WRIR 01-4210: Hydraulic-Property Estimates for Use With a Transient Ground-Water Flow Model of the Death Valley Regional Ground-Water Flow System, Nevada and California
Previous Section: Hydrogeologic Setting |
Next Section: Hydraulic Properties
Table of Contents | Conversion
Factors and Acronyms | Report Home Page
In this study, all aquifer-test results compiled from published reports were verified by re-analyzing the aquifer-test data using analytical solutions appropriate to the hydrogeologic setting in which those tests were conducted. If the published results agreed to within a factor of 2, the published results were accepted. If the difference between the published data and the independent calculations exceeded a factor of 2, and no independent justification was found for using the published data, the calculated values were reported. Because of the uncertainty associated with converting specific capacity data to transmissivity values, specific capacity data were not used. Because of the low volume of geologic material samples, results from the permeameter tests were not used in the analyses discussed in this report (with the exception of the clastic confining units). Following the elimination of suspect data and the addition of newly analyzed data, statistical methods were used to evaluate the distribution of hydraulic properties in the 11 DVRFS derived HGU's. Except for wells located on the Colorado Plateau in Utah, figure 2 shows the locations of the wells and boreholes used to collect data for the estimation of hydraulic properties presented in this report. The Colorado Plateau wells are not contained in the USGS National Water Information System (NWIS) database and do not have exact locations associated with them. These wells were included in the analysis of hydraulic properties because they are completed in the sedimentary confining unit (table 1), of which data are sparse in the DVRFS. The following hydraulic parameters are the primary focus of this study because of their use in ongoing numerical flow-modeling studies. The parameters were defined by Lohman (1979, p. 6 and 8):
Hydraulic conductivity (unit length per unit time): The coefficient that describes the ability of a geologic medium to ". . . transmit in unit time a unit volume of ground water at the prevailing viscosity through a cross section of unit area, measured at right angles to the direction of flow, under a hydraulic gradient of unit change in head through unit length of flow." Hydraulic conductivity can be calculated by dividing the transmissivity by the aquifer thickness (Lohman, 1979).
Transmissivity (square unit length per unit time): ". . . The rate at which water of the prevailing kinematic viscosity is transmitted [horizontally] through a unit width of the aquifer under a unit hydraulic gradient."
Specific yield (unitless): "The ratio of (1) the volume of water which after being saturated, it [rock or soil] will yield by gravity to (2) its [rock or soil] own volume." Specific yield is virtually the same as the storativity for unconfined aquifers.
Storage Coefficient or Storativity (unitless): "The volume of water an aquifer releases from or takes into storage per unit surface area of the aquifer per unit change in head."
Aquifer tests in unconsolidated sediments throughout the Death Valley region were analyzed by conventional methods developed for porous media (Dawson and Istok, 1991; Driscoll, 1986; Lohman, 1979). Because the consolidated sedimentary and igneous rocks of the region tend to be heavily fractured and the aquifer volume generally is large enough to permit an equivalent porous-media response to pumping, porous-media analysis methods were deemed adequate. This assumption is examined in more detail in the section "Fractured Media and Equivalent Porous Media." Once a match has been determined, a point is selected and the corresponding coordinate values for head, time, dimensionless head, and dimensionless time are selected.
Several different methods were used to analyze the data which were acquired from tests of constant-rate pumping, slug (injection and bailing), swabbing, and drill stem. Common analytical methods are briefly described below, while details can be found in the cited references. Uncommon analytical methods used in this study are cited with the aquifer-test results (app. A).
Constant-rate pumping and injection tests were analyzed by curve-fitting methods. Theoretical solutions to aquifer-test problems are represented as dimensionless curves. Data in the form of water levels or recovery are plotted as a function of elapsed time on log-log scales. These data curves are then matched to the dimensionless curves. These match-point values are then substituted into analytical equations to estimate hydraulic-property values. The Theis (1935) solution was used for aquifer tests in non-leaky confined aquifers. Residual drawdown in pumping tests and residual water-level rise in injection tests were analyzed to determine transmissivity, storativity, and, if the representative thickness of the aquifer is known, hydraulic conductivity. For this method, water-level change was plotted as a function of the log of the ratio of elapsed time since pumping or injection started to the elapsed time since pumping or injection ceased (Theis, 1935). The Theis method, as do those methods discussed below for confined aquifers, assumes that observation wells completely penetrate homogeneous, isotropic, confined aquifer of infinite extent. Curve-fitting techniques for estimating transmissivity and storativity for leaky, confined aquifers without storage in the confining unit were developed by Hantush and Jacob (1955) and Cooper (1963). Curve-fitting methods for estimating the transmissivity and storativity for leaky, confined aquifers with storage in the confining unit were developed by Hantush (1961) and Bourdet (1985). For unconfined aquifers with anisotropy but using the other assumptions previously mentioned for confined aquifer methods, Boulton (1963), Stallman (1965), and Neuman (1975) developed curve-fitting techniques to estimate transmissivity, anisotropy, and storativity. It should be noted that the Neuman (1975) method may not be appropriate for use with fractured rock. Fractured rock has a "dual-porosity" response that comes from the immediate de-watering of fractures (being the most permeable), followed by the delayed response of de-watering from the matrix. The Neuman (1975) method assumes that this delayed response is due to aquifer depressurization and dewatering. In fractured rock, the delayed response is believed to be from the exchange of water between fractures and matrix rock. Neuman analyses reported in the database are primarily from non-fractured media (e.g., alluvium). Where the Neuman (1975) method was applied to fractured volcanic rocks, the database (app. A) contains the previously published values. Because of the above-mentioned conditions, vertical anisotropy estimates for fractured rock using the Neuman method are suspect.
In fractured hydrogeologic media, fluid can be contributed to the system either from fractures or the matrix. This "dual-porosity" concept involves the exchange of water between the fractures and the matrix. Several specialized methods involving this concept have been developed, some of which were used in the published hydraulic-property estimates compiled for this report. The two methods whose results are reported in the database are by Moench (1984) and Streltsova-Adams (1978). Both methods use derived type curves for dual-porosity solutions to aquifer-test problems to match time-drawdown data from pumping and observation wells.
Straight-line fitting methods involve fitting a straight line through drawdown or residual drawdown data as a function of the log time or distance from the test well, and then substituting the slope of this line into analytical equations to estimate hydraulic-property values. Under the same assumptions applicable for the Theis (1935) solution, the slope of a straight-line fit to drawdown or recovery data plotted as a function of log time the values of transmissivity and storativity can be determined (Cooper and Jacob, 1946).
In bailing tests, water is bailed repeatedly for an extended period, but some recovery of water level in well occurs as the bailer is brought to the surface, emptied, and then returned to the test interval. The average withdrawal rate, which is the total volume of water removed divided by the time that the well was bailed, does not account for drainage back to the well between bailing runs or variations in the rate of bailing. In most bailing tests, residual drawdown from bailing can be analyzed using the recovery method of Theis (1935).
In swabbing tests, a mechanical device is lowered into the well to displace water. After repeated runs, the average withdrawal rate is calculated in the same way that the average bailing rate is calculated. Residual drawdown is then analyzed using the recovery method of Theis (1935).
In slug tests, a known volume of water either is instantaneously removed from or is injected into a well, and the time history of water-level recovery to the static water level is monitored. Cooper and others (1967) developed a method for analyzing slug tests, which was later modified by Bredehoeft and Papadopulos (1980). In the solution of Cooper and others (1967), ratios of the water-level drawdown or rise to the static water level (H/H0) are plotted as a function of log time since the test was initiated. Similar to the other curve-fitting techniques previously described, the data curve is then matched to a dimensionless type curve to obtain values of hydraulic properties.
Drill-stem tests are the standard way in which hydraulic properties of potential oil and gas reservoirs are evaluated by the petroleum industry (Bredehoeft, 1965). This test measures the pressure drop as the formation fluid (such as oil) moves from an isolated section of the borehole into a drill stem lowered into the borehole. In the method of Horner (1951), fluid-pressure recovery during the second shut-in period is plotted as a function of the ratio of the time elapsed during the shut-in period and preceding flow period to the time elapsed during the shut-in period.
Descriptive statistics, including the geometric and arithmetic means, range, and the 95-percent confidence interval (±1.96 standard deviations from the geometric mean) of the hydraulic conductivity, storage parameters, and anisotropy ratios are reported for each of the HGU's. These parameters will be used to aid in the calibration of the DVRFS transient ground-water flow model. Because hydraulic conductivity tends to be log normally distributed (Neuman, 1982), the geometric mean of the estimates is reported. The arithmetic mean also is reported. Storage parameters tend to be normally distributed (Neuman, 1982) and because of this, the arithmetic mean of the estimates is reported. Values of hydraulic conductivity derived from pumping well data, when an observation well was available, were not used in the statistical calculations to avoid bias from re-sampling the same aquifer test. For similar reasons, slug tests from intervals that overlapped each other, although present in the database (app. A), were not used in the statistical calculations.
Most of the analytical methods used in this work assume that an aquifer is a porous medium. However, the influence of fractures is fundamental to the flow of water in volcanic and carbonate rocks. In order to apply these aquifer-test methods to fractured rocks it is necessary to assume that the rocks are sufficiently homogeneously fractured and interconnected such that the rock being tested can be considered "an equivalent porous medium." The spacing of fractures, as well as their interconnectivity, can affect the results of an aquifer test. In areas where fractures are tightly spaced and interconnected, transmissivities generally are higher than in areas where the fractures are widely spaced and not interconnected. In a study on transmissivity in crystalline rock, slug tests using either porous or fractured media methods, provided estimates of transmissivity within an order of magnitude of each other (Shapiro and Hsieh, 1998). In the cases examined here, the equivalent-porous-medium assumption cannot be ruled out because plots of drawdown or recovery of water levels in wells conform to type curves derived for porous media.
Hydraulic-conductivity and transmissivity estimates are functions of test scale (Dagan, 1986; Neuman, 1990). As media test volume increases, more aquifer heterogeneity is encountered and influences the test results. For example, the potential exists to involve a larger network of fractures in the aquifer response to the imposed stress. In laboratory permeameter tests of core samples for determining rock matrix properties, unfractured core is needed for successful results. Because only matrix rock properties are determined from permeameter tests, the estimates generally are not useful for regional-scale ground-water flow models of fractured-rock aquifer systems. Thus, results for permeameter tests of core samples are not utilized in the descriptive statistical calculations of the hydraulic parameters (with the exception of the clastic confining units). Similarly, slug tests only examine a relatively small amount of aquifer material adjacent to the borehole. Because of this, hydraulic-property estimates from slug tests might not be representative of an entire unit. Single-well aquifer tests (including the pumping or injection well in multiple-well tests) optimally determine hydraulic properties in the near-borehole environment, but the accuracy of these tests can be decreased by inefficient borehole construction, convergence of flow lines and related head losses as water flows into or out of sections of perforated casing, and head loss as water moves between the test-interval depth and the pump-intake depth. As such, for the same set of wells transmissivity estimates derived from single-well tests tend to be less than those of multiple-well tests. Similarly, estimates of storage coefficients from single-hole tests are less reliable than those from multiple-well tests. Multiple-well aquifer tests tend to be more reliable because they manifest the influence of field-scale features, such as faults and fractures, as well as the water-transmitting properties of the rock matrix.
The hydraulic-property estimates presented in this report are based on the results of mostly field-scale tests involving wells. These tests include only a small amount of the volume of aquifer material within an HGU and thus are testing only a very small part of the HGU. The hydraulic-property estimates presented herein are intended to serve only as the basis for constraining flow estimates obtained from the simulation process. The scaling-up of these values for use in calibrating a regional ground-water flow model is problematic and is not explicitly addressed in this report.
General guidelines were used for selecting hydraulic-property data for compilation. These include: (1) the use of published aquifer-test results from wells in the DVRFS area. Selected unpublished data and aquifer-test results were evaluated and analyzed to fill spatial or hydrogeologic data gaps. (2) analyses of aquifer tests using methods appropriate to regional numerical ground-water flow models, and (3) analyses for each HGU should be sufficient to provide adequate spatial coverage and statistically describe variance resulting from differences in lithology, fracturing, and faulting. Based on Freund (1992), about 30 samples are a sufficient number to statistically describe parameters. Because wells and boreholes often are installed for purposes other than obtaining hydraulic-property data (such as water supply or monitoring), the above quidelines were not satisfied completely. Selected unpublished DVRFS area aquifer-test results and published data are from hydrologically similar areas.
Analytical methods used to determine the hydraulic-property estimates presented in this report rely on assumptions about the type and configuration of the aquifer. These assumptions are necessary to simplify the flow system so that mathematical equations representing ground-water flow can be solved analytically but result in some uncertainty in the computed hydraulic properties.
Most analytical methods assume that flow to a pumping well is derived from an aquifer of infinite extent. This assumption may not be accurate for many aquifer tests presented in this report because of faults in the study area that may act as either recharge or barrier boundaries.
The most commonly applied analytical methods for pumping tests in the study area, those of Theis (1935) and Cooper and Jacob (1946), assume radial flow to the pumping well under an axisymmetric hydraulic gradient. However, because of media heterogeneities, hydraulic gradients may vary directionally. Differing results in hydraulic-property values obtained from multiple-well aquifer tests involving multiple observation wells may arise as a result of non-radial flow occurring in a part of the flow system monitored by one or more, but not all observation wells. Disregarding a non-uniform hydraulic gradient seemingly would result in inaccurate computations of hydraulic properties, if the solutions of Theis (1935) or Cooper and Jacob (1946) are used. Only a single estimate of transmissivity and storage properties should be reported for these particular tests. To obtain these single results, the average of the property estimates could be used. Because the purpose of this report is to compile and report on estimates of hydraulic properties for use with a numerical flow model, all estimates are considered to be independent with respect to the descriptive statistics (central tendency and spread). Estimates from a pumping well are excluded when one or more observation wells were available due to inaccuracies inherent in the pumping well estimates. Several estimates from the same well (in the case of packer tests) or the same test (in the case of multiple observation wells) can give a range of values reflecting varying material properties of the unit. It is reasoned that because the statistics describe the central tendency and the spread of these parameters for a particular unit, use of most the estimates is appropriate (except where tested intervals straddle each other). One limitation to this approach is that the statistics may be biased toward the estimates obtained from multiple-well tests.
Single-well pumping or slug tests can provide estimates of storativity. These estimates, however, may vary up to an order of magnitude of the actual value (Cooper and others, 1967, p. 267). The slug-test solution of Cooper and others used in these analyses is very insensitive to storativity. Storativity values calculated from slug tests were not used in the statistical summaries of the hydraulic-property estimates and are not reported in the database.
Spatial bias could be significant for the hydraulic-property estimates compiled in this report. Wells and boreholes were drilled to meet the original goals of their respective studies, not to collect data to determine statistically representative regional-scale hydraulic properties. Most information was collected from wells clustered around Yucca Mountain and the NTS. Data were collected for studies of these areas and the number of wells decreases away from these areas. Many wells also were installed in relatively shallow formations because of the difficulties and cost associated with drilling deep wells.
To obtain hydraulic-conductivity estimates for use in the calibration of the DVRFS model, transmissivity estimates were divided by the thickness or length of the open interval of the tested or monitored well or borehole. The aquifer thickness was not used as this generally was unknown. Because most wells are open to the productive intervals, in a heterogeneous aquifer, coupled with usage of the open-interval thickness, hydraulic-conductivity estimates may be biased toward the larger values. Thus, the statistical means and variances presented here may be only representative of the hydraulic properties of the more productive zones within an HGU.
Other limitations of the hydraulic-property estimates involve the variability inherent in the hydrogeologic media. Lithologic factors, such as facies changes in sedimentary rock, welding in volcanic rocks, and degree of fracturing can cause hydraulic properties to vary greatly over relatively short distances. Variability also can be caused by sampling biases. For example, differences in the overlap between lithologic or sedimentologic bedding and the tested interval can cause estimates of hydraulic conductivity to vary. Sampling variability also can arise in fractured rocks as a result of a borehole failing to penetrate rock fractures especially for vertical boreholes penetrating rocks with steeply dipping (subvertical) fractures. Because of the inherent nature of variability, longer-term aquifer tests typically will produce more representative hydraulic-property estimates (hydraulic conductivity and storativity) than shorter-term aquifer tests or tests with shorter screened intervals (such as packer tests). Because of this, a smaller statistical constraint on the parameter estimates during calibration of the DVRFS model where this condition applies.
Previous Section: Hydrogeologic Setting |
Next Section: Hydraulic Properties
Table of Contents | Conversion
Factors and Acronyms | Report Home Page