults and “closed” reservoir depletion are all matched on the pressure derivatives. 14 Geological interpretation of well test analysis Fig. 15. The final match of the full DST-1 testing history corresponding the flow regime match shown in Fig.14. around the well in the centre, and the two together are shown on the right. Well Test Numerical Simulation With the reservoir model described above, a well test was simulated using the same time and rate schedule as in DST-1. This comprised an initial draw-down and an initial pressure build-up, followed by a main-flow period and a final pressure build-up. The total simulation time was about 66 hours with 1,343 time steps. Gas properties and composition were generated using correlations, according to the gas gravity and Z factor derived from the PVT lab. Skin factors of 7 and 7.6 were assigned to the two reservoir layers, with associated damaged zones having radii of 1.5ft and 1.8ft, respectively. The simulation was first tuned to match the flow regimes identified from DST-1; then the skin and well-bore storage were considered to match the full test history. WELL TEST INTERPRETATION Fig. 13 shows the simulated pressure distributions from the reservoir model. Red colours represent high-pressure values, while yellow, green and blue represent decreasing pressures. After well-bore storage, radial flow around the well is initiated (Fig. 13a). Then linear flow starts (Fig. 13b), followed by pseudo- steady-state flow due to reservoir depletion. As shown in Fig.14, the flow regimes correspond to the pressure distributions in Fig. 13. The dotted red and blue points are DST pressures and derivatives associated with well- bore storage, while solid blue and red lines are analytical solutions with zero well-bore storage. Radial flow, linear flow due to parallel faults and “closed” reservoir depletion are all matched on the pressure derivatives. These together are a result of a two-layered reservoir model with variable properties and irregular geometry. This interpretation is further S.Y.Zheng et al. 15 confirmed by the full test history match using the defined model and the corresponding derived parameters (Fig. 15). These results are closer to a realistic interpretation with a better representation of the reservoir’s geology, heterogeneity and the dynamic performance of the reservoir. They are more acceptable than those from the preliminary interpretation. CONCLUSIONS 1. Using data from an appraisal well in the Gulf of Thailand, we show how an integrated approach to well test interpretation can greatly reduce inherent uncertainties. In a defined geological background, numerical forward modelling, which integrates seismic, petrophysical and transient pressure data yields more realistic results and improved reservoir characterisation. 2. Test analysis and interpretation are two different steps. Analysis yields a non-unique solution, and integration of data from relevant disciplines is required to further explain or interpret the physical meaning of the results. Well test forward modelling and numerical simulation are essential for this integration. 3. From our case study, we show how geological interpretation of a well test can enhance the role of geology in reservoir modelling and also bridges the gap between dynamic reservoir performance and a static description. A transient pressure test is only useful when such an approach has been made. ACKNOWLEDGEMENTS Shell, Norsk Hydro, ExxonMobil, Unocal, Phillips, Wintershall and BP are acknowledged for funding this work within the GEOTIPE Project, and Unocal Thailand in particular is acknowledged for providing the data and permission to publish this paper. Edinburgh Petroleum Services Ltd (EPS) are thanked for the provision of software to undertake this study. Journal review was by P. Wong (Veritas) and R. Horne (Stanford University) whose comments on a previous version are acknowledged with thanks. REFERENCES BEYDOUN, W. B. and GUERILLOT, D. R., 1993. Identification of Reservoir Limits Integrating Seismic and Fluid Flow Responses. SEG’93, 341-344. BROWN, A.R., 1991. Interpretation of Three-Dimensional Seismic Data. 3rd edition. AAPG Memoir 42, 341 pp. CORBETT, P. W. M., MESMARI, A. and STEWART, G., 1996. A method for using the naturally-occurring negative geoskin in the description of fluvial reservoirs. Paper SPE 36882, SPE European Petroleum Conference, Milan, Italy. DUVAL, B.C. and GOUADAIN, J., 1994. Exploration and Development in the Bongkot Field Area (Gulf of Thailand): An Integrated Approach. Proc. 14th World Petroleum Congress. J. Wiley, pp.247- 257. HARDAGE, B.A., LEVEY, R.A., PENDLETON, V., SIMMONS, J. and EDSON, R., 1994. A 3-D Seismic Case History Evaluating Fluvially-Deposited Thin Bed Reservoirs in a Gas-Producing Property. Geophysics 59, 1650-1665. HORNE, R., 1995. Modern Well Test Analysis. Petroway, Inc., 257 pp. LIAN H.M. and BRADLEY, K., 1986. Exploration and Development of Natural Gas, Pattani Basin, Gulf of Thailand. Transactions of the Fourth Circum Pacific Energy and Mineral Resources Conference, Singapore, pp. 171-181. TAMHANE, D., WANG, L. and WONG, P.M., 1999. The role of geology in stochastic reservoir modelling: the future trends. SPE 54307. 16 Geological interpretation of well test analysis TREVENA, A.S., HOBBS, W.H., TURK, M.J. and WILLIAMSON, C.R., 1995. Core and 3D Seismic Study of Miocene Fluvial Deposits, Pattani Basin, Gulf of Thailand. Paper 5132, ACOPE Conference and Exhibition, Bangkok, 2-6 November. ZHENG, S.Y., CORBETT, P.W.M., RYSETH, A. and STEWART, G., 2000. Uncertainty in well test and core permeability analysis: a case study in fluvial channel reservoirs, northern North Sea, Norway. AAPG Bull. 84, (12), 1929–1954. ZHENG, S.Y., 1997. Well Testing and Characterisation of Meandering Fluvial Channel Reservoirs. Unpublished Ph.D thesis, Heriot-Watt University, 226 pp.