An Integrated Study to Characterize the Permeability Field in a Fault-Partitioned Siliciclastic Aquifer/Aquitard System as an Analog to Petroleum Reservoirs

Brann Johnson¹, Elena Zhurina^2
1Center for Tectonophysics and Department of Geology & Geophysics, Texas A&M University, MS 3115, College Station, TX 77843-3115, ([email protected])
¹International Exploration, Marathon Oil Company, 5555 San Felipe, Houston, TX 77056, ([email protected])

        A fault-partitioned, siliciclastic, groundwater aquifer/aquitard system in central Texas is an excellent analog for fault compartmentalized, siliciclastic, petroleum reservoirs. We have developed a unique field laboratory for study of the fluid-flow attributes and hydraulic signatures of stratigraphic heterogeneities and faults. The study integrates core-scale geological and permeability characterization, borehole geophysical logs (natural gamma, 2.5 cm electrode spacing electrical resistivity, velocity, and acoustic borehole televiewer), and reservoir-scale, hydraulic head data and numerical modeling analysis. 
        The aquifer is a sandstone-dominated unit with a few discontinuous mudstone interbeds (braided-stream fluvial deposit) that grades upward into interbedded sandstone and laterally extensive, mudstone strata (tide-dominated eusturine environment) representing the confining aquitard. At the test site an oblique-slip, normal fault with 20 m of displacement partially offsets the aquifer/aquitard system and impedes cross-fault flow. Eleven closely spaced (3-15 m), continuously cored boreholes have been drilled through the fault and adjacent aquifer/aquitard strata to basement to depths of 125 m. A high-resolution, stratigraphic and strucutural model has been constructed for the site based upon 1050 m of core and geophysical logs.
        The study fault is a linked fault system consisting of three major segments, which in turn are comprised of multiple linked subsegments. Two major segments overlap and locally hard link along both strike and dip and exhibit systematic spatial variations of displacement consistent with displacement transfer. In core, faults are manifest as well-defined shear zones with distinctive internal structure. Shear zone thickness is only weakly correlated to displacement. Permeability of fault rock in the shear zones is dominated by the effects of cataclasis of the sandstone and incorporation of mudstone; secondary mineralization occurs, but its intensity appears to be spatially variable. 
        Hydraulic head data are obtained from multilevel monitoring systems installed in eight boreholes that have 94 hydraulically isolated, pressure measurement zones distributed in a region 40 m by 20 m by 70 m encompassing the fault and adjacent aquifer/aquitard strata. In addition, 25 zones have “injection/withdrawal” ports, which can be used to perform localized, slug interference tests. A good geological model is critical to proper installation of the multilevel monitoring systems. Packers are positioned at all significant fault piercements and selected, laterally extensive, clay-rich strata. Packers in adjacent wells are positioned to bracket selected hydrostratigraphic intervals. Three types of hydraulic data are measured: (1) the quasi-steady state distribution of hydraulic head (natural flow conditions), (2) transient pressure histories at 85 measurement zones during full-reservoir pump tests, and (3) transient pressure histories in multiple zones during localized, cross-well, slug interference tests.
        The 3-D quasi-steady state distribution of hydraulic head clearly shows the effect on fluid flow of faults, the gross stratigraphy and local, low permeability stratigraphic heterogeneities. These effects are seen as anomalous head differences both within a single multilevel well and between adjacent wells. Head differences of 0.1 to 0.9 m occur across faults and exhibit systematic spatial variations that correlate either with stratigraphic position or fault displacement. In general the smaller head differences occur where the sandstone-dominate aquifer strata are juxtaposed across the fault. The largest head difference (0.9 m) occurs where the aquifer strata are faulted against the aquitard strata. The spatial distribution of fault-rock permeability is inferred from these data utilizing a geophysical inverse model coupled with a 3-D finite element model incorporating the geological model. The analysis shows a systematic variation of fault-rock permeability with a five order of magnitude change over a 50 m dip section and up to two orders of magnitude variation along a 15 m strike section. Model results show the flow is focused through the lower portion of the fault with only very minor flow across up-dip regions where the fault cuts the aquitard. The model calculates a fault-rock permeability of 10-30 md where the fault juxtaposes the aquifer (500 md) against itself. Where the aquifer is faulted against the aquitard, the fault-rock permeability ranges from 0.01 to 3.0 md. Within the aquitard, fault-rock permeability decreases from 0.01 to 0.0003 md up-section. Note that the 10 fold variation of head difference across the fault differs markedly from the calculated 100,000 fold variation of fault-rock permeability. Clearly, interpretation of the hydraulic significance of head differences across a fault requires care. At the field site, the distribution of the head difference across the fault reflects the 3-D flow field associated with the fault and not local flow across the fault.
        Full-reservoir pump tests also have been performed at the test site. Ideally, these data will permit a refinement of the hydraulic properties of faults deduced from the natural flow data. The tests utilize two, uncased, fully penetrating irrigation wells that straddle the fault and are in close proximity (7 to 65 m) to the multilevel monitoring wells. Drawdown and recovery transient pressure histories were measured for 85 zones and two open hole wells using pressure transducers with a resolution of 55 Pa (0.008 psi). Numerical modeling analysis of these data remains to be finished, but some qualitative inferences can be made at present. The transient pressure histories vary significantly from zone to zone in a single multilevel well as well as between adjacent wells. Derivative plots are especially useful for differentiating details of pressure histories. Using the constraints of the geological model, the derivative curve of a zone can reflect absolute vertical position, vertical stratigraphic position, and proximity to either a fault or a significant stratigraphic heterogeneity. Response curves for zones in a single hydrostratigraphic zone are similar in form, but different zones have discernibly different response characteristics. Comparison of the response curves of zones straddling a known fault shows the fault induces a time delay for the “downstream” zone relative to the “upstream” zone as well as a discernible change in the functional form of the curve. The spatial pattern of drawdown as a function of time is especially informative relative of the effect of stratigraphy and faults on the pumping-induced fluid flow field. Focused flow through the lower portion of the fault is very evident in these data. Quantitative inferences from this transient data must await the completion of the numerical modeling, because the pronounced 3-D nature of flow and the pronounced heterogeneity precludes use of the existing simple, well analysis models. One potential drawback of the full-reservoir pump test is the need to model a large region of the aquifer/aquitard system. 
        Slug interference testing utilizing the multilevel monitoring wells is currently on-going and is proving to be very effective at revealing effects of faults and local stratigraphic heterogeneities. The volume affected by a slug test is much smaller than a full-reservoir pump test, hence the region one must model during data analysis is smaller. Qualitative analysis of slug test responses at multiple zones also appears to be easier, because inference of the nature of the flow field is easier. Both falling-head and rising-head slug tests are performed, using initial slug magnitudes of 20 to 40 m. Using pressure transducers with 20 Pa (0.003 psi) resolution, a test series for a single slugged interval typically provides usable pressure histories for 30 to 40 zones located in a 3-D pattern at distances from 3m to 27m. Spatial attributes of the permeability field surrounding a slugged interval are readily seen by comparing the response curves at multiple localities. Comparison of response curves for zones straddling a known geological feature is especially useful for a quick qualitative assessment of the hydraulic attributes of the feature. Examples of the effect both of low permeability stratigraphic features and faults will be shown. Quantitative analysis requires numerical modeling. At present we are using forward modeling to gain insight into how simple geological heterogeneities affect the transient pressure field pattern. This is a first step toward development of an analysis strategy. Travel-time hydraulic tomography also is being investigated.

Copyright © 2004. The American Association of Petroleum Geologists. All Rights Reserved.