Evaluation of the Hydrologic System and Selected Water-Management Alternatives in the Owens Valley, California
HYDROLOGIC SYSTEM
- Precipitation and Evapotranspiration
- Surface-Water System
- Ground-Water System
Download the contents of this page as a pdf file (1MB)
Download all Hydrologic System pages as a single pdf file (3.3MB)
Download the entire report as a pdf file (8.3MB)
Ground-Water System
The ground-water system of the Owens Valley is unusual in comparison with that of other basin-and-range valleys in eastern California. The abundant precipitation in the Sierra Nevada and resulting runoff fills the basin to nearly overflowing each year. Historically, this abundance of water has eroded the surrounding mountains, filled the graben with highly transmissive deposits, and created a shallow water table beneath much of the valley, a water table which in turn supports a great density of native vegetation not found in other similarly formed basins. In nearby basin-and-range valleys, such as Indian Wells Valley to the south (Dutcher and Moyle, 1973) and Death Valley to the southeast (Hunt and others, 1966), the quantity of runoff is much less and most of the sparse native vegetation must subsist solely on precipitation.
As a result of the abundant runoff into the Owens Valley, the surface-water and ground-water systems are strongly linked. Much of the valley floor is characterized by surface-water conveyances that are in contact with the ground-water system (figures 3 and 10), and this connection facilitates a ready exchange of water. Native vegetation on the valley floor is dependent on a combination of water obtained from precipitation, subirrigation from surface-water conveyances, and ground water. Since 1970, when export of water from the valley was expanded to include ground water, the two systems have become linked even more closely politically as well as physically. Water management of one system typically has a noticeable effect on the other.
The following sections describe the hydrogeologic framework of the ground-water system; the hydraulic characteristics of the hydrogeologic units that compose the system; the source, occurrence, and movement of water through the system; and the valleywide ground-water flow model used to simulate the system and evaluate selected water-management alternatives. The hydrogeologic history of the ground-water system and related aquifer materials is described in detail by Hollett and others (1991). Many of the major components of the ground-water system are strongly linked to a surface-water feature, such as the rive-aqueduct system. For these components, the primary description, including quantification of ground-water recharge and discharge, is presented in an earlier section entitled "Surface-Water System."
Geometry and Boundary Conditions
Nearly all the recoverable ground water in the valley is in the unconsolidated to moderately consolidated sedimentary deposits and intercalated volcanic flows and pyroclastic rocks that fill the basin. Where saturated, these sedimentary deposits and volcanic rocks make up the ground-water system. The primary part of the ground-water system, defined by Hollett and others (1991) as the "aquifer system," is capable of yielding significant quantities of ground water to wells (Lohman and others, 1972). The defined aquifer system delineated in figure 14 is also the part of the groundwater system that was simulated with the valleywide ground-water flow model documented later in this report.
The aquifer system is a three-dimensional body of valley fill that is saturated with ground water. This saturated volume of valley fill is bounded on all sides by a "boundary surface" (Franke and others, 1987). The boundary surface allows water to either flow in or out of the system, such as at the water table, or acts as a flow barrier, which allows little or no water to enter or leave the system across the boundary surface, such as at a bedrock contact.
The upper boundary surface of the aquifer system is the water table and the lower surface is either a bedrock contact, the top of moderately consolidated valley fill, or an arbitrary depth based on the depth of pumped wells. The sides of the aquifer system are either bedrock or a part of a lateral boundary surface that allows ground water to flow in or out of the aquifer system, termed a "flow boundary." Thus, water can flow in (recharge) or out (discharge) of the aquifer system only through a flow boundary.
Flow also occurs into or out of the Owens Valley aquifer system at wells, springs, rivers, or as underflow through a cross section of the aquifer system. Lateral inflow boundaries (underflow) include sections along the southeast end of Round Valley, south end of Chalfant Valley, and that part of the two valleys overlain by the Volcanic Tableland (figures 4, 5, and 14). Underflow also enters the aquifer system from the drainages of Bishop and Big Pine Creeks and from Waucoba Canyon. The lateral outflow boundary from the system is a section that crosses the valley approximately east to west at the south end of the Alabama Hills.
Hydrogeologic Units and Subunits
The hydrogeologic framework of the aquifer system controls the vertical and horizontal flow of ground water in the system. The complex framework of the actual system was simplified by Hollett and others (1991) into a vertical series of units that represent either ground-water-producing zones or major zones of confinement to vertical flow. These units are referred to as "hydrogeologic units" and are numbered 1 to 3, from top to bottom in the aquifer system. Saturated valley fill that lies below the defined aquifer system and in contact with the bedrock is referred to as hydrogeologic unit 4 and is not part of the aquifer system. The primary purpose for simplifying the heterogeneous sedimentary and volcanic materials into hydrogeologic units was to be able to discretize the aquifer system for the threedimensional, ground-water flow model. Shown in figure 5 are typical hydrogeologic sections representing the major structural and depositional areas of the aquifer system and the division into hydrogeologic units. Additional sections and descriptions are presented by Pakiser and others (1964), Bateman (1965), Griepentrog and Groeneveld (1981), and Hollett and others (1991).
The criteria for dividing the aquifer system into hydrogeologic units are described in detail by Hollett and others (1991); only a summary is presented here. The first criterion used to divide the aquifer system is a method that defines the hydrogeologic units on the basis of uniform hydraulic properties, commonly represented by geologic or stratigraphic units. This method worked well for some parts of the aquifer system, such as the thick clay beds near Big Pine (section B—B', figure 5), but not for most of it. The second criterion defines hydrogeologic units on the basis of the distribution of vertical head. This method enabled the definition of units in the thick sequences of valley fill where interfingering and lateral discontinuity cause complex heterogeneity, such as beneath much of the valley floor. The third criterion defines hydrogeologic units on the basis of the depth at which significant recharge or discharge can occur. In areas of the Owens Lake Basin where little information is present to differentiate between hydrogeologic units 3 and 4 (section C—C', figure 5), the base of hydrogeologic unit 3 was chosen arbitrarily at 1.5 times the depth of the deepest production well in the area. The following is a brief description of the geologic, stratigraphic, and hydraulic characteristics of each of the hydrogeologic units.
Hydrogeologic Unit 1.—Hydrogeologic unit 1 represents the unconfined part of the aquifer system and includes the water table as the upper boundary surface. Unconfined conditions are areally pervasive throughout the aquifer system, although the depth of significant confinement varies with local conditions. Typically, the upper 100 ft of saturated deposits displays minimal restriction to the vertical movement of water, and differences in hydraulic head usually are less than 2 to 3 ft. In some parts of the aquifer system, confined conditions near the water table can be created by the less transmissive layers of the olivine basalt flows or by a fine-grained fluvial or lacustrine deposit (figures 4 and 14). This type of local confinement near the land surface is not typical of most conditions in the valley, and hydrogeologic unit 1 can be considered generally to have a saturated thickness of about 100 ft.
Hydrogeologic Unit 2.—Hydrogeologic unit 2 is the material, where present, that separates hydrogeologic unit 1 from hydrogeologic unit 3. In the middle of the valley, this material typically consists of fine-grained silt and clay beds that restrict the vertical movement of ground water. Near Big Pine, hydrogeologic unit 2 is composed of a massive, readily identifiable clay bed with a total thickness of more than 80 ft—referred to as the "blue-green clay" by Hollett and others (1991, p. 31 and figure 12). Vertical groundwater flow also is restricted by the volcanic materials of the Big Pine volcanic field even though they are depositionally much different from the fine-grained silt and clay beds. The volcanic material in the aquifer system near Bishop, in contrast, consists mostly of unconsolidated pumice (the lower member of the Bishop Tuff), which has hydraulic properties similar to sand and offers minimal restriction to vertical flow. Along the margins of the valley, the alluvial fan deposits are relatively homogeneous, displaying no dominant horizontal layering. In these areas, hydrogeologic unit 2 is virtually absent.
Hydrogeologic Unit 3.—Several confined zones that are present in the aquifer system have been combined into hydrogeologic unit 3. The confined part of the aquifer system generally extends from the toes of the alluvial fans along the Sierra Nevada to the toes of the alluvial fans along the Inyo and the White Mountains and extends along nearly the full length of the valley (figure 14). Confinement is created by a number of lenticular-to-continuous, flat-lying fluvial and lacustrine clay and silty-clay beds (hydrogeologic unit 2). Confinement also can be created by fine-grained material deposited by mudflows. These confining beds thin to extinction along the margins of the valley. Additional areas of confinement may be formed by the upper member of the Bishop Tuff, where present (figure 5), and by volcanic flows of the Big Pine volcanic field (figure 4), but an absence of data in these areas prevents a more detailed analysis. Saturated thickness of hydrogeologic unit 3 ranges from tens of feet along the margins of the basin to about 500 ft beneath most of the valley floor.
Hydrogeologic Unit 4.—Although not part ofthe defined aquifer system, hydrogeologic unit 4 occupies a large part of the valley fill (figure 5). Despite its large volume, the quantity of ground water flowing through or extractable from hydrogeologic unit 4 probably is minimal. Deep test drilling during 1988 by the Los Angeles Department of Water and Power (E.L. Coufal, oral commun., 1988) showed that most materials at depths greater than about 700 ft do not yield significant quantities of water to wells, generally less than 0.2 ft3/s. Deep volcanic deposits penetrated by drilling near Taboose Creek (figure 14) may yield greater quantities, although no aquifer testing was done. Except at the location of these deep test borings and a few previously drilled deep wells, the chemical and hydraulic characters of hydrogeologic unit 4 are largely undocumented.
Hollett and others (1991) further divided the hydrogeologic units into subunits on the basis of the type of geologic deposit (figure 4). For example, hydrogeologic unit 1 in section C—C' (figure 5) has subunits 1a representing alluvial fan deposits and 1c representing undifferentiated fluvial deposits. Hydrogeologic unit 3 in the same section has subunit 3a representing alluvial fan deposits; subunit 3t representing transition-zone deposits; and subunit 3c representing undifferentiated fluvial deposits. Additional subunits were defined for volcanic deposits and massive clay-bed deposits (figures 4 and 5). The combination of hydrogeologic units and subunits formed the basis of ground-water "model zones" discussed later.
Hydraulic Characteristics
The hydraulic characteristics of the aquifer system—transmissivity, saturated thickness, horizontal and vertical hydraulic conductivities, specific yield, and storage coefficient—were estimated from pumpedwell and aquifer tests, drill-hole data, and geophysical data. Detailed descriptions of the methods used to define the hydraulic characteristics and a general range of horizontal hydraulic conductivity and specific yield for different types of aquifer materials in the Owens Valley are presented by Hollett and others (1991, table 1). Additional confirmation of these values was obtained from preliminary ground-water flow models (Yen, 1985; Danskin, 1988; Hutchison, 1988; Hutchison and Radell, 1988a, b; Los Angeles Department of Water and Power, 1988) and from development and calibration of the final valleywide ground-water flow model documented in this report.
The areal distribution of aquifer characteristics was determined by analyses of all known pumped-well and aquifer tests, at more than 130 wells, in the valley. A complete list of the transmissivity, average horizontal hydraulic conductivity, and storage coefficient obtained from these analyses and the method of calculation (aquifer-test method) are given in table 9 (p. 155). In some cases, several calculations were made for a single well. Values calculated by the Los Angeles Department of Water and Power (M.L. Blevins, written commun., 1984—87) for some wells also were obtained. The values given in table 9 are those most representative of transmissivity unaffected by leakance and of a longer-term storage coefficient that reflects drainage of the aquifer system. These criteria were chosen in part to ensure consistency with the valleywide ground-water flow model. Leakance, if not taken into account in aquifer-test analysis, will tend to increase calculated transmissivity values. Storage coefficient, which is specific yield for water-table conditions, was difficult to calculate from the available tests. None of the values reach the 0.10—0.15 range that is characteristic of a true specific yield of these aquifer materials (Hollett and others, 1991; S.N. Davis, 1969). Much longer aquifer tests probably are required to achieve more representative values of specific yield. Calculation of storage coefficients for confined conditions was somewhat more successful; values typically ranged from 0.0005 to 0.005. Average horizontal hydraulic conductivity was calculated using an estimate of the total saturated thickness of transmissive deposits affected by the well—calculated as the depth of the well below the water table minus the total thickness of clay layers or, if data were available, as the total length of perforations.
The areal distributions of transmissivity and average horizontal hydraulic conductivity are shown in figures 15 and 16, respectively. Both sets of values are well correlated with the distribution of depositional materials (figures 4 and 5). Values for many of the wells near the Los Angeles Aqueduct in the Owens Lake Basin reflect the buried, more transmissive, transition zone deposits (figure 5) rather than the overlying, less transmissive, alluvial fan deposits.
In some cases, the transmissivity values in figure 15 and table 9 represent only a part of the transmissivity of the aquifer system. Some wells are not open to all of the transmissive aquifer materials, especially shallow materials, or the wells may not penetrate the entire depth of the aquifer system, especially in the volcanic areas. For these reasons, extrapolation of transmissivity values to the entire aquifer needs to be done cautiously. Alternatively, average horizontal hydraulic conductivity values (figure 16) multiplied by an estimate of the saturated thickness of the aquifer system may yield more reliable values of transmissivity. Gross estimates of saturated thickness in the center of the valley are 100 ft and 500 ft for hydrogeologic units 1 and 3, respectively. The thickness of hydrogeologic unit 2 is minimal, generally less than 15 ft, except near Big Pine.
Movement of Ground Water
Virtually all the ground water in the Owens Valley aquifer system is derived from precipitation that falls within the Owens Valley drainage basin area (figure 1). Ground-water recharge (deep infiltration) occurs primarily through the alluvial fans as water runs off the Sierra Nevada as a result of snowmelt or rainfall. Most of the runoff infiltrates through the heads of the alluvial fans and through the tributary stream channels. Lesser quantities of recharge result from seepage of water flowing in canals and ditches, from direct precipitation on the sparsely vegetated volcanic rocks, from runoff from bedrock areas within the valley fill, by leakage from the rive-aqueduct system, and as underflow from Chalfant and Round Valleys. Underflow to the Bishop Basin from Chalfant Valley also includes water moving south from Hammil and Benton Valleys. Most of the ground water from Chalfant, Hammil, and Benton Valleys is believed to enter the Bishop Basin near Fish Slough beneath the southeastern part of the Volcanic Tableland (Hollett and others, 1991, p. 63). Recharge to the aquifer system is minimal from percolation of water that moves through fractures in the surrounding bedrock to the zone of saturation or, because of the high evapotranspiration, from water that percolates directly to the water table from rainfall on the valley floor.
Ground water moves along permeable zones of the ground-water system from areas of higher head to areas of lower head. The direction of ground-water flow is approximately perpendicular to lines of equal head. The areal pattern of ground-water flow in the valley is shown in figure 14. The vertical flow directions in hydrogeologic units 1, 2, and 3 are shown in figure 5 and can be inferred from the relative position of equal-head contours for hydrogeologic units 1 and 3 in figure 14. The Darcian rate of flow along the illustrated flow paths is determined by the hydraulic gradient, the hydraulic conductivity, and the cross-sectional area of flow. Typical rates in the valley range from less than a foot per year in clay and silt to hundreds of feet per year in the more permeable basalt. Rates of horizontal flow of water in hydrogeologic units 1 and 3 generally range from 50 to 200 ft/yr. Additional studies of ground-water quality, particularly the analysis of hydrogen and oxygen isotopes, which can be used to determine the relative age of water, would help to confirm these rates of flow.
Ground water flows from areas of recharge to areas of discharge. Discharge can be from springs, wells, evapotranspiration, or seepage to the rive-aqueduct system and the lower Owens River. In general, ground-water flow is from the margins of the valley, mainly the west margin, toward the center of the valley and then southward toward the Owens Lake (figure 14). As ground water flows downgradient to the toes of the alluvial fans and the transition-zone deposits, the flow is primarily horizontal rather than vertical (figure 5). This horizontal flow of ground water is split by the confining beds of hydrogeologic unit 2 that interfinger with the alluvial fan and the transition-zone deposits and direct the flow of water into hydrogeologic units 1 and 3. Discharge from hydrogeologic unit 3 is generally upward through hydrogeologic unit 2 to unit 1, from pumped or flowing wells, or through the valley fill to the south end of the valley. Discharge from hydrogeologic unit 1 is principally to evapotranspiration, pumped wells, springs, the rive-aqueduct system, and the lower Owens River.
In the Bishop Basin, ground water that originates as underflow from Round and Chalfant Valleys and as underflow from the lower member of the Bishop Tuff enters hydrogeologic units 1 and 3. This water mixes with water recharged through alluvial fans and through the Big Pine volcanic rocks and moves southward along the center line of the valley (figure 14). In the Big Pine area, however, the direction of ground-water flow has changed, at least during some periods, since 1970. Increased pumpage from wells near Crater Mountain has shifted the ground-water gradient and caused ground water to flow northwest from the Tinemaha Reservoir and west from the section of the Owens River just north of the reservoir toward Crater Mountain.
In the Owens Lake Basin, water that enters the aquifer system as underflow through the narrows or as recharge through the alluvial fans moves south to the Owens Lake (dry). Most of the water is discharged to evapotranspiration, wells, or the lower Owens River. What happens to the remaining ground water that reaches the south end of the ground-water system at the Owens Lake (dry), however, is not known with certainty. The bulk of the ground water probably flows vertically upward and is discharged as evaporation from the dry lake. Minor quantities of water may flow at depth through the fractured bedrock beneath the Haiwee Reservoir to Rose Valley, which is south of the Owens Valley. Berenbrock and Martin (1991) estimated total underflow from Rose Valley south to Indian Wells Valley to be less than 50 acre-ft/yr, part of which is seepage from the Haiwee Reservoir (Danskin, 1988).
The presence of faults within the aquifer system (figure 4) may affect the movement of ground water, depending on the transmissive characteristics of the individual faults. The physical and chemical processes that cause one fault to retard ground-water movement more than another are discussed by Schaefer (1978), Freeze and Cherry (1979, p. 474) and Hollett and others (1991). Some faults in the Owens Valley, most notably the Owens Valley Fault (figures 4 and 14), significantly retard and deflect ground-water movement. For example, the Owens Valley Fault effectively splits the Owens Lake Basin into two halves. Most ground water flows southward down the west side of the fault; lesser quantities slowly seep over and through the fault to the east side of the basin. The effects of both recharge and pumping on the west side of the basin are isolated to a large extent from the east side of the basin—except in the northern part of the Owens Lake Basin, where the Owens Valley Fault does not appear to impede groundwater movement (compare figures 4 and 14).
Other faults that have a significant regional effect on ground-water flow were noted by Hollett and others (1991, p. 74). Additional water-retarding faults identified since that study was completed include a fault through Red Mountain (figures 3 and 14), en echelon sliver faults near Lone Pine (figures 4 and 14), and a probable, unexposed fault in the vicinity of west Bishop (figures 4 and 14).
Northwest-trending faults along the east side of Crater Mountain (Hollett and others, 1991, figure 15) have created additional fractures in the highly transmissive volcanic deposits. Calibration of the ground-water flow model required much higher transmissivities in this area than for other volcanic deposits in order to maintain the unusually flat water table along the edge of Crater Mountain. These fracture conduits appear to provide an enhanced pathway for ground water recharged in the Big Pine Creek drainage to move southward through Crater Mountain to the vicinity of Fish Springs.
Some of the water-retarding faults force ground water to rise to land surface, producing noticeable seeps and springlines. Many of these features can be identified readily by an increase in vegetation (Meinzer, 1927) and are indicated by linear red zones (false color) in figure 3. An excellent example is the sequence of faults just north of the Alabama Hills (figures 3, 4, and 14) described by D.E. Williams (1970).
In some parts of the Owens Valley, water-retarding en echelon faults have created flow compartments that are relatively isolated from the rest of the aquifer system. Areas with closely spaced faults near Lone Pine and just north of the Alabama Hills are typical of this phenomenon (figure 4). Recharge to the compartments typically is localized, such as from a stream. Discharge may be to a spring or well. Underflow into and out of the compartment depends on the retarding effect of the fault, which may vary with depth. Simulation of these areas, as discussed later, was difficult and not particularly successful.
Hollett and others (1991, figure 6) mapped numerous other fault traces, some of which may be locally important in affecting ground-water movement. Additional site-specific aquifer tests could be used to detect any significant retardation of ground-water flow caused by known or suspected faults in the Owens Valley. Ground-water-level data from an aquifer test show an unexpected change in the rate of drawdown if a flow-retarding fault is within the area of influence of the pumped well (Driscoll, 1986, p. 562).
The movement of ground water in the Owens Valley is controlled to a large extent by springs, seeps, evapotranspiration by native vegetation, and seepage to the rive-aqueduct system and the lower Owens River. Each of these features acts as a "hydraulic buffer" on nearby ground-water levels in hydrogeologic unit 1. As the altitude of the water table increases, discharge from the springs and seeps, by native vegetation, and to the rive-aqueduct system and the lower Owens River increases, thereby restricting the rise in water-table altitude. As the water table declines, discharge from each feature is reduced, thereby reducing the decline in water-table altitude. Without the broad areal distribution of these hydraulic buffers, which cover most of the valley floor, fluctuations in ground-water levels in response to changes in recharge and discharge would be much greater. The action of hydraulic buffers on ground-water levels and on recharge to and discharge from the aquifer system is a recurring theme that is exceptionally important in understanding the operation of the hydrologic system in the Owens Valley and in evaluating the effect of different water-management alternatives.
Ground-Water Budget
A ground-water budget is an accounting of the inflow to and outflow from a ground-water system (in this case, the defined aquifer system) and the changes in the volume of ground water in storage. If inflow equals outflow and if the change in the volume of ground water is zero, then the aquifer is in equilibrium or a steady-state condition. Equilibrium is reflected by nearly constant ground-water levels or by even fluctuations of levels with no long-term rise or decline. If total inflow does not equal total outflow, then the aquifer is in nonequilibrium or a transient condition, and the change in the volume of ground water in storage is reflected in the changing ground-water levels.
In several previous investigations, water budgets have been summarized for the whole hydrologic system in the Owens Valley. The investigators include C.H. Lee (1912), Conkling (1921), California Department of Water Resources (1960), D.E. Williams (1969), Los Angeles Department of Water and Power (1972, 1974b, 1975, 1976, 1978, and 1979), Griepentrog and Groeneveld (1981), and Hutchison (1986b).
Each of the water budgets, except that of Hutchison (1986b), was reviewed by Danskin (1988). In comparing the respective components of inflow and outflow, he noted that comparisons were difficult because each of the studies covered different areas or different periods of time. In addition, some of the water budgets used the same components of inflow and outflow, but with different definitions. A complete analysis of the hydrologic system of the Owens Valley, he concluded, would require at least three interrelated water budgets for the valley-fill part of the drainage basin area—a total budget for both saturated and unsaturated materials, including all precipitation and evapotranspiration; a budget for the surface-water system; and a budget for the ground-water system. To facilitate verification and comparisons, the budgets would need to cover the same area and time period and use similarly defined components.
The synthesis of three complex, interrelated water budgets was outside the scope of this study; however, significant progress in that direction has been made by development of a detailed ground-water budget (tables 10 and 11) [table 11 in pocket]. In addition, data have been collected and summarized and predictive relations have been developed for precipitation, evapotranspiration, and tributary streamflow. Eventual development of the three interrelated budgets would be needed to further refine the ground-water budget presented in this report.
The ground-water budget for the defined aquifer system shown in figure 14 is summarized in table 10. Each component of the ground-water budget is defined and discussed more fully by Hollett and others (1991). The values in table 10 are revised slightly from those presented by Hollett and others (1991, table 6), but they were developed using identical concepts and methods. Development of the ground-water budget involved using data from previous studies, new evapotranspiration and stream-loss data collected during this 6-year study, and results of simulation of the aquifer system described later in this report.
Average values for each component are given in table 10 for two time periods, water years 1963—69 and water years 1970—84. The first period represents average conditions in the aquifer system prior to increased pumpage and additional export of water from the valley (table 4). The second period represents conditions after pumpage and exports increased. The uncertainty of each value for the second period was estimated, and the likely range of values is given.
Ground-water budgets, such as the two given in table 10, can be useful in making semi-quantitative evaluations of an aquifer system, but budgets can be misinterpreted or misused quite easily (Bredehoeft and others, 1982). For example, the approximation of equilibrium is rarely satisfied over an entire system that has been modified by human activity. Localized areas in the Owens Valley likely will be undergoing change for years or decades as a result of human intervention. Changes in recharge or discharge, such as occurred in 1913 and 1970, are reflected in changes in the magnitude of several different components of the water budget (compare tables 4 and 10). In general, the interaction between the components is complex and the magnitude of the changes to the hydrologic system cannot be estimated from the budget alone. For this reason, numerical simulation is a critical part of understanding the operation of the aquifer system and the potential effects of water-management decisions.
The following components of the ground-water budget are not linked to a specific surface-water feature and were not discussed in previous sections of this report.
Discharge from Pumped and Flowing Wells
Discharge from wells includes discharge from both pumped and flowing wells, although the quantity from flowing wells is much less and is limited to a few wells along the Owens River south of Bishop and a few wells in the Independence area near the aqueduct. Several of the flowing wells also are equipped with pumps, and thus discharge sometimes is free-flowing ground water and sometimes is pumped ground water. In this report, all discharge from pumped and flowing wells is referred to informally as "ground-water pumpage."
Nearly all ground-water pumpage is from production wells owned and operated by the Los Angeles Department of Water and Power. Most of these wells provide water for export; a few wells supply water for ranching operations and to the four major towns; and four large-capacity wells supply water to two fish hatcheries. Some additional pumpage is from private domestic and agricultural wells. Distribution of the wells (figure 17) generally follows the rive-aqueduct system. In fact, a few of the present production wells were installed in the early 1900's for dewatering and water supply during construction of the first aqueduct. Division of the wells into well fields shown in figure 17 was done on the basis of general location of the wells and included all wells with production during water years 1963—88, as reported by the Los Angeles Department of Water and Power (M.L. Blevins, written commun., 1988; table 11). The well fields identified in figure 17 and used elsewhere in this report are similar to those defined by the Los Angeles Department of Water and Power (1979, figure 4-4; Hollett and others, 1991, figure 18).
Annual pumpage for individual wells for water years 1963 through 1988 was obtained from the Los Angeles Department of Water and Power (M.L. Blevins, written commun., 1988). Pumpage for water years 1963—69 was copied from typed summary sheets of well discharge per month. Pumpage for water years 1970—71 was estimated by interpolating between instantaneous discharge readings for each well. Pumpage for water years 1972—88 was obtained directly from computerized files.
Average pumpage in most areas of the Owens Valley changed dramatically after 1970, as shown by the inset graphs of well-field discharge in figure 17. Within the defined aquifer system (figure 14), total pumpage averaged about 20,000 acre-ft/yr during water years 1963—69 and about 98,000 acre-ft/yr during water years 1970—84 (table10). Much of this increase was caused by the switching from surface to ground water by two major fish hatcheries. The fish hatcheries, Fish Springs and Blackrock, are located near Fish Springs and Big Blackrock Springs, respectively (figure 17). Average pumpage changed again in 1987 with the addition of new "enhancement and mitigation" wells, which were used to provide water for selected recreation and wildlife projects throughout the Owens Valley (table 4; Los Angeles and Inyo County, 1990a, p. 5—20).
The total quantity of ground-water pumpage varies each year with the quantity of runoff. In years of greater runoff, less pumpage is required for in-valley uses or for export. Pumpage also depends on the quantity of runoff in the preceding year, as shown in figure 18. When antecedent conditions are wet, the rive-aqueduct system is full, and pumpage is less.
Discharge from different hydrogeologic units was investigated by analyzing each well. The first significant clay layer, as identified from the lithologic well log, was used to mark the separation between hydrogeologic units 1 and 3. Discharge from each well then was apportioned as withdrawal from hydrogeologic units 1 and 3 (upper and lower model layers) on the basis of length of perforations and estimated hydraulic conductivity of the adjacent material in hydrogeologic units 1 and 3, respectively. In most parts of the valley, well withdrawals are primarily from hydrogeologic unit 3 (figure 17). Near the Big Pine volcanic field, many wells tend to be shallow, and most water is withdrawn from the highly transmissive volcanic deposits near the land surface (figures 4 and 5).
Springs and Seeps
Most springs in the Owens Valley are near the toes of alluvial fans and along the edge of volcanic deposits near the Poverty Hills (figure 17). A few springs are caused by faulting as indicated by an obvious surface trace (figure 3; Hollett and others, 1991, figure 15). Historically, springs have discharged a large quantity of water, most of which eventually flowed into the rive-aqueduct system. For example, Fish Springs near Crater Mountain discharged as much as 22 ft3/s prior to 1970. When ground-water pumpage increased in 1970, discharge at springs dropped dramatically, to zero at some. Average discharge from major springs within the defined aquifer system was about 33,000 acre-ft/yr during water years 1963—69 and about 8,000 acre-ft/yr during water years 1970—84. About 20 percent of this discharge was estimated to return to the aquifer system as recharge in the immediate vicinity of the springs (Hollett and others, 1991). Net discharge from the aquifer system was about 26,000 and 6,000 acre-ft/yr for the two periods, respectively (table 10).
Seeps occur along some faults where groundwater flow is forced to the land surface and along the toes of alluvial fans where ground water flows out onto the valley floor. The major seeps (shown in figures 3 and 17) discharge an unknown quantity of water, nearly all of which is evapotranspired by nearby vegetation.
Springs and to a lesser extent seeps, such as the Independence "springfield" (figure 17), act as hydraulic buffers and exert a strong local influence on the aquifer system. The maximum altitude of the water table, particularly near the Poverty Hills, is controlled by the altitude of nearby springs and the transmissive properties of the adjacent deposits (figures 14, 15, and 17). Fish Springs, for example, prior to an increase in nearby pumpage in 1970, was exceptionally effective at dampening fluctuations in nearby ground-water levels [well 224, pl. 1 (in pocket)]. In the Big Pine area, an increase in recharge to the aquifer resulted in an increase in discharge from Fish Springs and only a minimal rise in ground-water levels near the spring; a decrease in recharge to the aquifer resulted in a decrease in discharge from Fish Springs and only a minimal decline in ground-water levels near the spring. After 1970, the buffering effect of springs near the Poverty Hills (figure 17) was reduced, and changes in aquifer recharge and discharge resulted in greater fluctuations in ground-water levels.
Underflow
Underflow into and out of the aquifer system occurs at several locations shown in figure 14. Underflow from three drainages (Bishop and Big Pine Creeks and Waucoba Canyon) originates as recharge from tributary streams outside the aquifer system. For that reason, the quantity of underflow from those areas, totaling about 500 acre-ft/yr, is included for waterbudget purposes as part of tributary stream recharge (table 10).
The quantity of underflow from Round Valley, the Volcanic Tableland, and Chalfant Valley is much greater and was estimated to average about 4,000 acre-ft/yr (table 10). Prior estimates of underflow from theseareas were significantly higher, totaling as much as 25,000 acre-ft/yr. These estimates were based on Darcy's law (Los Angeles Department of Water and Power 1972, 1976, 1978, 1979) and on steady-state ground-water-model simulations (Danskin, 1988). As shown in table 10, the quantity of underflow into the aquifer system is not known with certainty. However, the present estimates, which are consistent with results from several different ground-water flow models developed during the cooperative USGS studies, probably are more accurate than previous estimates. The models also are based on Darcy's law, but they have additional advantages; these include incorporating nearby ground-water recharge and discharge, accounting for changes in ground-water storage, and matching various historical conditions (calibration).
Underflow out of the aquifer system occurs only across an arbitrary east—west line south of Lone Pine. In the area east of the Alabama Hills, most ground water flows out of the aquifer system through hydrogeologic unit 3, which is thicker and more transmissive than hydrogeologic unit 1. In the area west of the Alabama Hills, hydrogeologic units 1 and 3 act together, and there is no clear distinction between the two units, or indication of the relative quantity of underflow from each. Total underflow from both areas was estimated to be about 10,000 acre-ft/yr. This estimate is based on calibration of the valleywide ground-water flow model and on a water-budget analysis of the Owens Lake area by Lopes (1988). No difference in the quantity of underflow before and after 1970 was detected (table 10).
Irrigation and Watering of Livestock
Irrigation of agricultural and pasture land is still (1988) prevalent in the Owens Valley (figure 3), although the total acreage of irrigated lands and the quantity of water applied to irrigated lands is much less than in previous years (D.E. Babb and R.H. Rawson, Los Angeles Department of Water and Power, written commun., 1988). The most recent change in watermanagement practices in the Owens Valley occurred in about 1968 in anticipation of providing sufficient water to fill the second aqueduct (table 4). Some land was taken out of production. Historical agricultural practices that resulted in an excessive application of water, such as using flood irrigation, were discouraged. Fields were leveled and irrigation sprinklers were installed. Water supplied by the Los Angeles Department of Water and Power to lessees was reduced from about 6 acre-ft/acre to about 5 acre-ft/acre. Watering of livestock, which typically involves diverting surface water from a canal or ditch and flooding a small area of the land surface, continued, but to a lesser degree. As a result, the total recharge from both irrigation and stock watering decreased, and the salvaged water was available for export.
Recharge to the aquifer system from irrigation and watering of livestock was estimated from maps of land use compiled by the Los Angeles Department of Water and Power (R.H. Rawson, written commun., 1988). Digitized map information was combined with assumptions about the quantity of water supplied and used per acre and the likely recharge rates on different types of soils. For years prior to 1970, water applied on volcanic materials was assumed to recharge at a rate of 24 in/yr, and water applied on other permeable materials, at a rate of 12 in/yr. For 1970—84, these rates were reduced to 12 in/yr and 6 in/yr, respectively. On the basis of these assumptions, the average recharge from irrigation and watering of livestock within the aquifer system (figure 14) was estimated to be about 18,000 acre-ft/yr in water years 1963—69 and about 10,000 acre-ft/yr in water years 1970—84 (table 10).
Ground-Water Quality
Ground water in most parts of the Owens Valley has a preponderance of calcium and bicarbonate ions, and the range of concentrations for dissolved constituents is small (Hollett and others, 1991, figure 21). Concentrations of dissolved solids are generally less than 300 mg/L. However, at the extreme southern end of the basin near the Owens Lake, ground-water quality is much different. A well named "Dirty Socks" (Hollett and others, 1991, figure 18) was found to have markedly different water quality—mostly sodium, chloride, and bicarbonate ions and a concentration of dissolved solids greater than 5,000 mg/L.
In 1973—74, the Los Angeles Department of Water and Power (1974a) conducted an areally extensive study of ground-water quality that included samples from selected wells in each well field (figure 17). Although the study focused primarily on drinkingwater standards (California Department of Health Services, 1983; U.S. Environmental Protection Agency, 1977a, b, 1986), results did not reflect any major differences in ground-water quality throughout most of the valley. It was also concluded in the study that no significant changes have occurred in ground-water quality in the valley during the past 10 to 35 years.
One area of exception was noted, however. On the basis of earlier data, ground-water quality just south of the Tinemaha Reservoir seemed to be different and possibly changing from 1972 to 1973 (Roland Triay, Jr., Los Angeles Department of Water and Power, written commun., 1973). Alkalinity for wells near the Taboose—Aberdeen well field (table 9, wells 118, 349, and 116) increased between June 1972 and April 1973 by as much as 90 percent. One possible explanation is that the extensive pumping from 1970 to 1973 (figure 17) induced movement of water from the east side of the valley toward the Taboose—Aberdeen well field. Ground water in contact with sedimentary and metamorphic rocks along the east side of the valley likely has a higher concentration of dissolved solids and a higher alkalinity than does ground water in contact with granitic rocks and near the dominant recharge areas on the west side of the valley. The significant drawdown observed at nearby wells (pl. 1, wells 362 and 347), a steep hydraulic gradient from east to west, and a pattern of increasing dissolved-solids concentration from west to east lend credibility to this explanation.
Another possible explanation is that dissolution and mobilization of soluble minerals in nearby finegrained deposits caused the observed changes in ground-water quality (Roland Triay, Jr., Los Angeles Department of Water and Power, written commun., 1973). Also, the increased hydraulic gradient may have induced vertical movement of ground water of different quality from an adjacent part of the aquifer. Additional localized water-quality studies would help in identifying the specific flow paths of ground-water movement, particularly as influenced by pumping and artificial recharge.
More generally, a complete inventory of groundwater quality in the Owens Valley is needed to confirm ground-water concepts presented in this report and by Hollett and others (1991). Many of the older wells are open to a combination of hydrogeologic units 1, 2, and 3. Water-quality data from these wells are ambiguous and difficult to interpret. Recently installed production and observation wells that are open only to specific strata offer the opportunity to sample ground-water quality for specific hydrogeologic units of the aquifer system. Also, some of the new wells are located near and some far from areas of recharge and discharge. Water-quality information from these new wells could aid considerably in confirming the areal and vertical ground-water flow paths (figure 14), and in identifying likely changes in flow paths. The water-quality characteristics of interest are major and minor ions; trace metals; nitrate and nitrite; hydrogen, oxygen, and carbon isotopes to date the water and identify different sources of recharge; and possibly pesticides or organic contaminants to document issues of public health.
Studies of oxygen- and hydrogen-isotope concentrations across much of southern California by Gleason and others (1994) revealed strong regional differences. Ground water from eight wells in the Owens Valley had less deuterium (that is, was much "lighter" in hydrogen isotopes) than did ground water in basins to the east and south. This trend implies that the dominant recharge to the Owens Valley groundwater basin comes from precipitation from storms that are moving westward. No trend within the Owens Valley could be detected from the scant number of samples. Although storm cells originating to the south may be important in providing water for native vegetation, the quantity of recharge to the groundwater system from such storms is much less than the quantity of recharge resulting from runoff from the Sierra Nevada.
Ground-Water Flow Model
A valleywide ground-water flow model was developed to integrate and test the concepts about the structure and physical properties of the aquifer system, the quantity of recharge and discharge, and the likely effects of water-management decisions. A numerical ground-water flow model, such as the valleywide model, is a group of mathematical equations that describe the flow of water through an aquifer. Variables (parameters) in the equations include hydraulic heads, transmissive characteristics, storage characteristics, and the rates of inflow and outflow. Different values for each variable, such as transmissivity or pumpage, can be distributed throughout the area being modeled in order to simulate observed spatial and temporal variations. This general technique is referred to as a distributed-parameter approach in contrast to a lumped approach, which uses a single value for each type of parameter.
Even when using a distributed-parameter approach, however, not all characteristics of the actual aquifer system can be included in the ground-water flow model. Simplifying assumptions are required to make the modeling effort manageable. Many of the assumptions used in developing the Owens Valley ground-water flow model are characteristic of most numerical ground-water flow models. Explanations of these assumptions are given by Remson and others (1971), Durbin (1978), Freeze and Cherry (1979), Wang and Anderson (1982), and Franke and others (1987). Assumptions underlying the particular computer program used in this study are described by McDonald and Harbaugh (1988). Additional assumptions made in the application of the computer program to the Owens Valley aquifer system are discussed in the next sections of this report.
For purposes of clarity in this report, hydraulic head (head) is used when referring to simulated hydraulic potential, which is well defined and has a precise x—y—z location. Ground-water level (level) is used when referring to general concepts of groundwater flow and to measured data, which are less well defined vertically and often represent a composite hydraulic potential.
Although a simulation model is only an approximation of the real world, it can be extremely useful in gaining an improved understanding of a complex system—in this case, a ground-water system interacting with many surface-water features. A ground-water flow model assures that estimates of local aquifer characteristics, the water budget, and hydraulic heads all are compatible. It is this attribute that gives additional confidence in the concepts and quantities presented in this report and in those described by Hollett and others (1991). In areas where data are sparse or uncertain, the ground-water flow model can be used to test the reasonableness of assumed values. Finally, a calibrated model—one for which all the parameter values are acceptable—can be used to compare the likely effects of different watermanagement alternatives.
General Characteristics
The computer program developed by McDonald and Harbaugh (1988) uses standard finite-difference techniques to approximate the partial differential equations that describe saturated ground-water flow. General characteristics of the numerical code include division of a ground-water system into finite-difference cells, each with uniform hydraulic properties. Multiple layers can be identified and linked with Darcy's law. A variety of different types of recharge and discharge can be simulated with constant-head, head-dependent, or specified-flux terms. Transmissivity can be constant or calculated as the product of hydraulic conductivity and saturated thickness. Both steady-state and transient conditions can be simulated, each with its own formulation. Several solvers are available, including those provided by Hill (1990a,b) and Kuiper (1987a,b) that constrain convergence of the solution using both head and mass-balance terms. The computer code is stable and flexible, and it is widely used in the public and private sectors.
Application of the numerical code to the aquifer system of the Owens Valley involved the use of two model layers. Flow between the layers was approximated by a relation that uses calculated head in vertically adjacent cells and an estimate of "vertical conductance" between the cells. Vertical conductance is calculated from vertical hydraulic conductivity, thickness between the layers, and horizontal area of the cell (McDonald and Harbaugh, 1988, p. 5—11). Transmissivity was varied between groups of model cells (model zones), but was assumed to remain constant over time. Specified flux terms were used to approximate discharge from wells and recharge from precipitation, tributary streams, canals, and ditches. Head-dependent relations were used to simulate springs, evapotranspiration, and interaction of the aquifer system with the rive-aqueduct system and the lower Owens River. A 26-year simulation period included water years 1963—88 and used annual approximations of recharge and discharge.
A geographic information system (GIS) was developed to ensure an accurate spatial control of physical features and the finite-difference model grid. This accuracy was critical in linking map information, such as the vegetative mapping by the Los Angeles Department of Water and Power (figure 9), the valleywide ground-water flow model, and the several more detailed ground-water flow models developed by Inyo County and the Los Angeles Department of Water and Power (table 2). The original digitizing of geologic and hydrologic information was done in latitude and longitude coordinates, using the North American Datum 1929, from maps with scales of 1:24,000 and 1:62,500. Replotting was done using a Universal Transverse Mercator (UTM) projection (Newton, 1985). This GIS methodology was used for all map illustrations in this report and in Hollett and others (1991). Because of the accuracy of the GIS method, subsequent computer scanning of the map illustrations should produce an accuracy of approximately 0.01 in. and permit registration with other maps drawn from a UTM projection. Detailed information on GIS and UTM mapping systems is given by J.P. Snyder (1982, 1985, 1987) and Newton (1985).
As part of the GIS system, the finite-difference model grid was linked mathematically to latitude and longitude and the UTM coordinate system. Coordinates of the finite-difference model grid are given in table 12. Projection and translation of coordinate systems (latitude-longitude, UTM, model) were done using computer programs based on those developed by Newton (1985). Use of the coordinates in table 12 and similar computer projection programs will enable future investigators to reproduce the model locations precisely. Use of this technique reduces any differences caused solely by spatial discretization and aids in duplicating specific results presented in this report.
Representation of the Aquifer System
Boundaries of the ground-water flow model conform to the physical boundaries of the Owens Valley aquifer system as shown in figure 14 and as described by Hollett and others (1991). Lateral underflow boundaries are present in eight locations: Chalfant Valley, the edge of the Volcanic Tableland, Round Valley, Bishop Creek, Big Pine Creek, Waucoba Canyon, and east and west of the Alabama Hills. All other boundaries of the aquifer system were assumed to be impermeable and were simulated with no-flow boundary conditions. The top of the aquifer system is the water table, and the bottom is either bedrock, the top of a partly consolidated unit, or an arbitrary depth based on the depth of production wells. Hydrogeologic unit 4 (figure 5) lies below the aquifer system in the center of the valley and is a poorly transmissive part of the ground-water system. Simulation studies by Danskin (1988) concluded that this unit could be eliminated from future ground-water flow models with little loss of accuracy in the upper 1,000 ft of more transmissive materials. Round Valley and the Owens Lake area also were excluded as suggested by Danskin (1988), primarily for computational reasons and because the areas were peripheral to the specific objectives of this study. Future simulation studies with more powerful computer capabilities may find that including both areas is an advantage in analyzing some water-management questions as well as in eliminating the use of specified-flux boundary conditions.
Division of the aquifer system into hydrogeologic units and model layers is more complex and somewhat more arbitrary than the selection of boundary conditions. For this study, the aquifer system was simulated using two model layers. The upper model layer (layer 1) represents hydrogeologic unit 1, the unconfined part of the aquifer system. The lower model layer (layer 2) represents hydrogeologic unit 3, the confined part of the aquifer system. Each model layer is composed of 7,200 cells created by 180 rows and 40 columns (pl. 2, in pocket). The active area of ground-water flow (active model cells) is the same in both model layers.
This division of the aquifer system permits simulation of the measured ground-water levels, which generally are either for shallow wells that monitor unconfined conditions or for deeper wells that monitor a composite confined zone. The use of two layers is consistent with the assumption that both unconfined and confined storage conditions are present in some parts of the valley (figure 14).
To test the value of additional model layers, a smaller, more detailed ground-water flow model was developed to simulate conditions in the Big Pine area (P.D. Rogalsky, Los Angeles Department of Water and Power, written commun., 1988). Although three layers were used in the model in order to more closely approximate the complex layering of volcanic and fluvial deposits described by Hollett and others (1991), results from the more detailed model were not significantly different from results obtained using the valleywide model.
Hydrogeologic unit 2, as defined by Hollett and others (1991), usually represents either a massive clay bed, such as the blue-green clay near Big Pine (figure 5, section B—B'), or overlapping lenses or beds, which are more typical of the valley fill. The Darcian relation that simulates vertical flow between the model layers was used to approximate the vertically transmissive properties of hydrogeologic unit 2. Storage characteristics of hydrogeologic unit 2 were included in the storage coefficients of the surrounding model layers. This formulation is typical of most models used to simulate groundwater movement in unconsolidated, poorly stratified deposits, such as those in the Owens Valley (Hanson and others, 1990; Berenbrock and Martin, 1991; and Londquist and Martin, 1991).
Along the edge of the basin, the clay beds thin, and hydrogeologic unit 2 virtually disappears (figure 5, section C—C'). In these areas, a high value of vertical conductance was used, allowing water to move between the model layers with minimal resistance. The spatial distribution of vertical conductance and its relation to hydrogeologic model zones are shown on plate 2.
In some parts of the valley, hydrogeologic unit 2 represents volcanic deposits, such as those near Big Pine (figure 5, section B—B'). The volcanic deposits have a high transmissivity but can restrict the vertical movement of water as a result of the depositional layering of individual volcanic flows. Where faulted or highly brecciated, the volcanic deposits of hydrogeologic unit 2 were represented by a high value of vertical conductance. As with other deposits represented by hydrogeologic unit 2, the transmissivity of the volcanic deposits was included in the model layer that best approximates the storage properties of the deposit—usually the upper model layer, which represents unconfined conditions.
To facilitate modeling, the aquifer system was divided into model zones, each representing part of a hydrogeologic unit or subunit (Hollett and others, 1991, pl. 2). This technique was shown to be effective in preliminary model evaluations (Danskin, 1988), although the use of additional model zones was suggested in order to simulate key areas of the basin, such as along the toes of alluvial fans. Therefore, development of the valleywide model included additional model zones—specifically, zones to represent the transitionzone deposits. Each model zone represents similar geologic materials that have fairly uniform hydraulic properties. In the volcanic areas of the basin, maintaining this uniformity was not possible. Instead, a single model zone included highly transmissive volcanic deposits along with other much less transmissive fluvial deposits (figure 5). For these zones, the presence of volcanic deposits dominated the hydraulic properties. Outcrops of volcanic flows and cinder cones on the land surface identified likely locations of volcanic deposits in the subsurface. The actual presence of volcanic deposits was confirmed using lithologic information from well logs wherever possible. Calibration of the model was necessary to refine the locations and hydraulic properties of the volcanic zones.
A likely range of transmissivity for each model zone was determined by using the values given in table 9 and the distribution shown in figure 15. In some areas of the basin, however, little or no data were available. In these areas, the depositional models described by Hollett and others (1991, figure 14) were used to extrapolate data and concepts. This technique based on general depositional models with specific data points throughout the aquifer system worked surprisingly well. Values of average horizontal hydraulic conductivity (figure 16) times estimated saturated thickness were compared with estimated transmissivity values in each zone in order to ensure consistency of hydraulic conductivity, saturated thickness, and transmissivity. Other methods of interpolating transmissivity, such as kriging (Journel and Huijbregts, 1978; Sampson, 1978, 1988; Yeh, 1986), were evaluated and found to be of little use in the faulted, complex structure of the Owens Valley (figures 4 and 5).
The transmissivity of volcanic areas was determined by means of arithmetic weighting of the estimated hydraulic conductivity and thickness of volcanic deposits with that of the surrounding sand, gravel, and silt deposits. Not surprisingly, the exceptionally transmissive volcanic deposits dominated the value of all zones where they were present (pl. 2). Only a few electric logs were available, but lithologic well logs were of great value in identifying the general type of depositional material and its appropriate zone.
Transmissivity in all areas of the model was assumed to remain constant over time (pl. 2). This assumption implies that saturated thickness of the model layer—particularly the upper, water-table layer—does not change significantly during model simulations. Changes in saturated thickness may result in differences in computed heads as a result of a mathematical nonlinearity in the ground-water-flow equations (Bear, 1979, p. 308). Because of the relative thinness of hydrogeologic unit 1, a 20-foot change in saturated thickness of unit 1 produces a 10-percent greater fluctuation in nearby water-table altitude than that predicted by the model. The modeling option to vary transmissivity over time (McDonald and Harbaugh, 1988, p. 5—10), however, creates its own set of problems. These problems include the need for significantly more detailed data for model construction and the conversion from active to inactive model cells when dewatered conditions are simulated. For the Owens valleywide model, these problems outweighed the benefits gained by varying transmissivity over time.
Vertical conductance between the two model layers was estimated from aquifer tests, development of preliminary dewatering and cross-sectional models (figure 2), and calibration of the final valleywide model. A high correlation was found between the value of vertical conductance and the type of material in the lower model layer. In most instances, the thicker lower model layer contributed most of the impediment to vertical ground-water flow. As a result, the values of vertical conductance were keyed to the model zones representing the lower model layer (pl. 2).
Faults that restrict ground-water movement (figure 14) were represented by lower values of transmissivity in model cells. The ratio of reduced transmissivity caused by the fault to transmissivity of adjacent aquifer materials is noted on plate 2. For example, a section of the Owens Valley Fault (F20) was determined to reduce transmissivity of the aquifer materials for that zone by a factor of 20—from 80,000 to 4,000 (gal/d)/ft.
Approximation of Recharge and Discharge
The physical characteristics of recharge to and discharge from the aquifer system are described in detail in earlier sections of this report, specifically in the sections entitled "Surface-Water System" and "Ground-Water Budget." The following discussion describes only the approximations of ground-water recharge and discharge that were made in order to simulate these processes in the ground-water flow model. The type of boundary condition and method of approximation for each recharge and discharge component are given in table 13. Annual values for each component for water years 1963—88 are given in table 11, along with the derivation of the value (measured, estimated, or calculated by the model). The areal distribution of each recharge or discharge component in the model and the average values for each model cell for water years 1970—84 are shown on plate 3 (in pocket).
Well package.—Most of the recharge and discharge components were simulated using the well package of McDonald and Harbaugh (1988, p. 8—1). This package simulates extraction of a defined quantity of water from a specific cell in the ground-water flow model. Annual estimates for several recharge and discharge components (table 13) were combined in a pre-processing program, and the net result was used as input for the well package. In most areas of the model, only a few values in the well package represent actual discharge from wells (pl. 3F). Estimated flux for individual items, such as for a stream or an area of ground-water recharge, was distributed uniformly to all model cells related to that item. For example, recharge for a specific stream was the same for each model cell along its length. The individual items are listed in table 11. A few components (precipitation, spillways, and underflow) were assumed to have a virtually constant recharge or discharge rate from one year to another, and were simulated with a constant value for water years 1963—88. Recharge from irrigation and watering of livestock was simulated as having a constant rate for each of two periods, water years 1963—69 and 1970—88. All other components were simulated as having different annual values. Any major changes that were made to initial estimates of recharge and discharge components simulated by the well package are described below.
Some canals, ditches, and ponds probably gain water from the aquifer system, at times, instead of acting as recharge components (table 13). To attempt to account for this dual character, a head-dependent relation (in particular, the river package described below) was used to approximate some of the larger canals during development of the detailed groundwater flow model of the Bishop area (Hutchison, 1988). This technique, however, was found to dampen fluctuations in ground-water levels too severely, and it was abandoned.
Estimates of recharge from ponds were not changed, except for an initial estimate of a 90-percent percolation rate for purposeful ground-water recharge in the Laws area. This rate produced poor model results, and it was reduced during calibration to 75 percent.
Pumpage for each well was assigned to individual model cells using the map-projection and translation programs described in the previous "General Characteristics" section of this report and the well-location information given in table 9. Distribution of average measured pumpage from both model layers is shown on plate 3F.
Underflow was approximated, at first, using Darcy's law. The calculated quantities of underflow were distributed along the flow boundary on the basis of estimated transmissivities. These initial estimates of underflow had a high degree of uncertainty associated with them, and they did not work well in the model; subsequently, they were reduced significantly during calibration plate 3G.
River package.—Permanent surface-water bodies exchange water with the aquifer system—gaining water if nearby ground-water levels are higher than the surface-water stage, and loosing water if nearby levels are lower. A head-dependent relation, referred to as "the river package" by McDonald and Harbaugh (1988, p. 6—1), permits simulation of this type of interaction. The quantity of water exchanged is calculated by the model from the average stage of the stream, altitude of the bottom of the streambed, transmissive properties of the streambed, and model-calculated head for the upper model layer.
In order to simulate different surface-water features (table 13), the average stage and altitude of the bottom of the streambed (or equivalent riverbed or lakebed) were estimated for each model cell from values of land-surface datum obtained from 1:62,500-scale USGS topographic maps. For the Owens River, the Los Angeles Aqueduct, and the lower Owens River, the slope of the river stage from upstream to downstream model cells was checked to ensure that the slope was relatively smooth and uniformly downhill. The concrete-lined, nearly impermeable section of the Los Angeles Aqueduct near the Alabama Hills was not included in the model.
A "conductance" term is used in the river package to incorporate both the transmissive properties of the streambed and the wetted area of the surfacewater feature. The transmissive properties of the streambed (bottom sediment) for each feature were estimated from typical values for valley-fill deposits (table 9; Hollett and others, 1991, table 1) and later were modified during calibration. For example, values of conductance for the lower Owens River were decreased somewhat from values for the Owens River in the Bishop Basin because deposits near the river in the Owens Lake Basin are characteristically finer and less transmissive. The wetted area of each feature was estimated from topographic maps, photographs, and field reconnaissance.
The Pleasant Valley Reservoir was not simulated explicitly in the model, although recharge from the reservoir was considered in selecting values of underflow and in evaluating the simulated gain of water by the Owens River immediately downstream from the reservoir. Use of the river package to simulate sewage ponds near the four major towns was physically realistic, but the parameters and results are highly uncertain.
Evapotranspiration package.—Evapotranspiration was calculated in the model from a piecewiselinear relation, a series of connected straight-line segments, that is based on depth of the water table below land surface (McDonald and Harbaugh, 1988, p. 10—3). An assumption was made that evapotranspiration ceases when the water table is more than 15 ft below land surface (Groeneveld and others, 1986a; Sorenson and others, 1991). When the water table is at land surface, a maximum evapotranspiration rate is reached. At intermediate depths, the evapotranspiration rate linearly decreases from the maximum rate to zero.
The average maximum evapotranspiration rate for vegetation on the valley floor was estimated to be 24 in/yr for the period prior to 1978. This estimate is based on measured evapotranspiration (table 5), results from previous modeling (Danskin, 1988), and measurements of transpiration by Groeneveld and others (1986a, p. 120). The dramatic increase in average pumping after 1970 and the drought of 1976—77 were assumed to permanently decrease the maximum vegetative cover on the valley floor. As a result, the maximum evapotranspiration rate was reduced by 25 percent from 24 in/yr to 18 in/yr for the period after 1977. This reduction was based on the reduced quantity of water available for evapotranspiration (table 10), on the variability of maximum evapotranspiration rates (table 5), and on the observed response to decreased water availability (Sorenson and others, 1991).
The maximum evapotranspiration rates used in the ground-water flow model (28 or 24 in/yr) were chosen to represent the broad areas of native vegetation covering most of the valley floor. These rates tend to underestimate evapotranspiration from riparian vegetation, for which evapotranspiration exceeds 40 to 60 in/yr (D.P. Groeneveld, Inyo County Water Department, written commun., 1984; Duell, 1990). In particular, along the lower Owens River, evapotranspiration is influenced greatly by an abundance of high-water-use cattails (figure 10C). As a result, evapotranspiration calculated by the model underestimates the actual evapotranspiration near the lower Owens River, possibly by as much as 2,000 acre-ft/yr. Most of this extra discharge, however, is simulated by the river package as a gain to the lower Owens River. The net effect on the aquifer system is the same although the accounting is different. This artifact of the model is recognized as potentially confusing, but it does not alter any of the basic conclusions presented in this report.
Drain package.—Springs and seeps were simulated with the head-dependent relation referred to as "the drain package" by McDonald and Harbaugh (1988, p. 9—1). This relation uses a value of the transmissive properties (conductance) of the spring and the simulated model head to compute a discharge—if the model head is higher than a specified drain altitude. If the model head is lower, discharge is zero. The drain altitudes were chosen on the basis of a leveling survey of each spring (R.H. Rawson, Los Angeles Department of Water and Power, written commun., 1988), or on a value of land surface obtained from 1:62,500-scale USGS topographic maps.
Simulation Periods
Simulation periods were chosen to calibrate and verify the ground-water flow model, to evaluate past water-management practices, and to predict the likely condition of the aquifer system after 1988. Historical periods of similar water use, as summarized in table 4, were used as an aid in selecting simulation periods that capture the main elements of water management in the Owens Valley and rigorously test the model.
Water year 1963 was chosen to calibrate the ground-water flow model under equilibrium or steady-state conditions. This particular period was chosen for three reasons. First, ground-water levels did not seem to change significantly during water year 1963, a prerequisite for a steady-state analysis. Second, the percent of valleywide runoff for water year 1963 was about average (107 percent of normal). Third, although water year 1963 was preceded by a short-term increase in ground-water pumpage, the year was sufficiently isolated from major runoff or pumping effects that the aquifer system was assumed to be in a quasi-steadystate condition—that is, sufficiently stable to begin a transient simulation.
Water years 1963—84 were chosen to calibrate the ground-water flow model under nonequilibrium or transient conditions. Stable initial conditions were ensured by beginning the transient simulation with results from the steady-state simulation of water year 1963. The first part of this period, water years 1963—69, represents conditions in the valley prior to completion of the second aqueduct (table 4). The Los Angeles Department of Water and Power (1972) showed that the valleywide system was in approximate equilibrium for water years 1935—69 and, except for brief periods of heavy pumping during the 1930's and early 1960's, probably in near-equilibrium for most of the period between the completion of the first aqueduct in 1913 and the second in 1970. Therefore, the first part of the calibration period, water years 1963—69, was assumed to be fairly analogous to the entire period prior to operation of the second aqueduct.
The second part of the calibration period, water years 1970—84, represents the significantly different conditions in the valley after completion of the second aqueduct and the related changes in water use (table 4). This second period was a time of significantly increased pumpage, a decrease in water supplied for agricultural and ranching operations, a severe drought (1976—77), and extremely wet conditions following the drought. The ability of the model to simulate such diversity of conditions within the same calibration period reflects on its appropriate design and helps to confirm that the model is a fairly complete representation of the actual aquifer system.
Water years 1985—88 were chosen to verify that the ground-water flow model was not uniquely tuned to the calibration period and could be used to evaluate non-calibration periods. The verification period, although short, is a good test of the calibrated groundwater flow model because there are significant fluctuations in runoff and pumpage. Also, new highproduction "enhancement and mitigation" wells were put into service. The verification period was simulated after calibration of the model was complete. Recharge and discharge components required for the verification period were calculated in the same way as for the calibration period. No changes were made to recharge, discharge, or other parameters in the ground-water flow model. In fact, as it turned out, all model simulations for the verification period were completed prior to obtaining and reviewing measured ground-water-level data for the period—a rather unnerving, if somewhat fortuitous sequence for verification.
A final simulation period was defined to represent "1988 steady-state conditions" —that is, the equilibrium that the aquifer system would reach if operations as of 1988 were continued well into the future. Preliminary evaluation at the beginning of the cooperative studies identified water year 1984 as a likely period that could be used to simulate average present conditions. Subsequent analysis, however, determined that the Owens Valley was in the midst of significant vegetation and hydrologic changes and that stable quasi-steady-state conditions did not exist in 1984. Therefore, a more generalized steady-state simulation was designed, taking into account longterm average runoff and new enhancement and mitigation wells that were installed after 1984. This simulation and the related assumptions and approximations are described later in this report in a section entitled "Alternative 1: Continue 1988 Operations."
Calibration
Calibration of the ground-water flow model involved a trial-and-error adjustment of model parameters representing aquifer characteristics and certain recharge and discharge components in order to obtain an acceptable match between measured ground-water levels and computed heads and between estimated and computed recharge and discharge. For example, more than 200 hydrographs displaying levels and heads were reviewed throughout the calibration process; 67 of these hydrographs for 56 model cells are shown on plate 1. Also, simulated recharge and discharge were reviewed extensively on a "cell-by-cell" basis (McDonald and Harbaugh, 1988, p. 4—15) to ensure that the magnitude and distribution of computed ground-water flows (fluxes) were appropriate. The calibration process was continued until further changes in the ground-water flow model did not significantly improve the results and until the model parameters, inflows and outflows, and heads were within the uncertainty of historical data.
The philosophy of model development and calibration was to use general relations for as many components of the model as possible. These relations, or conceptual themes, permit an improved understanding of the overall model and its more than 100,000 parameters. For example, the hydraulic characteristics of the model were based on hydrogeologic subunits (model zones), each with uniform hydraulic properties. Reductions in transmissivity caused by faults were calculated as a percentage of the transmissivity of the faulted material (pl. 2). Recharge and discharge commonly were related to a more general concept, such as the percent of average valleywide runoff. Detailed, site-specific adjustment of parameters or relations was done rarely, if at all. Because of the way it was calibrated, the model is most useful for evaluating valleywide conditions, not for predicting small-scale effects covering a few model cells. Site-specific ground-water flow models or multivariate regression models, such as developed by P.B. Williams (1978) and Hutchison (1991), can give more accurate predictions at selected sites. However, these models in turn are less useful for evaluating valleywide hydrogeologic concepts or predicting valleywide results of watermanagement decisions.
The calibration procedure first involved estimating initial values of inflow and outflow to the aquifer system for the steady-state period, water year 1963. Many of the estimates were obtained from preliminary work by Danskin (1988). Adjustments were made in some of the initial estimates in order to ensure a balance of inflow and outflow as well as to match the distribution of measured ground-water levels. An assumption in the calibration of steady-state conditions was that ground-water levels in 1963 were similar to those in 1984 for most parts of the basin (figure 14). This assumption was necessary because of the absence of virtually any ground-water-level data prior to 1974 for hydrogeologic unit 1.
The bulk of the calibration involved making adjustments to the model that are based on the transient behavior of the aquifer system during the 22-year period, water years 1963—84. To ensure stable initial conditions, the steady-state period was resimulated each time changes were made to the model. Also, the distribution of head and the pattern of ground-water flow were reevaluated for each steady-state simulation to ensure that they remained conceptually valid and similar to those shown in figure 14.
Transmissivity values were adjusted within the general range indicated by aquifer tests (figure 15 and table 9) and related studies (Hollett and others, 1991; Berenbrock and Martin, 1991). Calibrated values of transmissivity were slightly higher than initial estimates for highly transmissive volcanic deposits, especially in the area of Crater Mountain near Fish Springs (figure 15 and pl. 2).
Values of vertical conductance were constrained to approximately the same values derived from the preliminary models (figure 2) and from aquifer tests described by Hollett and others (1991). Values were adjusted until simulated heads in the upper and lower model layers matched measured ground-water levels indicated on contour maps (figure 14) and on hydrographs (plate 1). For most of the area covered by alluvial fan deposits, measured levels were not available. In these areas, values of vertical conductance were adjusted so that simulated heads in the two layers differed by less than 1 ft.
Storage coefficients were held constant at 0.1 and 0.001 for the upper and lower model layers, respectively. For the upper model layer, the storage coefficient is virtually equivalent to specific yield. Values determined from aquifer tests (table 9), as expected, were lower than model values. Aquifer tests, even those extending several days, are affected most by the compressive response of the aquifer and expansion of ground water and are affected very little by actual drainage of the aquifer materials. This drainage, which accounts for nearly all of the specific-yield value, is delayed and occurs slowly over a period of weeks, months, or years. As a result, storage coefficients obtained from model calibration of long-term conditions usually are much more indicative of actual values than are those calculated from aquifer tests. Attempts at specifying unique storage coefficients for each hydrogeologic unit proved to be tediously unproductive.
All recharge and discharge components had conceptual or semi-quantitative bounds associated with them. These bounds (which are discussed in greater detail in other sections of this report, including "Surface-Water System" and "Ground-Water Budget" ) restricted model calibration in much the same way as did measured ground-water levels (plate 1). Some recharge and discharge components (recharge from precipitation, recharge from spillgates, and underflow) were assigned constant rates on the basis of their uniform characteristics from one year to another (tables 11 and 13). All other components were varied annually on the basis of a general concept such as percent annual runoff.
Most recharge and discharge components did require some degree of adjustment, often minor, during calibration. This adjustment was needed not only to match measured conditions, but also to ensure that a consistency between different recharge and discharge components was maintained. For example, changing recharge from a narrow canal on the valley floor required re-evaluating the quantity of recharge from narrow tributary streams on alluvial fans and from broad river channels on the valley floor. The philosophy of calibration did not permit adjusting values in individual model cells in order to match historical conditions.
The location and type of model boundaries were assumed to be known and were not varied. The quantity of underflow, however, was reduced considerably from previous estimates by Danskin (1988) and the Los Angeles Department of Water and Power (1976). Recharge from canals was slightly less than original estimates. Recharge from purposeful water-spreading operations was about two-thirds of the initial estimate. Conductance of both the rive-aqueduct and the lower Owens River were increased during calibration, thereby increasing ground-water recharge to or discharge from them. The quantity of evapotranspiration was less than original estimates. Pumpage was assumed to be known and was not changed.
Land-surface datum was used in many parts of the model, particularly in defining head-dependent relations and estimating precipitation (figure 7B). Attempts at computing land-surface values from 1:250,000-scale AMS (Army Mapping Service) point data sets obtained from R.J. Blakely (U.S. Geological Survey, written commun., 1986) required fitting a surface to the point data; results were not satisfactory, especially in areas of abrupt change in slope of the land surface, such as near the Tinemaha Reservoir. Therefore, the values were interpolated by hand from 1:62,500-scale USGS topographic maps and held constant during calibration.
Results of the model calibration are displayed in figures 19 and 20, which show comparisons of measured ground-water levels and simulated heads during spring 1984 for the upper and lower model layers, respectively. This was a time when levels were higher than they had been for several years, dormant springs had resumed some discharge, and the basin was assumed to be in a nearly full condition (Hollett and others, 1991). The match between measured levels and simulated heads for both the upper and the lower model layers seems to be quite good for most parts of the basin. A notable exception is the area west of Bishop near the Tungsten Hills.
Measured water levels and simulated heads for individual wells are compared on plate 1. Although more than 200 wells were used extensively in the calibration process, only 67 wells are included on plate 1. The 67 wells were selected to represent different well fields, different model layers, and different hydrogeologic subunits (model zones). Some wells were included on plate 1 to illustrate those parts of the valley where the ability of the model to simulate actual conditions is not as good as in other locations—for example, well 278 near Bishop and well 172 near Lone Pine (plate 1).
Precise tracking of the measured and simulated hydrographs (plate 1) was not deemed necessary, and might not be desirable or correct depending on the characteristics of the well, the surrounding aquifer material, and the model cell approximating the well. Of primary importance was that the measured and simulated hydrographs be of the same general shape and trend. Shape of a hydrograph is influenced by aquifer characteristics, recharge, and discharge; trend is influenced most by change in aquifer storage. The magnitude of vertical deflection likely will be different for measured and simulated hydrographs because of spatial discretization required for the model. The ratio of vertical deflections between the two hydrographs, however, should remain similar over time. Vertical offsets might or might not be important depending on the specific well. For example, an acceptable vertical offset can result when a well is located away from the center of a model cell; this type of offset is particularly noticeable in areas of steep hydraulic gradients, such as on the alluvial fans.
During calibration of the valleywide model, the comparison between estimated and simulated recharge and discharge was as important as the comparison between measured ground-water levels and simulated heads. Recharge and discharge components that act as hydraulic buffers respond to changes in other model parameters and reflect the dynamics of the aquifer system—sometimes much better than do changes in head. The simulated recharge and discharge for the dominant fluxes in the model after calibration are shown in figure 21.
As an aid in using and extending the work presented in this report, simulated values for each component of recharge and discharge in the groundwater flow model are given in table 11. The individual values are important aids in compiling water budgets for specific parts of the valley; developing linked water budgets for the surface-water and ground-water systems; defining the relative degree of confidence to be placed in model results in different parts of the valley; identifying how to revise and improve the model; and making local water-management decisions.
In places where concepts or data were uncertain, the ground-water flow model was not calibrated forcibly to produce a match between simulated heads and measured levels. For example, in the area north of Laws, something is missing in the ground-water flow model. Simulated heads in layer 1 do not recover after 1974 as fully as do the measured levels (well 107T, plate 1). The actual recovery could be caused by any of several processes—increased underflow during the drawdown period, induced flow of water from Fish Slough or the Bishop Tuff, increased percolation of operational spreading of surface water, or changes in the operation of nearby canals. Without a valid reason to pick one process rather than another, none was altered during calibration—thus highlighting an area of uncertainty and an area where further work is necessary. This approach was a major philosophy of the modeling study and the rationale for including some of the hydrographs shown on plate 1.
Verification
Water years 1985—88 were used to verify that the calibrated ground-water flow model will duplicate measured data for a non-calibration period. The 4-year verification period included significant stress on the aquifer system because of unusually wet and dry conditions. Valleywide runoff varied from 158 to 68 percent of normal (table 7). In addition, new enhancement and mitigation wells were put into production in various locations throughout the valley (tables 9 and 11). Initial conditions for the verification were simulated heads for water year 1984 at the end of the calibration period. Recharge and discharge data were developed for the ground-water flow model in exactly the same way and using the same relations as had been done for the calibration.
A comparison of measured ground-water levels and simulated heads during the verification period is shown on plate 1. In general, the match is very good, particularly in the Laws area where the aquifer was highly stressed. The model also simulates the return of spring discharge during the period (figure 21). The close agreement between measured ground-water levels and simulated heads and between measured and simulated spring-discharge rates was achieved without any adjustment of model parameters. This ability to reasonably match data from another time period suggests that the ground-water flow model can be used to predict results from stresses that are similar in type and magnitude, but not exactly the same as those used during calibration—a prerequisite for a predictive model.
Sensitivity Analysis
Sensitivity analysis is a procedure to determine how sensitive the model solution is to a change in each model parameter, including transmissivity, vertical conductance, storage coefficients, and inflow and outflow rates. As is always the case with numerical models, not all parameters of the model were known completely. Because some uncertainty is present in each parameter, there is some uncertainty in the model solution. This uncertainty is reflected in heads and inflow and outflow rates that are somewhat in error. A sensitivity analysis identifies which parameters exert the most control over the model solution and, therefore, have the potential to generate the largest errors. An improved understanding of those parts of the aquifer system represented by the most sensitive parameters yields the greatest improvement in the ground-water flow model.
One of the sensitivity tests that was most illuminating is presented in figure 22. For the test, water years 1963—88 were resimulated with slight modifications in recharge and discharge. For the first part of the test (figure 22A), recharge from tributary streams, recharge from ungaged areas between tributary streams, and recharge from runoff from bedrock outcrops within the valley fill were held constant at 100 percent of long-term average conditions (100-percent runoff year). In the second part of the test (figure 22B), calibration values were used for everything except ground-water pumpage, which was held constant at the values for water year 1963. Effects from each test were observed at wells in recharge areas, near well fields, and away from both recharge areas and well fields. As expected, the effects in recharge areas are most dependent on recharge, and the effects near well fields are most dependent on pumpage. Away from either area, heads are relatively unaffected by changes in either recharge or pumpage, probably as a result of the many hydraulic buffers in the aquifer system. What is somewhat surprising is the degree to which both recharge areas and well fields are affected by pumpage. Clearly, pumpage plays the dominant role in affecting heads (ground-water levels) in the valley.
For the rest of the sensitivity analysis, each of the model parameters was altered by a certain amount from the calibrated values. The amount of the alteration was determined by estimates of the likely range of the data (Hollett and others, 1991, table 1) (figures 15 and 16; tables 9, 10, and 11). To simplify the analysis, similar variables, such as transmissivity on the alluvial fans, were altered together. The variables associated with the most change in the model solution were identified as the most sensitive. Similar sensitivity analyses were done using a ground-water flow model of the Bishop Basin (Radell, 1989) and a model of the Owens Lake Basin (Yen, 1985). Those analyses are presented graphically for several of the model parameters and depict results similar to those discussed here for the valleywide model.
Although useful, this method of testing sensitivity is subject to a potentially significant flaw. Because each variable in the model is tested separately, the additive effects of changes in more than one variable are not considered. For example, the simultaneous overestimation of both recharge and evapotranspiration in the model would tend to be self-correcting. However, overestimating recharge and underestimating evapotranspiration would produce a considerably different model solution. If neither recharge nor evapotranspiration by itself were a sensitive part of the model, the conclusion from a routine sensitivity analysis would be that additional refinement of these rates is unnecessary. Nevertheless, the additive effects of errors in recharge and evapotranspiration might produce significantly erroneous results in some simulations of the aquifer system.
This type of error can be prevented by means of a more subjective analysis of sensitivity during development and calibration of the ground-water flow model. The modeling technique chosen for the valleywide model took advantage of this method. Those characteristics of the aquifer system believed to be most important were analyzed first using different-scale models (figure 2). Then, the valleywide model was developed by adding sequentially greater complexity to the model—one recharge or discharge component, or one additional model zone at a time. In this way, during model development and calibration, the sensitivity of each model parameter could be identified more easily. These observations, which are as valuable as a post-calibration sensitivity analysis, also are included in the following discussion of the sensitivity of each parameter.
Transmissivity.—The areal distribution of transmissivity in the valley is based on scattered data (figure 15) and an assumption of uniformity within each model zone (pl. 2). Model errors can be associated with the values of transmissivity chosen for an individual zone and with the choice of zone boundaries. The sensitivity of the model to the locations of the zone boundaries is best evaluated by altering the locations, recalibrating the model, and observing the differences. Although this time-consuming process was not part of this investigation, the location of the transition zone was found, during model development, to be a sensitive parameter. Equally sensitive was the location and, in particular, the continuity of volcanic deposits near the Taboose—Aberdeen and the Thibaut—Sawmill well fields (figure 17).
Variations in the value of transmissivity within a model zone produced less effect on heads and groundwater discharge than was hypothesized initially. An exception to this was the area of highly transmissive volcanic materials between Big Pine and Fish Springs (pl. 2). Lower values of transmissivity produced much lower discharge from Fish Springs and unrealistically steep gradients from north to south along the edge of Crater Mountain. From a valleywide perspective, the addition of the more transmissive model zones representing transition-zone and volcanic deposits produced a much greater effect on heads than did variations of transmissivity within individual zones.
Vertical conductance.—Calibrated values of vertical conductance (the model equivalent of vertical hydraulic conductivity) were based on sparse field data and model calibration. To test a wide range of possible values, vertical conductance in each hydrogeologic area was varied by two orders of magnitude. However, the effect on heads was not as pronounced as was expected. In fact, the model seemed to be rather insensitive to changes in vertical conductance (Radell, 1989, figure 6.4). Part of the reason for this may be the relatively large size of the model cells and use of an annual approximation of recharge and discharge. Both of these model characteristics, which require averaging simulated recharge and discharge over space or time, result in less change in simulated ground-water levels for a given recharge or discharge than would occur in the actual aquifer system. A greater sensitivity in vertical conductance might be expected in an analysis using smaller distances and shorter timeframes, similar to those used to analyze an aquifer test. During calibration, the value of vertical conductance was noted as being closely tied to the rate of evapotranspiration, which tends to dampen changes in heads near the valley floor. Lower values of vertical conductance result in less flow from the lower model layer to the upper, which in turn results in less water available for evapotranspiration. This spatial correlation between vertical conductance and evapotranspiration can be seen by comparing the vertical difference in head (figures 19 and 20) with evapotranspiration rates (pl. 3A)
Storage coefficient.—Storage coefficient was determined to be one of the least sensitive variables. This result corresponds to similar findings by Yen (1985, p. 150). Sensitivity analysis showed that storage coefficients higher than the calibrated values did not change heads significantly, but values less than about 0.0001 for the lower model layer (hydrogeologic unit 3) produced unrealistic variations in heads at many locations in the basin.
Precipitation.—Precipitation records for the Owens Valley, in general, are very good, except for an absence of precipitation stations on the east side of the valley (figure 7A). Nearly all precipitation falling on the valley floor is assumed to be used by native vegetation, and recent monitoring of the unsaturated zone tends to confirm this assumption (Groeneveld and others, 1986a; Sorenson and others, 1991). Therefore, the effect of recharge from precipitation falling on the valley floor was not tested in the sensitivity analysis.
In contrast, recharge from precipitation is assumed to occur along the mountain fronts, but the quantity is completely unknown. The present assumption is that about 95 percent of precipitation is evapotranspired, and 5 percent, or about 2,000 acre-ft/yr, is recharged (table 10). Variations of 3 to 4 times this value produced minor effects on model simulations, primarily increasing evapotranspiration from the valley floor and gains of water by the rive-aqueduct system. Similar results were found by Radell (1989, figure 6.10). If the present assumption is largely incorrect, then recharge from precipitation could be a sensitive model parameter with respect to ground-water flow rates as found by Danskin (1988). However, a large increase in recharge from precipitation probably would require a similar decrease in mountain-front recharge between tributary streams (tables 10 and 11) in order to maintain a calibrated model.
Tributary stream recharge.—Measurements of tributary stream discharge are among the most complete and most accurate hydrologic measurements in the valley. Because most tributary streams are measured at both a base-of-mountains gage and a rive-aqueduct gage (figure 11), estimates of tributary stream recharge do not vary greatly. An increase of 10 to 20 percent in tributary stream recharge for streams in the Owens Lake Basin resulted in moderate to significant changes—generally, higher heads on the fans and a greater gain of water by the rive-aqueduct system. Heads and evapotranspiration rates on the valley floor showed much less effect. In the Bishop Basin, particularly near Big Pine, accounting for each stream is more difficult, and the uncertainty in recharge estimates is greater than in the Owens Lake Basin. Variations of as much as 50 percent in tributary stream recharge near Big Pine and Taboose Creeks resulted in a minimal change in heads in this area of high transmissivities, but an important change in the discharge of nearby springs (figure 17).
Mountain-front recharge.—Mountain-front recharge between tributary streams is a large, poorly quantified component of the ground-water budget (table 10). Sensitivity analysis of this item included variations of a 50-percent increase or decrease and resulted in significantly different heads and groundwater fluxes along the west side of the basin. Results are similar to a 15-percent error in recharge from all tributary streams. The lack of measured data suggests that errors in estimating mountain-front recharge are more likely than for most other components of the ground-water flow model. This large degree of uncertainty makes the high sensitivity of this component even more important. During calibration of the Bishop area, an inverse correlation was observed between the quantity of mountain-front recharge and the quantity of recharge from canals and ditches; an increase in recharge for one component probably requires a decrease in recharge for the other.
Evapotranspiration.—Evapotranspiration data are sparse, even in the most intensively studied parts of the valley (figure 2). Correlations of selected evapotranspiration data with extensive mapping of vegetation has permitted a far more detailed examination of evapotranspiration than was possible a few years ago. Even so, valleywide evapotranspiration remains a largely unquantified, highly variable component of the groundwater flow model. Given this uncertainty, variations of as much as 25 percent were investigated during the sensitivity analysis. Not surprisingly, these variations produced the greatest overall variations in heads, inflows, and outflows of any parameter in the groundwater flow model. This effect results primarily from the large role that evapotranspiration plays in the groundwater budget and from its broad areal distribution. Changes in evapotranspiration rates were most evident in the simulated gain of water by the rive-aqueduct system and the lower Owens River.
Variations in the maximum evapotranspiration rate for the head-dependent evapotranspiration relation (McDonald and Harbaugh, 1988, p. 10—1) produced most of the change in the model. Variations in the depth below land surface at which evapotranspiration was assumed to be zero did not significantly affect the model solution—except that the solution became numerically less stable for depths less than 10 ft.
Underflow.—The quantity of underflow is relatively small in comparison with that of other components of the ground-water budget, but unlike many components, underflow in the model is concentrated in areas of limited extent. Variations in the quantity of underflow from Round Valley (figure 14) significantly affected heads in that part of the basin. Variations in the quantity of underflow from the Chalfant Valley resulted in slightly different quantities of evapotranspiration near Bishop and some gain or loss of water by the Owens River near Laws. Variations in the quantity of underflow along the Volcanic Tableland made little difference in either nearby heads or gains by the Owens River.
Variations in the quantity of underflow south to the Owens Lake area produced a significant change in heads west of the Alabama Hills and relatively little change in heads east of the Alabama Hills. Much of the potential change in heads east of the Alabama Hills was dampened by changes in gains to the lower Owens River. Values of underflow near Bishop and Big Pine Creeks and near the Waucoba Canyon were locally less important and were not varied as part of the sensitivity analysis.
As was typical of much of the sensitivity analysis, changes in the quantity of underflow were not as evident in heads as in the distribution and quantity of other inflow and outflow components. The hydraulic buffering of heads by evapotranspiration, springs, and surface-water features was repeatedly demonstrated in the sensitivity testing. An analysis of sensitivity of the valleywide model, or similar models (Yen, 1985; Hutchison, 1988; Los Angeles Department of Water and Power, 1988; Radell, 1989), with respect only to changes in head would miss much of the response of the model.
Pumped and flowing wells.—Discharge from pumped and flowing wells was assumed to be known and was not varied as a part of the sensitivity analysis. The effect of withdrawing water from different model layers, however, was investigated. Initially during model development, all water was withdrawn from the lower model layer, and the model matched measured ground-water levels surprisingly well. Subsequently, discharge for each well was split between the upper and lower model layers on the basis of the length of perforations and the estimated hydraulic conductivity of adjacent aquifer materials. The match with measured data did not improve significantly. This is a curious result for a topic that has been thought to be critical in isolating the water table and native vegetation from the effects of pumping. The case of withdrawing all pumpage from the upper model layer was deemed physically impossible and was not simulated.
The causes of the lack of model sensitivity to the vertical distribution of pumpage may be the same as those suggested for the lack of sensitivity to changes in vertical conductance—that is, model cells are large in comparison with individual wells and the simulation period is long. A preliminary simulation model of the Independence fast-drawdown site (figure 2; tables 1 and 2) used model cells as small as 10 ft on a side and simulated a time period of a few weeks. Results indicated that the smaller model was highly sensitive to changes in the pumpage distribution between layers. Similar results have been suggested by the Inyo County Water Department (W.R. Hutchison, oral commun., 1989).
The lack of sensitivity also may result from the proximity of many production wells to the edge of the confining unit (compare figures 14 and 17). Over a longer timeframe, the pumping influence reaches the vertically transmissive alluvial fans and is transmitted vertically to both model layers. The confining clay layers are effectively short-circuited because of the geometry of the aquifer and the location of the production wells.
Surface water.—The head-dependent method of simulating the interaction of the aquifer system with the Owens River, the Los Angeles Aqueduct, and the Tinemaha Reservoir allows for adjustments in the prescribed stream stage, altitude of the bottom of the streambed, and conductance of the streambed. Stream stage and altitude of the bottom of the streambed were assumed to be known and were not varied. Variations in streambed conductance identified this parameter as important and narrowly defined. Increasing or decreasing streambed conductance resulted in significantly different gains to or losses from the aquifer system. This response implies that the head-dependent surfacewater features exert a strong control on the simulated aquifer system, but do not act as constant heads (McDonald and Harbaugh, 1988, p. 3—16; Franke and others, 1987; S.A. Leake, U.S. Geological Survey, oral commun., 1989).
Springs.—Springs are simulated in the model using the drain package (table 13). Spring discharge is controlled mostly by a conductance term representing the transmissive properties of the spring conduit, such as fractured lava or lava tubes, and by nearby recharge or discharge. A decrease in the conductance of individual springs produced remarkable, although somewhat localized, results. Much of this sensitivity results from the high natural discharges for several springs (figure 21). In contrast, increases in the conductance of individual springs produced much less effect. These results indicate that the transmissive properties of the spring conduits are much greater than those of the surrounding aquifer materials.
Use, Limitations, and Future Revisions
The valleywide ground-water flow model is best used to help answer questions of regional water use, ground-water flow, and surface-water/ground-water interaction. The conceptualization of the aquifer system described by Hollett and others (1991) provided the basis for a consistent, logical model for nearly the entire basin. This translation from qualitative concepts to quantitative testing was a major purpose for constructing the valleywide model and remains an important use of the model. Additional or alternative concepts of the aquifer system can be tested using the model as presently constructed or using the model as a skeleton for a somewhat different model. If changes to the present model are significant—for example, change in number of model zones, in transmissivities, or in areal extent—then recalibration will be required.
The philosophy and methodology of developing the valleywide model indicate its strengths and possible uses. The modeling technique used in this study was the development of successively more complex models to simulate the aquifer system. The initial model resembled that documented by Danskin (1988). Subsequent site-specific models (figure 2) were developed to investigate specific questions about the aquifer system (table 2), and information gained from these smaller models was incorporated in the design of the valleywide model. Final refinements in the valleywide model were critiqued in concert with ongoing modeling studies by Inyo County and the Los Angeles Department of Water and Power. In this way, important information was obtained at several different scales and from several different viewpoints. As a result, the valleywide model reflects this technical and numerical consensus. During the cooperative studies, the model played an important role as a neutral, technical arbitrator in answering complex and often volatile water-use questions. Future beneficial use of the model may be in a similar way.
Valuable information gained from design, development, calibration, and sensitivity analysis of the ground-water flow model is not complete. Additional information and insight certainly can be obtained without any new model simulations simply by additional review of model data and results presented in this report. Additional sensitivity analysis may be helpful in identifying which new data are most beneficial in answering water-management questions. Although regional by design, the valleywide model does include many small-scale features and site-specific data and concepts. Future analysis of these smaller-scale features or issues—such as a volcanic deposit, a facies change, or a question of local water use—might best be done by use of smaller-scale models or field studies, in combination with simulations from the valleywide model.
The most appropriate use of the valleywide model is best illustrated by the results presented in this report. The goal in designing both water-management alternatives and figures was to maintain the "regional" character of the model, focusing on larger issues, over longer periods of time. Results are presented precisely (table 11) in order that they can be duplicated and extended; however, use of model results needs to be more schematic—for example, more change occurs in this part of the basin, less in that part. The specific value of drawdown at a well (plate 1) or for an area of the basin (figure 23) is far less important than the relative value (more drawdown or less drawdown) in comparison with other areas of the basin. Use of the model in this way will maximize its utility and minimize the limitations.
The primary limitation of the valleywide groundwater flow model is that it is regional in nature. Interpreting results at a scale of less than about 1 mi2 is inappropriate. The model also is "regional" with respect to the time scale that was chosen for calibration. Interpreting results at a scale of less than a single year is inappropriate. Many limitations of the valleywide model are common to all numerical models and are described by Remson and others (1971), Durbin (1978), Wang and Anderson (1982), Franke and others (1987), and McDonald and Harbaugh (1988). Despite these general limitations of modeling and the specific limitations of the valleywide model of the Owens Valley, as described below, no other methodology provides such a complete testing of ground-water concepts and data.
Interpretation of model results in selected areas of the basin requires special caution. In particular, the area west of Bishop and the area near Lone Pine are simulated poorly. The area west of Bishop has a combination of faults, buried Bishop Tuff, terrace gravel deposits, and abundant recharge. The measured levels and simulated heads (figures 19 and 20; plate 1) do not match well, indicating that the model does not represent actual conditions. It is not clear at this point whether a more detailed simulation of the complex geometry of the Bishop Basin described by Hollett and others (1991) is needed, or if refinement of present hydrogeologic concepts is necessary.
The area around Lone Pine also is simulated poorly. Any number of changes in the model—in the location or hydraulic properties of nearby en echelon faults, in underflow rates, or in recharge from Lone Pine Creek—did little to improve the match for wells in the immediate area, such as well 172 (plate 1). A basic problem may be that the wells are in small, isolated compartments created by the en echelon faulting. This same phenomenon probably is present north of the Alabama Hills near well 363T (plate 1). These wells do not interact with the rest of the aquifer system in a way readily approximated by this model. The complex hydrogeology of the areas requires extensive data collection in order to provide the concepts, spatial definition, and parameters necessary to design and calibrate a more accurate numerical model. An alternative method for predicting local ground-water-level changes is to use a simple regression model that avoids many of the spatial and conceptual issues. However, as noted by Hodgson (1978), use of a regression model does not obviate the need for a more rigorous groundwater flow model, at least at a regional scale.
In some parts of the valley, critical hydrologic features are located within a few thousand feet of each other. In the Independence area, for example, the aqueduct, pumped wells, changes in transmissivity and vertical conductance, and changes in vegetation from dryland sagebrush to valley-floor phreatophytes (xerophytes) all are present within about 3,000 ft of each other. Abrupt changes, such as these, result in differences between measured ground-water levels and simulated heads (figures 19 and 20). From a regional perspective, the differences are acceptable; however, an evaluation of specific local conditions may require a better match.
In the area north of Laws, measured groundwater levels in the immediate vicinity of the boundary of the aquifer system (wells 107T and 252, pl. 1) recover more rapidly than do heads predicted by the model. Although noted, this discrepancy does not affect model simulations or the related results significantly. Simulation of the western alluvial fans and the area east of the Owens River produced reasonable results that seem to validate the basic hydrogeologic concepts about each area; however, an absence of measured data in each area suggests that results in these areas should be interpreted cautiously.
Some of the chosen methods for approximating the aquifer system may produce undesirable effects in some parts of the basin under some conditions. The choice of simulating a constant saturated thickness for hydrogeologic unit 1 may lead to differences in drawdown near sites of significant recharge or pumpage when compared with simulated results that account for changes in saturated thickness. Simulation of canals and ditches only as sources of recharge underestimate their capacity to drain the aquifer system during extended periods of high runoff. The simulation of underflow as a specified, constant rate limits the accuracy of the model for predicting effects of recharge or discharge near a flow boundary, such as north of Laws.
The valleywide model, which simulates the saturated aquifer system, does not incorporate the complex process of vegetative growth and water use as explicit variables, nor does the model simulate the unsaturated soil-moisture zone. Vertical onedimensional models with these capabilities were developed for selected areas of the valley (table 1 and figure 2) as a related part of the comprehensive studies of the Owens Valley (Welch, 1988). Incorporating these features in a valleywide model would make it numerically far too large to be useful. The ground-water flow model, however, does simulate changes in the water table and extraction of water from hydrogeologic unit 1 by various processes, including evapotranspiration. With these capabilities, the model can be used to predict areas of the valley where hydrologic stress, such as a decline in the water table or a decrease in ground-water flow rates or discharge, probably will occur.
A key assumption in using the saturated groundwater flow model to evaluate likely effects on native vegetation is that areas of significant hydrologic stress correspond to areas of vegetative stress. In related studies, researchers found that a significant decline in the water table corresponded to a significant stress on native vegetation, particularly rubber rabbitbrush (Chrysothamnus nauseosus) (Dileanis and Groeneveld, 1989; Sorenson and others, 1991). Other factors, including alkalinity and salinity (table 3), are acknowledged to play an important role in the health of native plant communities (figure 6). Therefore, results from the ground-water flow model should be viewed in general terms as areas of the valley where stress on native vegetation is likely.
A simplification of how the ground-water flow model simulates water use by plants may contribute to an underestimation of water-table recovery during wet periods immediately following dry conditions. During a drought, plants drop leaves in order to limit transpiration and loss of water. During the year following a drought, use of water by plants is restricted (because number of leaves is fewer) until more leaves can be grown. If abundant precipitation falls during this time when the plants have fewer leaves, then the precipitation may satisfy the bulk of the water needs of the plants. Relatively little ground water will be transpired even though ground-water levels are rising because of increased recharge. The ground-water flow model assumes that higher ground-water levels always result in higher evapotranspiration from the ground-water system. This feature may overestimate evapotranspiration during some wet years, and may not allow the simulated water table to recover as rapidly as measured data indicate.
During development of the valleywide model, the simulation of evapotranspiration by native vegetation was studied extensively. Several different approaches were tested, including use of a piecewise-linear, head-dependent relation with a fixed maximum evapotranspiration rate, as described for the final calibrated model; the same relation with a spatially varying maximum evapotranspiration rate based on mapped native vegetation; an evapotranspiration rate based on a separate soil-moisture-box accounting; and an evapotranspiration rate related to precipitation. Each method had its own advantages and disadvantages but yielded surprisingly similar results. This unanticipated conclusion probably stems from the annual approximation of recharge and discharge, the long simulation period, and the regional character of the model. In order to better simulate some transient conditions, future revisions of the valleywide model may consider use of a more complex evapotranspiration package with spatially varying parameters to simulate direct precipitation on the valley floor, antecedent soil moisture, and vegetative growth and water use.
Spatial and temporal discretization of the valleywide model generally does not adversely affect the simulation of regional or subregional watermanagement issues. The two-layer approximation of the aquifer system produced good results in nearly all areas of the valley. However, a three- or four-layer approximation of the Big Pine and the Taboose—Aberdeen areas, paralleling the conceptualization documented by Hollett and others (1991), would yield a more physically based and possibly more reliable model. Addition of more layers to the model allows a better spatial representation of the complex geometry between pumped volcanic deposits and nearby fluvial and lacustrine deposits, and might result in a more accurate simulation of pumping effects on different parts of the aquifer system. The approximation of numerous individual clay layers by a single confining layer, such as for the fluvial and lacustrine deposits (figures 4 and 5), yielded good results and does not need to be changed in future revisions of the valleywide model. The present approximation of the massive bluegreen clay near Big Pine with a simple Darcian relation is likely to result in inaccurate results for some simulations that are sensitive to the transient propagation of hydraulic head through the thick clay and the concurrent release of ground water from storage in the clay.
The use of model zones to group areas with similar geologic materials (hydrogeologic subunits) was a simple technique that produced good results. Identifying transition-zone deposits as a unique hydrogeologic unit (figure 5) and incorporating the unit as a separate model zone, as suggested by Danskin (1988), substantially improved simulation along the toes of the western alluvial fans. Additional drilling east of the Owens River would help to confirm the presence and configuration of hydrogeologic subunits and related model zones in that area (pl. 2). A more detailed definition of the hydrogeology of the area west of Bishop is needed and might prompt a redefinition of model zones in that area.
One method of solving some limitations of the valleywide model is to decrease the size of the model grid. A finer grid-spacing facilitates a more gradual change in hydraulic parameters, which produces a better simulation of the aquifer system. Microcomputer capabilities as of 1988 permit design of a valleywide model with three or possibly four layers using a uniform grid size of 1,000 ft on a side. Use of finite-element techniques facilitates increased spatial resolution in key areas (Danskin, 1988). However, prior to redesigning the present model, certain questions about hydrogeologic concepts need to be answered or the increased numerical resolution will not be accompanied by a commensurate increase in reliability. These questions are itemized in a later section entitled "Need for Further Studies."
Another method of improving the predictive capability of the valleywide model in selected areas of the basin is to use smaller, more detailed models, such as those developed by Inyo County and the Los Angeles Department of Water and Power (table 2). An important caveat in the use of this type of model became apparent during the cooperative studies when a detailed model of the Thibaut—Sawmill area was developed by Inyo County (Hutchison and Radell, 1988a, b). Although the boundary conditions of the smaller model were chosen carefully, the model could not be successfully calibrated. Inspection of the valleywide model revealed that the boundaries of the smaller model, although reasonable under steady-state conditions, were too dynamic under transient conditions to be simulated using the standard modeling techniques described by McDonald and Harbaugh (1988). Only transient specified-flux boundary conditions obtained from the valleywide model were sufficient to achieve a reliable transient simulation. Thus, use of more detailed models may offer advantages, particularly near well fields or spatially complex areas, but the models need to incorporate boundary conditions from a valleywide model.
Both the spatial distribution and method of simulating stream recharge worked well. Although ground-water-level data are sparse for the upper slopes of alluvial fans, the general distribution of recharge along individual streams produced reasonably good results in areas of known levels (figures 19 and 20; plate 1). Because of the considerable distance between land surface on the alluvial fans and the underlying water table, a noticeable lag may occur between a measured loss of water in a stream and the resulting response of the aquifer system (well 1T, plate 1). Although recognized, this lag did not affect simulation results significantly. Future revisions that use stress periods of 6 months or less may need to account for this time lag.
The addition of spring discharge to the model, in comparison with previous modeling efforts by Danskin (1988), produced major improvements in simulating areas along the toes of alluvial fans and edges of volcanic deposits. These areas also are characterized by a relative abundance of water and native vegetation (figure 3), which might indicate that evapotranspiration rates are higher than in most other parts of the valley. Simulation of these areas might be improved further by locally increasing the maximum evapotranspiration rate.
Future modeling also might benefit from a more detailed simulation of the interaction between the major surface-water bodies and the aquifer system. A variety of physically based relations are available that incorporate the wetted surface area of the interface, the hydraulic conductivity of intervening materials, and temporal variability in the hydraulic head of the surface-water body (Durbin and others, 1978; Yates, 1985; Prudic, 1989). Use of an explicit surface-water model linked to the ground-water flow model would allow more detailed mass balancing of the surfacewater system than was possible in this study and would facilitate the development of integrated surfacewater/ground-water budgets as suggested by Danskin (1988).
Discussion of Simulated Results, Water Years 1963—88
Calibration and verification of the ground-water flow model for water years 1963—88 enabled both a critique of model performance and an analysis of a critical period of basin operation—in particular, the conditions before and after the second aqueduct was put into operation. Because measured ground-water levels for hydrogeologic unit 1 (upper model layer) were collected at only a few sites prior to 1974, a quantitative analysis of the period requires the use of simulated results.
The simulated change in water-table altitude between water years 1963 and 1984, both times of a relatively "full basin," is shown in figure 23. Simulated conditions for water year 1963 generally reflect average conditions prior to 1970 (table 4). In some parts of the valley, antecedent pumping seems to have affected measured ground-water levels (plate 1). Because this antecedent pumpage is not included in the model, simulated heads for water year 1963 may be slightlyhigher than measured levels in those areas. Simulated conditions for water year 1984 also reflect a nearly full basin, but one after the substantive changes in basin management that occurred in 1970.
Major changes in the simulated water table between water years 1963 and 1984 are obvious in the Laws and the Big Pine areas (figure 23), and are visible in measured levels (plate 1). Equally major changes also are suggested beneath western alluvial fans, particularly near the Taboose—Aberdeen well field (figure 17). Because no measured levels are available in the fan areas, this simulated result is less certain. However, the result is consistent with the large increase in pumpage from the Taboose—Aberdeen and the Thibaut—Sawmill well fields (figure 17), the decrease in discharge from nearby springs (figure 21), and the reasonable simulation by the model of other conditions during water years 1963—88.
The relatively wet conditions in 1984 are reflected by the blue areas in figure 23, indicating a rise in the simulated water table. It is important to note that many areas of the valley floor had a rise in the simulated water table between water years 1963 and 1984—even though elsewhere in the valley, the simulated water table declined. This duality of response is typical of the complexity observed in the valleywide system.
One of the primary questions at the beginning of the study was, "What effect does pumping have on ground-water levels and native vegetation in the middle of the valley?" The ground-water flow model was used to investigate this question for the Independence area, an area of intensive monitoring and modeling during the USGS studies (figure 2 and table 1). Shown in figure 24 are simulation results from the valleywide model for water years 1963—88 at the Independence fast-drawdown site (site K, figure 2; table 1). Values of ground-water-flow vectors for two periods, water years 1963—69 and water years 1970—84, are shown in figure 24A.
The principal components of the vectors show that the dominant ground-water flow direction is horizontal and generally eastward, although there is a significant southward component in hydrogeologic unit 3. These results are comparable to those depicted in figures 14, 19, and 20. As is typical of a layered aquifer, vertical flow rates are significantly less than the total horizontal flow rate in either unit. The difference in flow rates between the two periods is most evident as a decrease in the vertical flow rate, decrease in the evapotranspiration rate, and increase in the southward flow rate in hydrogeologic unit 3.
It is important to note that the vertical flow rate, and the related decrease in vertical flow rate, is a larger percentage of flow in hydrogeologic unit 1 than it is in hydrogeologic unit 3. Pumping may produce relatively minor effects in hydrogeologic unit 3, and at the same time, have a much greater effect on flow rates into and evapotranspiration from hydrogeologic unit 1. Native vegetation depends on the continuous flow of water into hydrogeologic unit 1 and is affected by a change in flow rates. Shown in figure 24B is the simulated change in flow rates and evapotranspiration for water years 1963—88. The effect of pumping is clearly evident, beginning in 1970, in simulated flow rates and evapotranspiration at the Independence fast-drawdown site.
The importance of maintaining an adequate ground-water flow rate into hydrogeologic unit 1 also is illustrated in figure 25, which shows a schematic east—west section in the same general area of Independence shown in figure 24. Two conditions are shown in the section (figure 25)—ground-water levels with and without ground-water pumping. With no pumping, ground-water levels are fairly static. Ground water recharges hydrogeologic units 1 and 3 from the western alluvial fans in proportion to the saturated thickness of each unit. With pumping, the saturated thickness of hydrogeologic unit 1 is decreased, which in turn decreases the quantity of ground water flowing into hydrogeologic unit 1.
Eventually, this decrease will reduce the rate of evapotranspiration from the middle of the valley (figure 24). This aspect of a fluctuating saturated thickness (time-variant transmissivity) was not simulated by the ground-water flow model; as a result, changes in actual ground-water flow rates into hydrogeologic unit 1 may be somewhat greater than those shown in figure 24.
In summary, the aquifer system, particularly the discharge components, changed significantly with the increase in pumping and export of ground water after 1970. Although changes in water use and distribution of surface water also were made in 1970, most of the changes in the aquifer system resulted primarily from increased ground-water pumpage. The increased efforts at ground-water recharge after 1970 did not compensate for the increased pumpage (table 10).
Questions about Owens Valley Hydrogeology? Please contact Wes Danskin (email or address). 619.225.6132
California Water Science Center Home Page
