Synoptic Overview

The modern Lucerne Valley is a closed, desert basin that is flanked to the south by the San Bernardino Mountains and ringed to the north by highlands of the Mojave Desert, including the Granite Mountains, Ord Mountains, and Cougar Buttes. The basin—a bolson—has a central playa fed by alluvial fans that emanate from surrounding bedrock mountain ranges and inselbergs. Intervening between the principal fans are broad piedmont slopes covered by veneers of Pleistocene and Holocene slope wash and alluvium. In many cases, the piedmont slopes are pediments that have been beveled across tilted Pliocene and early Pleistocene sedimentary strata and underlying igneous and metamorphic rocks.

The Lucerne Valley basin is part of a regionally extensive trough that developed during the Pliocene and Quaternary coincident with uplift of the San Bernardino Mountains. Basement rocks capped by saprolite that formed beneath a widespread, deeply weathered erosion surface and overlain by late Miocene basalt and older strata were all warped downward during the early Pliocene between the rising San Bernardino Mountains to the south and highlands in the Mojave Desert to the north (Sadler, 1982e; figure 12 in Meisling and Weldon, 1989; Powell and Matti, 2000). The developing trough received sediment from both these upland sources. Uplift of the San Bernardino Mountains was accentuated during the late Pliocene and Quaternary in concert with the onset of northward-vergent thrusting within the North Frontal Fault Zone along the San Bernardino Mountains front (figs. 1, 4; for fuller documentation of thrust faults, see Sadler, 1982e, fig. 1; Bortugno, 1986; Miller, 1987, 2004; Meisling and Weldon, 1989; Powell and Matti, 2000; Spotila and Sieh, 2000; Spotila and Anderson, 2004; equivalent to North Frontal thrust sytem shown in U.S. Geological Survey, 2020).

Throughout its history, the Lucerne Valley basin-fill, like other basins north of the San Bernardino Mountains (fig. 1; Cox and others, 2003), has been characterized by three interfingering sedimentary facies: (1) alluvial deposits derived from the Mojave Desert highlands to the north, (2) alluvial deposits derived from the San Bernardino Mountains to the south, and (3) intervening basin-floor lacustrine and playa deposits that interfinger northward with the Mojave Desert-derived alluvial deposits and southward with the alluvial deposits derived from the San Bernardino Mountains. The geologic sections (fig. 5) indicate that these three sedimentary facies are arranged in a diachronous sequence of Mojave Desert-derived strata interfingering with and overlain by basin-floor strata to the south, in turn interfingering with and overlain by strata derived from the San Bernardino Mountains farther south. This stratigraphic sequence indicates that, generally, the depocenter of the Lucerne Valley closed basin was migrating northward with continuing uplift of the San Bernardino Mountains. There are, however, significant retrogressive variations in the generally northward-migrating position of facies transitions that are likely due to fluctuations in climate and tectonic activity.

The Lucerne Valley basin has evolved in conjunction with uplift of the San Bernardino Mountains to the south and displacement on northwest-trending right-lateral faults of the of the Mojave Desert (Spotila and Anderson, 2004), in particular the Helendale section of the Helendale-South Lockhart fault zone (U.S. Geological Survey, 2020; herein called the Helendale Fault), the Old Woman Springs Fault, and the Cougar Buttes Fault (figs. 1, 4) that form part of the Eastern California Shear Zone (Dokka and Travis, 1990a, b). Where the regional trough that flanks the San Bernardino Mountains to the north is transected by these faults of the Mojave Desert province, that trough has been segmented into discrete basins like Lucerne Valley.

Movement on the northeast-trending faults was accompanied by the growth of dome-shaped and anticlinal folds on the San Bernardino Mountains piedmont (figs. 1, 4; Eppes and others, 1998, 2002; Powell and Matti, 2000; Pearce and others, 2004). These folds warped the desert floor and the sediments beneath it in the latest Pliocene and Quaternary. In contrast to the north vergence of thrusting along the range-front, the asymmetry of the anticlinal folds on the piedmont indicates south-vergent strain.

Within this evolving tectonic setting, local and regional climatic fluctuations have contributed to an evolving landscape underlain by a wide variety of Quaternary sedimentary deposits (Powell and Matti, 1998a, b, 2000). The stratigraphic and structural frameworks outlined here are further developed in the “Geologic Units” and “Geologic Structure” sections that follow.

Geologic Units

The geologic map of the study area (fig. 4) shows the geologic units that frame and define the aquifer system in the Lucerne Valley groundwater basin. These units fall into three major categories that have been generalized from the work of previous studies (Powell and Matti, 2000; Miller and Matti, 2001; Miller and others, 2001; Miller, 2004): (1) crystalline basement rocks exposed around the basin and encountered in wells that penetrated through overlying basinal strata, (2) late Cenozoic basinal strata (basin-fill deposits; fig. 5), and (3) thin to very thin surficial deposits that unconformably overlie the basement and basin-fill overlying the basement rocks.

Geologic Structure

Within the broader context of the regional tectonic framework presented in the “Synoptic Overview” section, the discussion of geologic structures here focuses on faults and folds that deform the late Cenozoic sedimentary fill of the Lucerne Valley as shown on the geologic map on figure 4.

Natural Recharge

Natural recharge to the Lucerne Valley occurs primarily by the infiltration of sporadic runoff from ephemeral washes at the base of the San Bernardino Mountains to the south, mostly during the winter months (Western Regional Climate Center, 2017a). Some surface water could enter the basin from washes emanating from the lesser mountains surrounding the valley during the occasional brief summer monsoonal storms, but most of the runoff into the valley occurs from the south through Cushenbury Creek and the small wash along the west side of the Blackhawk landslide (fig. 4). Measured streamflow records from the streamgage station at Cushenbury Creek (USGS streamgage 10260400; fig. 1) are available only for when the streamgage was in operation between 1957 and 1971 but show an annual mean discharge during that time of 0.03 cubic foot per second (ft³/s), which mostly occurred during November–May, in response to storms in the San Bernardino Mountains. Based on these existing records at Cushenbury Creek, any streamflow in the other minor washes to the south of Lucerne Valley likely is very small. The CA-DWR estimated that the average annual surface-water inflow to the Lucerne Valley was 1,050 acre-ft/yr during 1936–61, about 600 acre-ft of which was derived from the south and about 450 acre-ft from the north (California Department of Water Resources, 1967). However, geochemical data indicate that infiltration and recharge from the northern mountain sources probably do not occur under present-day climatic conditions (Izbicki, 2004).

Sources of Natural Recharge

Stable isotope ratios of oxygen (¹⁸O/¹⁶O) and hydrogen (²H/¹H) have long been used as standard hydrologic tools to provide a record of the source and evaporative history of water and can be used as a tracer of the movement of the water (International Atomic Energy Agency, 1981; Friedman and others, 1992; Izbicki and others, 1995). Precipitation that forms at cooler temperatures is isotopically lighter than precipitation that forms at warmer temperatures, which is often associated with lower altitudes, warmer climatic regimes, or lower latitudes (International Atomic Energy Agency, 1981). Lighter isotopes of oxygen and hydrogen are preferentially evaporated; water that falls as precipitation and infiltrates beyond the root zone as recharge in arid climates is isotopically heavier. The isotopic composition of recharge ceases to change below the root zone; therefore, any subsequent changes in the isotopic composition of groundwater along a flow path generally indicate only the mixing in the aquifer system or concentration by evaporation in a discharge area.

Radioactive isotopes of hydrogen (tritium; ³H) and carbon (carbon-14; ¹⁴C) are additional geochemical tools that are used to determine the movement and age (time since recharge) of groundwater. Tritium is a naturally occurring radioactive isotope of hydrogen that has a half-life of 12.43 years; it was introduced into the atmosphere in abundance when it was released as a result of nuclear testing during 1952–62 (International Atomic Energy Agency, 1981). Tritium concentration of precipitation was first demonstrated as a tracer of the movement of surface water, groundwater, and oceanic circulation during the 1950s (Begemann and Libby, 1957). Tritium concentrations are especially useful in arid environments for identifying the movement of groundwater and identifying areas where groundwater recharge occurs since about 1952 (Froehlich and Yurtsever, 1995) and have been used to estimate the movement of water through thick unsaturated zones present throughout most of the Mojave Desert (Izbicki and others, 2000). Carbon-14 present in groundwater as dissolved inorganic carbon has a half-life of about 5,730 years and has been used in hydrologic studies to determine the time since recharge of older groundwater (Vogel and Ehhart, 1963). The combination of the stable and radioactive isotopes is particularly effective in arid environments for identifying sources of recharge, determining the age (time since recharge) of groundwater, and geologic controls on the movement of water (Froehlich and Yurtsever, 1995; Izbicki and Michel, 2004).

Izbicki (2004) and Izbicki and Michel (2004) analyzed water samples from wells in the southeastern Mojave Desert collected between 1992 and 2000 to evaluate the mechanisms of recharge and identify the regional sources, movement, and age of groundwater. Their research compared the geochemistry of regional precipitation and local groundwater that included samples from about 40 wells in the Lucerne Valley. The areal distribution of stable isotope ratios of oxygen and hydrogen, and carbon-14 in the Lucerne Valley groundwater basin indicated that the groundwater was very old in areas away from the base of the San Bernardino Mountains, and recharge occurred when the climate of the western Mojave Desert was different from the present-day climate (Izbicki, 2004). Despite pumping and subsequent drawdown, which can induce the mixing of water types in some aquifers, the areal distribution of stable isotope ratios from the Lucerne Valley reflects contributions of predevelopment sources of recharge (Izbicki, 2004). The isotopic ratios are lighter than those found in present-day precipitation or water from wells sampled in adjacent groundwater basins, such as in the Upper Mojave River Valley groundwater basin (fig. 1), where recent recharge occurs from the Mojave River. The stable isotope data indicate that the groundwater in the Lucerne Valley probably was recharged as infiltration from streams draining the mountains in the Mojave Desert to the north and that the infiltration and recharge from these sources probably does not occur during present-day climatic conditions (Izbicki, 2004).

Water containing tritium, interpreted as water recharged after 1952, commonly is used as a tracer of the movement of groundwater recharged less than about 70 years before present (Izbicki and Michel, 2004). The presence of tritium can be an appropriate estimate of the initial carbon activity when carbon-14 activities are not available. Izbicki and Michel (2004) used tritium to infer carbon-14 activities in the Lucerne Valley (fig. 6A). Wells sampled in the study area yielded tritium only in a small area near Cushenbury Springs, at the base of the San Bernardino Mountains (fig. 4); measurable tritium was not detected in any other wells sampled in the Lucerne Valley (Izbicki and Michel, 2004; fig. 6A). As expected from the stable isotope and tritium data, the inorganic carbon-14 activity of groundwater was highest along the mountain front at the southern edge of Lucerne Valley, where the groundwater contained greater than 90-percent modern carbon (pmC; fig. 6B). Based on uncorrected ages, groundwater with inorganic carbon-14 activities of 90-pmC would have recharged about 370 years before present, and groundwater with 50-pmC would have recharged about 5,730 years before present. Most of the Lucerne Valley contains water with less than 30-pmC, which indicates uncorrected groundwater ages of greater than about 10,000 years before present (Izbicki and Michel, 2004).

Recharge from present-day precipitation in the Lucerne Valley groundwater basin likely is negligible, an interpretation that is supported by the absence of tritium in the groundwater in the valley and the fact that evaporation far exceeds precipitation in this arid environment. As discussed earlier, the average rainfall on the valley floor was 4.04 in. for 1919–73 (Western Regional Climate Center, 2017a), and the estimated average annual pan-evaporation rate was 111.59 in. for 1948–2005 for Mojave, California, which is also in the southwestern part of the Mojave Desert (Western Regional Climate Center, 2017c).

Natural Recharge Estimates from the Basin Characterization Model

In arid basins where surface-water data are lacking, such as the Lucerne Valley, the potential in-place recharge and potential runoff into the groundwater system can be estimated using the regional-scale California Basin Characterization Model (CA-BCM; Flint and others, 2013). The CA-BCM uses a monthly regional water-balance model to simulate hydrologic responses to climate and renders estimates of basin recharge and runoff. This model has been used in other desert basins to estimate underflow from adjacent mountains and basins and potential runoff in stream channels to downgradient groundwater basins where surface-water data are sparse or nonexistent (Rewis and others, 2006; Flint and others, 2012; Flint and others, 2013; Faunt and others, 2015). Using a suite of hydrologic components that included, but were not limited to, streamflow data from similar nearby basins, snowpack and the timing of snowmelt, temperature, types of vegetation, and precipitation in the San Bernardino Mountains, the CA-BCM simulated average annual potential recharge for 1981–2010 ranging from 0 in. on the valley floor to a maximum of about 16 in. on the north side of the San Bernardino Mountains (fig. 7A). The potential annual recharge estimated for 2000–13 was lower, and the CA-BCM simulated essentially no recharge to the valley floor and less recharge in the surrounding mountains areas overall (fig. 7B). However, the CA-BCM model domain included the watersheds that drain into the Lucerne Valley and the adjacent, but separate, Upper Mojave River Valley groundwater basin, west of the Helendale Fault (fig. 7). Therefore, the estimated potential recharge to the Lucerne Valley groundwater basin and study area (active model area for the Lucerne Valley Hydrologic Model) was calculated only from the CA-BCM cells east of the Helendale Fault. Applying the CA-BCM to this area, potential recharge to the study area for 1942–2016 (the period discussed in the “Lucerne Valley Hydrologic Model” section) was estimated to range from about 2 to 1,890 acre-ft/yr, with an average of about 635 acre-ft/yr for the study area (area of the active Lucerne Valley Hydrologic Model grid). For the same period, the potential annual recharge to the area covered by the entire Lucerne Valley groundwater basin was estimated to range from about 325 to 2,340 acre-ft, with an average of about 940 acre-ft (fig. 8), and an estimated cumulative total of about 70,600 acre-ft for 1942–2016.

Although the CA-BCM is a useful tool for estimating recharge in areas lacking surface-water data, the potential recharge simulated by the model is not necessarily equivalent to actual recharge and can be affected by many physical factors. Because water that infiltrates past the root zone does not always reach the groundwater table (Rewis and others, 2006; Flint and others, 2012), the potential differences between estimated net infiltration and actual groundwater recharge tend to increase with increases in the unsaturated-zone thickness, travel time of infiltration through the unsaturated zone, climate variability, and geologic heterogeneity in the unsaturated zone. In mountainous areas, where the unsaturated zone is likely to be more geologically heterogeneous, the potential for localized perching and lateral groundwater flow in the shallow subsurface increases. Lateral groundwater flow in the shallow subsurface (underflow) can divert a portion of net infiltration to downgradient springs or to subsurface locations within the zone of ET. Underflow is most likely in steep mountain watersheds underlain by low-permeability bedrock, where water from the higher elevation areas of the watersheds can drain into basins at lower elevations, such as to the Lucerne Valley. The net effect of underflow on a basinwide scale is a decrease in recharge balanced by an increase in ET and streamflow. Flint and others (2012) showed this effect in a study that was done near the City of Big Bear Lake (in the high elevations of the San Bernardino Mountains; fig. 1) by comparing results from an infiltration model that was modified to simulate lateral flow in the shallow groundwater system to those from the CA-BCM. In the Flint and others (2012) study, it was estimated that approximately 35 percent of the potential recharge resulted in groundwater recharge. Likewise, Rewis and others (2006) reported that mountain-front recharge estimated using an infiltration model without lateral underflow needed to be reduced to about 43 percent of the upstream recharge. Although the percentage of recharge that becomes underflow to the Lucerne Valley is unknown, the percentage of potential recharge to the groundwater system from contributing watersheds is probably even lower than percentages estimated for these studies because of the steep mountain drainage basins underlain by low-permeability rock in an arid environment (Rewis and others, 2006; Flint and others, 2012).

The CA-BCM simulated recharge on the north side of the basin (fig. 7), which contradicts the results of the geochemical studies and results presented by Izbicki (2004) and Izbicki and Michel (2004), are summarized earlier in this report section. Removing the estimates from the northern part of the basin slightly lowers the simulated potential annual recharge. Thus, the potential annual recharge estimates presented here were considered as initial values and were adjusted during the calibration process for the groundwater-flow model that was done as part of this study (see the “Lucerne Valley Hydrologic Model” section).

Anthropogenic Recharge

After the Lucerne Valley was developed in the 20th century, anthropogenic sources of artificial recharge were introduced, such as the infiltration of effluent from domestic septic systems, treated wastewater effluent, and irrigation-return water, which is infiltrated water applied to agricultural fields that is not used by plants or lost through evaporation and reaches the water table. The potential for the delivery of imported water from the California State Water Project (California Department of Water Resources, 2022) through a pipeline to spreading basins proposed south of State Route 247 near well 4N/1E-17L1 (fig. 4) was evaluated by Schaefer (1979) in the mid-1970s to alleviate groundwater-level declines, and although a pipeline was built in 1994 to convey California State Water Project water to groundwater basins to the east (not shown; Mojave Water Agency, 2022), there were no facilities or recharge operations in the Lucerne Valley at the time of this study.

The town of Lucerne Valley (fig. 2) and surrounding basin is an unincorporated rural area (U.S. Census Bureau, 2017) and therefore relies on private septic disposal systems to dispose of wastewater. Potential recharge from this water use is difficult to quantify but is assumed to be negligible. A rough estimate of the potential recharge from private domestic septic systems can be made based on the population and an assumed volume of effluent discharged to septic systems. Umari and others (1995) used an average septic-tank effluent of 70 gallons per person per day to estimate the quantity of septic wastewater that mixed with the underlying groundwater in Victorville, a desert community with similar hydrogeology located about 30 mi to the west. Using the 1928 and 2010 population estimates for the Lucerne Valley of 250 and 5,810 residents, respectively (described earlier in the “Land and Groundwater Use” section), the amount of potential recharge from septic effluent ranged from about 20 to 455 acre-ft/yr during those years. Indoor water-use rates from a minimal-producer study for 2007 and 2014 (Mojave Water Agency, 2017) yielded estimates of 0.204 and 0.178 acre-ft/yr, respectively. Mills (2009) estimated a 50-percent loss rate of septic-return flow caused by evaporation and transpiration in the Borrego Valley (not shown); considering these assumptions, the infiltration from septic systems in the Lucerne Valley could range from 0.089 to 0.102 acre-ft/yr per parcel. The actual average amount of this type of anthropogenic recharge could be less after accounting for a lower population before 1928, losses from evaporation of near-surface septic systems, and the travel time necessary for the effluent to migrate through the thick unsaturated zone to the groundwater table.

Treated wastewater effluent from the City of Big Bear Lake, which is in the high elevations of the San Bernardino Mountains south of the study area (fig. 1), has been piped down the mountain and discharged to the Lucerne Valley since 1980. This anthropogenic source of recharge from the Big Bear Area Regional Wastewater Agency (BBARWA) has been released to infiltration ponds and also has been used to irrigate alfalfa fields in the southeastern part of the valley (figs. 2I–K). Between 1980 and 2016, the amount of treated effluent transferred to the Lucerne Valley annually ranged from about 1,470 to 4,000 acre-ft (N. Flores, Big Bear Area Regional Wastewater Agency, written commun., 2017; fig. 9). Although some loss through evaporation and consumptive use by alfalfa occurs, the total volume of water from this source of artificial recharge is about 3.5 times greater than the natural recharge estimated by previous studies (California Department of Water Resources, 1967; Schaefer, 1979) or by the CA-BCM (Flint and others, 2013).

Irrigation return is groundwater that has been pumped to irrigate agricultural fields but is not consumed by the crops and subsequently returns to the groundwater system. The amount of water that is lost through plant use and evaporation and does not return to the groundwater table is considered consumptive use. The consumptive use, or net pumpage, associated with each land-use type was simulated in this study based on the percentage of irrigated land, reference ET, and crop coefficients. The methods used to estimate net pumpage, and the magnitudes of these estimates, are described in more detail in the “Lucerne Valley Hydrologic Model” section of this report. Estimates of recharge based on crop type, crop efficiencies, and irrigation requirements for 2005–16 from the calibrated groundwater-flow model are discussed later in the “Landscape-Water Use and the Farm Management Process” section.

Natural Discharge

Before groundwater development in the valley, natural discharge from the basin occurred through a few springs along the Helendale Fault (fig. 7), as evaporation at the playa surface, and by ET from the sparse desert vegetation. Evaporation likely occurred from several springs that existed along the Helendale Fault and its splays; Mendenhall (1909) observed that many areas had small and constant flow when he visited Lucerne Valley in about 1905. Because of the high predevelopment-groundwater table and areas with artesian conditions, farming was established early in the valley. The extraction of groundwater by pumping since the turn of the 20th century has lowered the elevation of the groundwater table and has caused most, if not all, of the artesian wells and springs to go dry, eliminating this type of discharge. Current surface-water evaporation is limited primarily to the moisture in the clay-rich soils of the dry lakebed and any water that ponds on the playa surface after sporadic storm runoff.

Except for stands of cottonwood (Populus spp.) trees at the springs south of Lucerne Lake and at the southern part of the Lucerne Valley that were reported by Mendenhall (1909), there are few documented observations of phreatophytes in the valley by previous researchers; it is likely that early development of the valley and groundwater extraction lowered the groundwater table beyond the root zone of phreatophytes in most places by the beginning of the 20th century. In 1917, Thompson (1929) observed some flowing wells and springs next to areas with saltgrass along the Helendale Fault, but noted that saltgrass (Distichlis spp.), which requires the groundwater table to be within a few feet of land surface, and phreatophytes were generally absent in the southern part of the valley. Later, the CA-DWR reported that about 12.5 acre-ft/yr of water was used by phreatophytes during 1936–40 (California Department of Water Resources, 1967). Schaefer (1979) reported the presence of pickleweed (Salicornia spp.), salt-loving plants that commonly grow near shores and playas along the edge of Lucerne Lake. Plants growing near the dry lakebed are tolerant of soils with high salinity and have small root systems that survive on moisture stored in the playa clays, in areas where competing plants are absent (Ellison, 1987).

The CA-DWR estimated that the average annual subsurface groundwater outflow from the Lucerne Valley to the Upper Mojave River Valley groundwater basin during 1936–61 was 100 acre-ft/yr (California Department of Water Resources, 1967). However, the Helendale Fault separates these two adjacent groundwater basins, and its effectiveness as a groundwater-flow barrier is well documented (Riley, 1956; Schaefer, 1979; Stamos and others, 2001; Teague and others, 2016). Thompson (1929) noted that the groundwater table was about 50–75 ft higher on the west side of the fault than on the east side; more recent groundwater-level data and flow patterns depicted by Schaefer (1979), Stamos and Predmore (1992), Trayler and Koczot (1995), and Dick and Kjos (2017), and also noted in the “Groundwater Levels, Flow, and Movement” section of this report, showed large differences in groundwater levels on either side of the Helendale Fault. Carbon-14 activities and tritium values indicate that groundwater on the west side of the fault flows toward the Upper Mojave River Valley groundwater basin, away from the Lucerne Valley groundwater basin (Izbicki and Michel, 2004; fig. 6). Some groundwater could be exchanged across the fault near the base of the mountain, where historical differences in the groundwater table have been smaller (Thompson, 1929), but the steep drop of the groundwater table along most of the fault trace indicates that any groundwater crossing the fault would flow from the Upper Mojave River Valley groundwater basin eastward to the Lucerne Valley groundwater basin.

Pumpage

As mentioned previously, groundwater development and the use of wells in the valley probably began in the late 1880s because Mendenhall (1909) noted that several wells had been drilled and flowing wells were reported at established ranches by 1905. Groundwater use in the basin was primarily for irrigated agriculture, with some cattle and poultry farming and a small amount for homesteads. Schaefer (1979) estimated that pumpage from agricultural sources during 1950–76 ranged from 4,000 to 13,000 acre-ft. The estimated irrigated area reported in Schaefer (1979) was about 3,500 acres in 1954 and 2,500 acres in 1976, which is a considerably larger area than what was determined from the aerial imagery discussed in the “Land and Groundwater Use” section. Pumpage for industrial use did not begin until 1960 (fig. 10). Estimates of total pumpage for 1942–2016 ranged from about 3,010 acre-ft in 1942 to about 18,300 acre-ft in 1984.

Agricultural pumpage estimates were estimated from aerial land-use photographs between 1942 and 1993 (figs. 2A–H) and estimates reported by the Mojave Water Agency for 1994–2016 (figs. 2I–K; Mojave Water Agency, 2017). The annual agricultural pumpage from 1942 to 1993 was estimated by multiplying the amount of irrigated acreage and the consumptive-use requirements for alfalfa. Consumptive use for agriculture is the unit amount of water consumed in a unit area by transpiration, building of plant tissue, and evaporation from adjacent soil. Estimates of consumptive use for alfalfa, the main crop in the Lucerne Valley since the late 19th century, vary from about 6.0 feet per year (ft/yr) in Arizona (Erie and others, 1965) to as high as 6.50 ft/yr in fields along the Colorado River between California and Arizona (Owen-Joyce and Raymond, 1996). Stamos and others (2001) assumed a consumptive-use rate of 7.0 ft/yr to estimate total pumpage for alfalfa. For this study, an assumed consumptive-use rate of 6.0 ft/yr was used to estimate total pumpage. Estimates of agricultural pumpage ranged from about 3,000 acre-ft in the early 1940s to a high of about 17,600 acre-ft in 1984 (fig. 10A). Within 10 years, total pumpage had declined by about two thirds (to about 6,800 acre-ft in 1993). This decrease in agricultural pumpage (since the mid-1980s) likely resulted from the increase in costs associated with pumping groundwater from greater depths as the groundwater table declined. Agricultural pumping continually decreased after 1993, and by 2016, the estimated annual pumpage was about 2,900 acre-ft, the lowest volume estimated during the study period.

Pumpage for domestic groundwater use was estimated by previous researchers; for this study, pumpage was estimated by using historical population estimates and results from verified pumpage studies in later years. Schaefer (1979) estimated that domestic pumpage was about 200 acre-ft in 1954 and about 1,200 acre-ft in 1976, but other domestic-use pumpage estimates were not available; therefore, domestic pumpage between 1942 and 1993 was estimated based on population estimates (Schaefer, 1979; San Bernardino County, 2007). Linear interpolation was used to estimate the population between years when population data were available. Based on values from verified production or pumpage from 1994 (Mojave Water Agency, 2017), average domestic water use was assumed to be 169 gallons per person per day; accordingly, domestic pumpage was estimated to have ranged from about 90 acre-ft in 1942 to about 670 acre-ft in 1993 (fig. 10B). From 1994 through 2016, verified pumpage was used to estimate most of the domestic groundwater use in Lucerne Valley (Mojave Water Agency, 2017). As mentioned in the “Land and Groundwater Use” section, a minimal producer study was done by the Mojave Water Agency Watermaster to estimate any additional unverified, rural, residential water usage in the Lucerne Valley for 2007 and 2014 (Mojave Water Agency, 2017); this study estimated that the average domestic water use was 0.204 and 0.178 acre-ft per year per parcel, respectively. These additional estimates for unverified rural water use were included in the estimates of domestic pumpage from 2007 through 2016 (fig. 10).

Groundwater withdrawal for industrial use mainly has been for a cement plant located at the base of the San Bernardino Mountains. Industrial pumpage estimates between 1960 and 2017 ranged from about 1 acre-ft in 2011 to about 545 acre-ft in the late 1980s (fig. 10B; Mojave Water Agency, 2017).

The total cumulative amount of groundwater removed from the basin by pumping between 1942 and 2016 was estimated to be about 700,000 acre-ft (fig. 10A), which was about 10 times greater than the cumulative amount of recharge to the entire Lucerne Valley groundwater basin (of about 70,600 acre-ft), as estimated by the CA-BCM (Flint and others, 2013; fig. 8). This difference in volume between inflow and outflow in the basin, about 630,000 acre-ft, represents the approximate cumulative total amount of groundwater that has been removed from aquifer storage by pumpage between 1942 and 2016. As a comparison, this volume approaches the capacity of Lake Havasu on the Lower Colorado River, which can store about 646,000 acre-ft (U.S. Bureau of Reclamation, 2018).

Groundwater Levels, Flow, and Movement

The installation of production wells for irrigation began sometime in the 1890s. The earliest comprehensive assessment of the region’s hydrology was conducted decades later in 1916–17 by Thompson (1929), who reported groundwater-level measurements for 43 wells and observed many places where the groundwater table was within a few feet of land surface. Thompson’s data indicated that the direction of groundwater flow in Lucerne Valley was from the outer margins of the basin toward the discharge area of the dry playa, Lucerne Lake, which is the topographically lowest part of the valley. In 1916–17, groundwater-level elevations (hydraulic heads) ranged from higher than 2,900 ft above NAVD 88 in the southern areas to less than 2,820 ft above NAVD 88 near Lucerne Lake (fig. 12A), where the depth to groundwater was between 20 and 30 ft bls. Schaefer (1979) constructed a groundwater-level contour map for the valley using Thompson’s 1916–17 groundwater-level data (contours from map reproduced herein with modification on fig. 12A). Thompson’s groundwater-level data, reviewed for this study, showed some inconsistencies between the groundwater-level data reported by Thompson and the contours drawn by Schaefer; these inconsistencies may be attributable to the effects of local pumping and inexact well locations.

In 1954, Riley (1956) inventoried and measured about 460 wells in the Lucerne Valley to provide records for as many wells as possible in anticipation of subsequent groundwater investigations and the valley’s development. Groundwater levels in the inventoried wells ranged from flowing to more than 380 ft bls, with the greatest depths to groundwater in the southern part of the basin. During his field reconnaissance, Riley observed 12 wells near the Helendale Fault that were either flowing or had hydraulic heads above land surface.

Using Riley’s data from 1956, Schaefer (1979) constructed another groundwater-level contour map for 1954 (contours from map reproduced herein with modification on fig. 12B). A comparison of the 1917 and 1954 groundwater-level contour maps (figs. 12A, 12B) indicates that the continued and increased pumping since 1917 had lowered the groundwater table in most parts of the valley, particularly in the area southeast of Lucerne Lake playa where most of the groundwater pumping continued to occur. In the area southeast of the playa, Schaefer’s maps show that the groundwater table had declined by about 30 ft. The maps also indicate that the predominant direction of regional groundwater flow—from the basin margins to the playa—had not changed since 1917, but that the area with the lowest groundwater-level elevations had expanded and migrated south toward the pumping center.

During an assessment for potential artificial recharge in the Lucerne Valley to offset groundwater-level declines, Schaefer (1979) collected groundwater-level data in 1976 and constructed a third groundwater-level contour map (contours from maps reproduced herein with modification on figs. 12C). On a basinwide scale, the groundwater-level data and contours from 1976 show the continued steepening of the horizontal hydraulic gradient and the enlargement and southward migration of the pumping depression as pumping continued to remove large quantities of groundwater from aquifer storage.

Riley (1956) brought to light the complexity of the hydrogeology of the basin by recognizing the presence of separate aquifers in the area about 2 mi south of the Lucerne Lake, where he noted the largest withdrawals had occurred and where pumping was “heavy.” Riley documented higher groundwater-level elevations in what he called the “shallow body” or upper aquifer than those measured in the “deeper body,” or middle and lower aquifers. Although Schaefer (1979) noted “strong evidence of a two-aquifer system consisting of a deep and a shallow aquifer separated by a confining layer,” he did not distinguish the groundwater-level measurements taken from the upper, middle, or lower aquifers, so the contours constructed by Schaefer (1979; figs. 12A, 12B, 12C) in the southern part of the basin represent a composite of hydraulic heads from the three water-bearing units. However, because of the poor water quality of the shallow aquifer (Schaefer, 1979), most wells were completed in the middle and lower aquifers; therefore, in this study, we assumed that the contours from Schaefer represented the middle and lower aquifers in most parts of the Lucerne Valley.

When constructing the maps for 1917, 1964, and 1976, Schaefer considered the “Lucerne Lake Fault,” which was first proposed by Riley (1956), but this fault is not supported by the geologic findings from this study (see the “Lucerne Lake Fault” and “Hydrogeologic Units and Framework” sections). As discussed earlier, Schaefer’s groundwater-level contour maps for 1917, 1954, and 1976 (figs. 12A, 12B, 12C) represent a mixture of data from the upper, middle, and lower aquifers that likely is the cause for abrupt differences in groundwater elevations measured in adjacent wells. During earlier investigations, well-depth information was available for less than one-half of the wells measured, and well-construction information, such as perforated interval, was sparsely available for existing wells; consequently, groundwater-level data were sometimes compared between wells of varying depths or reported from wells perforated in multiple water-bearing units.

Groundwater-level data were collected in the Lucerne Valley as part of 12 previous regional studies that included nearby and adjacent groundwater basins in the western Mojave Desert on the Mojave Groundwater Resources website (https://ca.water.usgs.gov/mojave/; U.S. Geological Survey, variously dated). In those previous regional studies, groundwater-level data were collected to construct 13 regional groundwater-level contour maps, biennially, for calendar years 1992–2016. The Mojave Groundwater Resources website can be used to view these historical data and contour maps; it also describes the methods used to collect the data used for this study and how the contours were drawn (U.S. Geological Survey, variously dated).

For this study, groundwater-level data for 1990–96 and 2014–16 were used to construct groundwater-level contour maps because of the insufficient amount of annual data for a single year during those periods (figs. 12D, 12E). These data and the associated contour maps indicate groundwater-level declines in the middle and lower aquifers were largest during 1990–96, when groundwater levels were more than 100 ft lower than in 1917 (figs. 12A, 12D). Where available, the data for 1990–96 and the associated contour map also indicated that groundwater levels from the upper aquifer and confining unit were at least 45 ft, and in places greater than 100 ft, higher than those in the deeper aquifers. By 2014–16, the depth to groundwater in the area south of the playa, where groundwater-level elevations were lowest in 1917, increased from between 20 and 30 ft bls to as much as 100 ft bls, and the cone of depression caused by pumping had migrated more than 3.5 mi to the southeast since development in the valley began (figs. 12A, 12E). The difference in hydraulic head between the upper and lower aquifers ranged from about 20 to 25 ft near the playa to about 140 ft in the southern part of the basin.

Long-Term Trends in Groundwater Levels

Long-term groundwater-level data from individual wells with multiple measurements provide insight into how stresses on the aquifer system affect groundwater levels over time. The historical data available in the Lucerne Valley are invaluable for documenting the long-term groundwater-level elevation trends throughout the basin. To show trends during the longest possible period, groundwater-level data were combined from wells that are near and similarly constructed so that the effects of subsurface geologic structures on groundwater movement and the consequences of long-term pumping could be evaluated. Using groundwater-level data from 1905 to 2016 (figs. 13, 14), 20 long-term hydrographs were constructed for wells located throughout the basin. Most of the hydrographs show a long-term groundwater-level decline, particularly for wells in the center of the basin where pumpage historically has been the greatest. In this area, the hydrographs show groundwater levels have declined more than 100 ft between the mid-1940s and the mid-1990s. Wells located farther from the pumping center and along the margins of the basin showed less decline during the same period. Since the early 1990s, some wells in the southeastern part of the basin had very slight declines in groundwater levels (for example, wells 4N/1E-9D4 and 4N/1E-15R1 [hydrographs 14 and 19, respectively, on figs. 13, 14]). The slight increase in groundwater levels in well 4N/1E-2Q2 (hydrograph 10 on figs. 13, 14) may be attributable to the effects of the BBARWA-treated sewage effluent that is discharged to the south; however, no data are available to evaluate the migration of this source of artificial recharge and its potential effect on groundwater levels.

Groundwater-level declines were largest in the early-to-mid 1990s, following decades of pumping rates that often exceeded 10,000 acre-ft/yr (figs. 10, 14). Since the early 1950s, groundwater-level declines had exceeded 100 ft in the middle of the basin. After the early-to-mid 1990s, the hydrographs show an upward trend in groundwater levels; these trends followed a period when pumpage sharply decreased from previous years. Pumping rates continued to decline since the mid-1990s and were comparable to those in the early 1940s and late 1950s. The lower groundwater withdrawal rates decreased the stress on the aquifer system; consequently, groundwater levels rose in the middle of the basin and groundwater levels at the margins of the basin continued to decline as the hydraulic gradient in the basin approaches a new equilibrium.

Groundwater-level data from three separate periods (1911–35, 1990–96, and 2014–16) were used to evaluate and illustrate groundwater-table decreases throughout the basin through time (fig. 15). Data were evaluated and selected to ensure groundwater-level elevations were representative of the groundwater table for these three periods. The profiles of the groundwater-level elevations shown on figure 15 were interpreted using data from wells along or near the section lines shown on figure 4. Some data could reflect composite groundwater-level elevations because wells often are completed in multiple aquifers; however, any difference in hydraulic head between the aquifers would be dwarfed by the greatly exaggerated vertical scale necessary to show the data (fig. 15).

As indicated by the data shown on figure 15, the elevation of the groundwater table from 1911 to 1935 was higher than in later years. During the early-to-mid 1990s, the groundwater levels were at historical lows; the largest declines occurred in the middle of the basin where most of the pumping was concentrated (figs. 15A, 15C). After pumping rates were reduced, groundwater levels rose slightly in the middle and lower aquifers by 2014–16, mostly in the middle of the valley (figs. 14, 15C). The decrease in the amount of pumping after the early-to-mid 1990s (fig. 10) lessened the hydraulic stress on the middle and lower aquifers and enabled hydraulic heads to recover slightly in the middle of the basin as groundwater from storage flowed from the basin margins toward the pumping depression. This migration of groundwater caused groundwater-level elevations in the center of the basin to increase since 1990–96 and resulted in the continued decline in groundwater-level elevations at the margins. As much as 25 ft of groundwater-level decline occurred near the margins of the basin since 1990–96 as the groundwater table throughout the basin gradually flattened (figs. 15A, 15C). The decline of groundwater levels along the margins of the basin (shown on figs. 13–15) is expected to continue as groundwater from aquifer storage migrates toward the pumping depression that has been created in the middle of the basin.

Several hydrographs on figure 14 show how the vertical distribution of pumpage had affected the hydraulic head in the individual aquifers and reveal the extent and hydrologic influence of the confining unit. In early field studies, groundwater-level differences between adjacent wells completed at different depths were difficult to distinguish because the hydraulic head in the aquifers probably were similar and most wells were completed in more than one aquifer. However, the long-term extraction of groundwater, mostly from the middle and lower aquifers, lowered their hydraulic heads below those that were measured in the upper aquifer. These differences in hydraulic head became more pronounced over time because the lacustrine deposits that are present below the upper aquifer (fig. 15) impeded the downward movement of water. As discussed earlier, Riley (1956) noted differences in hydraulic heads between adjacent wells south of the playa that were completed at different depths; these differences were evident in adjacent wells completed in the different aquifers and have become more pronounced over time. For example, the hydraulic head measured in 1954 in well 4N/1W-1P2 (59 ft deep and completed in the upper aquifer) was about 13 ft higher than in well 4N/1W-1P1 (450 ft deep and completed in the deeper aquifers; hydrograph 8 on figs. 13, 14). By the mid-1990s, differences in the hydraulic heads in that same area were more than 30 ft between the upper aquifer (well 4N/1W-1R7) and the middle aquifer (well 4N/1W-1R5; hydrograph 8 on fig. 13).

Spatial Discretization

The finite-difference grid, designed to represent the land surface and the subsurface alluvial deposits, consists of orthogonal square cells oriented in the north-south direction that are 2,000 by 2,000 ft (92 acres per cell). The total model grid covers an area of about 92.83 square miles (59,413 acres) that was represented by 64 rows, 50 columns, and 4 layers (12,800 cells; fig. 19A). About 18 percent of the cells (561 cells in layer 1, 528 cells in layer 2, 631 cells in layer 3, and 538 cells in layer 4) were within the active area of the hydrological model; only active model cells simulate groundwater flow, and only areas with a sufficient amount of data were included in the active model domain (fig. 19B). The grid orientation and cell size were chosen to be roughly parallel to the general direction of regional groundwater flow and align with the adjacent groundwater-flow model grid for the Upper Mojave River Valley groundwater basin created by Stamos and others (2001). The coordinates for the LVHM grid are shown in table 3.

The model layers were aligned to represent the individual hydrogeologic units of the aquifer system (figs. 5, 20). The top of the hydrologic model (layer 1) represented the elevation of the land surface. Layer 1 ranged in thickness from a minimum of about 32 ft near the southern margin of Lucerne Lake to a maximum of about 320 ft adjacent to the southernmost part of the Helendale Fault. This top layer represented the shallow alluvial-fan and lacustrine deposits (units Qvof, QTf, Tmf, and TI) along the edges of the model and the lacustrine deposits (unit Qvol) near the center (fig. 5A). Layer 2 mainly represented the lacustrine deposits (units Qvol and QTI) of the confining unit and ranged in thickness from about 1 ft in the northern part of the model to a maximum of about 172 ft in the southern part (fig. 5B). Layer 3 represented the middle aquifer and ranged in thickness from about 58 ft in the northeastern part of the basin to about 380 ft in the center (figs. 5B, 5C). Layer 4 represented Mojave Desert derived alluvial fan deposits and fan-delta deposits (unit Tmf) of the lower aquifer and ranged in thickness from about 3 ft adjacent to the Cougar Buttes Fault to about 400 ft near the Helendale Fault (fig. 5C).

Temporal Discretization

The total simulation period for the LVHM spanned 76 years, from January 1941 through December 2016. The first stress period (representing 1941) was a steady-state simulation that defined the initial condition for the transient-state simulation (1942–2016). Periods of user-specified model inflows, outflows, and boundary heads are referred to as stress periods. Variations in stresses were simulated by changing these model inputs from one stress period to the next. The temporal discretization of the model was designed to better simulate the agricultural water demands during the growing season, the dynamics of varying recharge and discharge, and potential evapotranspiration (PET) that collectively drive the major water components of the basin. The inputs that included groundwater pumping, precipitation, PET, and water applied to irrigate crops, were assumed to be constant in each stress period; stress periods were further divided into time steps to make the model run faster while still maintaining acceptable accuracy. The transient-state simulation (from 1942 to 2004) had two stress periods each year to represent a non-growing season from January 1 through March 31 and a growing season from April 1 through December 31, for a total of 126 stress periods (stress periods 2–127). Starting in 2005, finer resolution land-use and pumpage data were available, and the transient-state stress periods from 2005 to 2016 (stress periods 128–271) were discretized into 144 monthly stress periods.

Hydraulic Conductivity

Hydraulic conductivity is a function of the physical properties (porosity, skeletal structure, saturated hydraulic radius, and pore connectivity) of a porous medium and the physical properties of the fluid passing through the porous medium. A porous medium will transmit groundwater through a geologic section of unit area, measured at a right angle to the direction of flow, under a hydraulic head gradient if the medium transmits (Lohman, 1972). The OWHM2 requires horizontal and vertical hydraulic conductivity values for confined and unconfined aquifers. Initial estimates of hydraulic conductivity values (table 4) were derived from specific-capacity tests and driller’s logs; initial horizontal hydraulic conductivity values ranged from 0.001 to 100 feet per day (ft/d), and initial vertical hydraulic conductivity values were 0.0001 and 0.00001 ft/d. Calibrated horizontal hydraulic conductivity values ranged from 0.0013 to 40 ft/d, and calibrated vertical hydraulic conductivity values ranged from 0.000068 to 0.269 ft/d (table 4). These horizontal and vertical hydraulic conductivity values were assigned to model cells according to the zonation patterns shown on figure 21.

Specific Yield and Specific Storage

Water is stored and released from an aquifer by two processes: (1) the filling and draining of pores due to gravity and (2) the expansion and compression of water and the aquifer skeleton (Freeze and Cherry, 1979). Specific yield is associated with the filling and draining of pores by gravity and is a much larger contribution to the total storativity of an aquifer than specific storage. Specific yield applies only to unconfined systems and represents the drainable porosity of an aquifer as the groundwater table changes. For confined aquifer systems, water is released from storage when pumpage causes a decrease in hydraulic head. The change in pressure, from a decrease in hydraulic head, is transmitted into the confined aquifer system; water is produced by a slight compression of the aquifer skeleton and the expansion of water as an effect of the pumpage stress.

The upper aquifer, confining unit, and middle aquifer (layers 1, 2, and 3, respectively) were simulated under convertible conditions; therefore, specific yield and specific storage were specified in the Upstream Weighting Package (UPW; Niswonger and others, 2011). The lower aquifer (layer 4) was simulated under confined conditions and only specific storage was parameterized. Initial estimates of specific yield and specific storage were taken from groundwater-flow models done for the Antelope Valley (Siade and others, 2014) and Joshua Tree (Nishikawa and others, 2004) groundwater basins. Initial estimates of specific yield ranged from 0.05 to 0.125, and initial estimates of specific storage were 4.20E-07 ft^–1. Initial estimates of specific yield and specific storage for model zones are listed in table 4, and model zones are shown on fig. 21.

Storage Properties

The Subsidence and Aquifer-System Compaction Package (SUB; Hoffmann and others, 2003) was used to simulate elastic (recoverable) compaction and expansion and inelastic (permanent) compaction of compressible fine-grained beds or interbeds within the aquifer; the SUB was used to simulate the instantaneous drainage of the fine-grained interbeds in model layers 1–3. As discussed earlier in the “Land Subsidence” section, removing water from storage within the fine-grained silts and clays interbedded in the aquifer system causes these highly compressible sediments to compact, resulting in land subsidence. The deformation of the interbeds is caused by hydraulic-head or pore-pressure changes and, thus, by changes in effective stress within the interbeds. Interbeds in model layer 1 consist primarily of the playa and lake deposits visible on the land surface (fig. 4); these interbeds were assumed to compact instantaneously when dewatered. Layer 2 consists of Quaternary fine-grained lacustrine deposits (mainly unit Qvol and some unit QTI; fig. 5) that compact when dewatered. Layer 3 contained fine-grained interbeds in a few places but was mainly composed of sand- and granule-size particles. Layer 4 is assumed to have low compressibility, and therefore, was assumed to be minimally susceptible to compaction because the deposits of this aquifer range from moderate- to well-consolidated fluvial coarse-grained deposits associated with the geologic unit Tmf described in table 1. Delayed drainage was not simulated in this model because we were unable to simulate the magnitude and trend of subsidence near Lucerne Lake while using delayed drainage. Subsidence was therefore simulated using only instantaneous drainage.

Storage and subsidence related properties associated with instantaneous drainage (fluid-pressure equilibration) of interbeds were estimated from Siade and others (2014) and were used as initial values for the SUB package in the LVHM. These initial parameter values included the properties of elastic and inelastic storage coefficients ($S_{ke}$ and $S_{kv}$, respectively).

In general, the storage coefficient ($S$, dimensionless) is the sum of the skeletal storage ($S_k$) and the storage attributed to the compression/expansion of water ($S_w$). These storage coefficients can be expressed as a product of sediment thickness, $b$, and the corresponding specific-storage value, $S_{sk}$ for the aquifer skeleton, and $S_{sw}$ for water. Generally, because it is assumed that only the fine-grained deposits (interbeds and confining units) deform inelastically (Hoffmann and others, 2003),

Values for $S_{ke}$ can be defined for the fine-grained $\left(f\right)$ and coarse-grained deposits $\left(c\right) as$:

For this model, only the fine-grained component was specified in the SUB package, and only deformation of the fine-grained deposits was simulated. The aquifer-system specific storage, $S_s$, specified in the UPW, accounts for the volume of elastic storage changes resulting from $S_{sw}$ and $S_{ke}\left(c\right)$ in model cells where aquifer-system compaction and subsidence were simulated. Generally, $S_{kv}$ values were on the order of 10^–4 and $S_{ke}$ values were on the order of 10^–6, similar to specific storage $S_s$.

Preconsolidation head ($h_c$), or preconsolidation stress, is another parameter used by the SUB package that strongly affects storage changes, particularly the timing of those changes. Poland (1984, p. 335) defined preconsolidation head as “the maximum antecedent effective stress to which a deposit has been subjected, and which it can withstand without undergoing additional permanent deformation.” The preconsolidation stress represents the threshold value that determines whether changes in stress deform the aquifer skeleton elastically or inelastically. If the stress is less than the preconsolidation stress of the sediments, the deformation is elastic; if the stress is greater than the preconsolidation stress, the deformation is inelastic. Therefore, once the hydraulic head in an interbed, or confining unit, falls below the preconsolidation head, permanent inelastic compaction of that interbed occurs. During the modeling process, a new preconsolidation head is calculated at the end of each transient-state time step and is then used in the subsequent time step (Hoffmann and others, 2003). For the LVHM, the initial preconsolidation head values for all layers were estimated at 2,850 ft, which were about equal to the steady-state groundwater levels that were present near the center of the basin. The preconsolidation head values were adjusted using trial-and-error calibration techniques; they did not vary spatially within an individual layer in the LVHM.

Horizontal-Flow Barriers

As discussed in the “Hydrogeology” section, faults in the Lucerne Valley have been documented as, or are suspected to be, barriers to groundwater flow. The faults within the model domain were simulated using the Horizontal-Flow Barrier (HFB6) package (Hsieh and Freckleton, 1993; Harbaugh and others, 2000; Harbaugh, 2005; table 10). The HFB6 package allows for the simulation of thin, vertical, low-permeability geologic features situated between adjoining model grid cells. The HFB6 package assumes that the width of the fault is negligible in comparison with the horizontal dimension of the model grid cell. Conductance through the fault is controlled by a hydraulic characteristic that implicitly includes fault width (Harbaugh and others, 2000). Initially, the hydraulic characteristic value for each fault was set to a large value, allowing groundwater to freely flow across each fault. Through calibration, the hydraulic characteristic values were lowered until simulated groundwater levels matched closely to measured observations.

Three mapped faults (HELENFAUL, CUSHENBURY, SCOUGFAUL) and two previously unidentified barriers (HELENSPLAY, GMFAUL) that affected groundwater flow in the Lucerne Valley were simulated using the HFB6 package (fig.19; table 10). Model fault HELENFAUL simulated the Helendale Fault with a hydraulic characteristic value of 3.6E-06 ft/d; however, its trace was adjusted near the southern end of the simulated extent because measured groundwater levels by Thompson (1929) indicated a discontinuity in this area. Model fault CUSHENBURY simulated the Cushenbury Fault with a hydraulic characteristic value of 2.27E-06 ft/d. The horizonal-flow barrier HELENSPLAY was associated with the San Andreas fault zone (figs. 1, 19), which was described earlier in the “Geologic Structure” section and was simulated with a hydraulic characteristic value of 4.54E-06 ft/d. HELENSPLAY was inferred from groundwater-level data in the southern part of the basin (fig. 12E), which showed large groundwater-level differences between adjacent wells. Model fault SCOUGFAUL follows the trace of the Cougar Buttes Fault (fig. 19; table 10) and was simulated with a hydraulic characteristic value of 6.0E-05 ft/d. An additional horizontal-flow barrier (GMFAUL) was used in the northern part of the basin (fig. 19B) to more accurately simulate the discontinuities shown by the measured groundwater-level data (figs. 12, 14); this horizontal-flow barrier was simulated with a hydraulic characteristic value of 9.1E-06 ft/d.

Delivery Requirement

The TFDR is determined as the CIR of all WBS cells for irrigated crops and is increased sufficiently to compensate for irrigation efficiencies and ET from groundwater. The amount of ET from groundwater is a function of soil type, groundwater-table elevation, land-use type, root depths, and the anoxia and wilting points assigned to each land-use type. The TFDR data requirements for computing consumptive use on a cell-by-cell basis include soil type, land use (distribution of crop types and natural vegetation, including phreatophytes), and climate data. The consumptive use of each land-use type is simulated in this study based on root depths, irrigation efficiencies, runoff coefficients, crop coefficients, and the percentage of agricultural land use (tables 5–7). Although the exact value of these variables is uncertain, the basis of these estimates comes from various agricultural water-use and plant physiology studies (Rasse and others, 2000; Schenk and Jackson, 2002; Irrigation Training and Research Center, 2003; Food and Agriculture Organization of the United Nations, 2004; Sun and others, 2012; Ma and others, 2013).

Soils

The LVHM soils were simplified into six texture categories based on soil textural properties—sand, sandy loam, loamy sand, silty sand, silty loam, silty clay loam—from the Soil Survey Geographic Database (SSURGO; Natural Resources Conservation Service, 2017; fig. 23). Soil textural properties (percent sand, silt, and clay) were used to parameterize the soil textual properties program, Rosetta (Schaap and others, 2001), and estimate soil-water characteristic parameters using pedotransfer functions (Wösten and others, 2001). Rosetta is a program that models soil hydraulic properties, based on textural information, using neural network analysis (Schaap and others, 2001). The capillary fringe was estimated from the Brooks and Corey (1964) air-entry coefficient. These soil attributes were used for the entire simulation period, and the cell-by-cell distribution was independent of the crop type (table 5) and WBS. The FMP associates the distributed soil types with the specified capillary fringes and internal coefficients that allow individual analytical solutions for the calculation of ET (Schmid and others, 2006a).

Land Use

The FMP can be used to estimate components of consumptive use for a wide variety of land uses, including vegetation in irrigated or non-irrigated agriculture, fallow fields, phreatophytes and other natural vegetation, golf courses, and residential and municipal landscape settings. The FMP also can be used to simulate an assortment and spectrum of irrigation methods, including sprinklers and drip irrigation.

In the LVHM, the land-use attributes were defined on a cell-by-cell basis and included agricultural and native uses. Multiple land-use fractions were assigned to each cell when multiple land uses were present. Land use can change gradually or rapidly in response to changes in climate, urbanization, zoning, or farming practices. For the LVHM, the land-use patterns from 2005 to 2016 were changed on a yearly basis using crop distribution data (Mojave Water Agency, 2017). During this period, simulated land use had a maximum irrigated agricultural coverage (including landscaping) of 2.06 percent of the basin in 2013 (fig. 2; table 7). There were 14 land-use categories, referred to as “virtual crops,” that were defined based on crop-type indexes; these virtual crops were represented by an index number in the FMP (tables 5–7). For the simulation period for 2005–16, these virtual crops were used to drive the use and movement of water for each WBS.

Natural Discharge

Natural discharge occurred primarily as evaporation from the playa surface of Lucerne Lake (fig. 16). Historically, phreatophytes were sparse; some saltgrass survives along the edges of Lucerne Lake playa. Natural discharge was simulated near Lucerne Lake using the Drain package (DRN), but only during pre-development conditions and the first few years of the transient-state model (fig. 19) because by 1929, areas of natural discharge were no longer observed in the basin because of declining groundwater levels (Schaefer, 1979).

Groundwater Pumpage

Groundwater pumpage is the largest stress on the aquifer system of Lucerne Valley and is grouped into three categories for this study: (1) agricultural; (2) industrial; and (3) domestic, which includes municipal, domestic, and rural-residential supply. Agricultural pumpage estimates include pumpage from wells used to supply water for irrigation of crops and recreational uses and were either specified in the MNW2 package for 1942–2004 that pump from as many as three aquifers or estimated by crop demand using the FMP for 2005–16. Pumpage for industrial, municipal, and domestic water supply was specified by reported values (Mojave Water Agency, 2017) and estimated pumpage based on population described in the “Mechanisms of Discharge” section.

Agricultural Pumpage

Discharge from agricultural wells has not been metered in Lucerne Valley, and production estimates were not available before 1994. Therefore, agricultural water use was simulated by a combination of direct and indirect estimates of pumpage. Consumptive use derived from aerial photography data gave estimates of irrigated acreage, and a consumptive-use factor was applied to derive pumpage. Aerial imagery estimates are limited because they do not account for the combined consumption of precipitation and water applied for irrigation and do not capture the variability in consumption with climate (Schmid and others, 2006a; Hanson and others, 2014). The estimation of pumpage by using the FMP provides physically based, dynamic, and linked pumpage estimates as an alternative to indirect methods (Schmid and others, 2006a).

Before 1994, agricultural pumpage was estimated using aerial photography because detailed pumpage data were not available for Lucerne Valley. Agricultural water use was estimated to be 6 ft/yr based on the consumptive-use estimates for alfalfa (Erie and others, 1965). Because of limited agricultural well information, MNW2 well cells were distributed throughout the model domain based on land-use imagery, and pumpage was specified for each of these wells. From 1942 through 1977, aerial imagery was limited to seven datasets; changes in land use were only applied for years in which data were available, otherwise the last available aerial dataset was used. Starting in 1984, aerial imagery data were available for each year. Annual pumpage estimates by well were available from 1994 to 2016 (Mojave Water Agency, 2017) and were used to specify the agricultural pumpage volume and distribution from 1994 through 2004. Pumpage for agricultural supply from 2005 to 2016 was estimated using the FMP.

Pumpage for agricultural supply using the FMP is estimated as a combination of CIR and efficiencies required to satisfy the TFDR for all wells that deliver water to a particular WBS. Efficiencies include those from irrigation-water conveyance (canals or irrigation pipes) and potential losses from runoff and deep percolation below the root zone during irrigation. The groundwater pumpage required to satisfy TFDR can be estimated by considering any potential surface-water supply, the efficiency of irrigation, effective precipitation, and fractions of transpiration and evaporation within each model cell. Because all irrigation is supplied by groundwater pumpage in Lucerne Valley, surface-water supplies were not simulated. Pumpage is estimated by the FMP based on the TFDR. As many as 37 wells were used to simulate pumpage for irrigation; the number of wells and the pumpage varied during 2005–16 (fig. 24). Verified estimates of pumpage (Mojave Water Agency, 2017) were used as additional calibration data for agricultural pumpage.

For each well, the aquifer for which pumpage was simulated was based on the available construction information. Specifically, the drillers’ logs were analyzed, and the top and bottom screened intervals for each well were assigned where available. The FMP allocated pumpage to wells within a WBS by the average fraction of total required pumpage. Total pumpage for each WBS was distributed to each of the wells within the WBS based on the average pumping rate (Schmid and others, 2006a). Pumping rates are then limited by the maximum pumping capacity of each well.

PEST Observation Groups and Parameter Weights

Four observation groups were defined in PEST++: (1) groundwater levels for initial conditions, (2) transient-state groundwater levels, (3) drawdown, and (4) agricultural pumpage calculated by the FMP. Groundwater-level observations for initial conditions, which represented predevelopment conditions (steady state), included measurements in 32 wells documented in 1917 by Thompson (1929) and 21 “virtual” observations created from interpolated hydraulic head observations in the 1950s were used for steady-state calibration (fig. 25). Transient-state groundwater-level observations included 5,290 measurements in 86 wells (fig. 26) that represented conditions during 1942–2016. The 4,761 drawdown observations were based on the transient-state groundwater levels and were calculated as the groundwater-level rise or decline compared to 1942. Agricultural pumpage observations were based on 384 reported pumping volumes from 37 wells during the 2005–16 period. An additional 12 pumping observations were calculated as basin-wide annual agricultural pumping from reported pumping volumes at individual wells and used as calibration observations from 2005 to 2016.

Although calibration weights varied throughout the calibration process, final calibration weight values in the transient-state model were constant for hydraulic-head and drawdown observations. For the final calibration, sites with long-term observation records were weighted more heavily than sites with few observations so that long-term head and drawdown trends in the basin could be matched. The highest weights (1, 6, or 10) were given to steady-state hydraulic-head observations, and the transient-state hydraulic-head observations were given a weight equal to 1. Pumping observations were given a weight of 0.1, and basinwide annual agricultural pumping was given a weight of 1.

Comparison Between the Simulated Results and Observed Data

The ability of the transient hydrologic-flow model to accurately represent the observed data was evaluated throughout the calibration process. Observed data were the groundwater-level, drawdown, and land-subsidence measurements observed in the field, in addition to the reported agricultural pumpage. A comparison of the simulated and the observed values for pumpage was done to assess the capacity of the LVHM to reasonably simulate the effects of changing stresses on the aquifer system over time. The comparison of measured and simulated data provided an indication of model performance spatially and temporally. The resulting error distributions constrained the values of each simulated parameter, and the comparisons between the observed and simulated values provided a basis for the sensitivity values of selected parameters, discussed later in the “Sensitivity Analysis” section.

Simulated hydraulic heads were compared directly with measured groundwater-level elevations if the wells were perforated in a single model layer. For wells perforated in multiple model layers, OWHM2 calculated a composite, simulated equivalent hydraulic head, which is a function of the simulated model layers. The equivalent hydraulic head is calculated by using the hydraulic properties of the perforated model layers with the hydraulic-head observation (HOB) package (Hill and others, 2000).Measures of model fit included the following:

These statistics are measures of model fit; if the model results matched the measured groundwater-level data and observed land-surface deformation exactly, they would be equal to the simulated hydraulic head and subsidence. Similarly, when plotting the residuals against the simulated hydraulic heads, all the points would be at zero residual if the model results matched the data exactly.

Simulated Hydraulic Head and Measured Groundwater Levels

To determine how well the LVHM results reflected the measured data, the average RMSE between the simulated hydraulic head and the measured groundwater-level elevations (“groundwater levels”) were compared; the average RMSE for the entire transient-state simulation was calculated from RMSE values in table 8 as 14.91 ft, indicating a good fit, overall, of the simulated heads to the measured groundwater levels. The RMSE was small relative to the observed and simulated hydraulic head values that ranged more than 100 ft during the study period. The RMSE was lowest at well 5N/1E-15C1 in zone 33 (0.45 ft) and largest at well 5N/1E-22H1 in zone 33 (65.24 ft; figs. 21A, 21C; table 8), which was due to noisy observation data.

Table 8.    

Summary of model-fit statistics for differences between simulated hydraulic heads and measured groundwater-level elevations (U.S. Geological Survey, 2021) for the transient-state simulation, Lucerne Valley Hydrologic Model (Larsen, 2022), 1942–2016, Lucerne Valley, California.

[NWIS, National Water Information System; USGS, U.S. Geological Survey; ft, foot]

Table 8.    Summary of model-fit statistics for differences between simulated hydraulic heads and measured groundwater-level elevations (U.S. Geological Survey, 2021) for the transient-state simulation, Lucerne Valley Hydrologic Model (Larsen, 2022), 1942–2016, Lucerne Valley, California.

State well number (NWIS site name)	USGS site number (NWIS site number)	Average (ft)	Median (ft)	Minimum (ft)	Maximum (ft)	Root-mean-square error (ft)	Number of observations
4N/1E-2Q2	342732116504701	−13.35	−13.34	−17.07	−7.72	13.58	11
4N/1E-4N1	342737116532401	4.39	4.85	1.31	7.00	4.97	3
4N/1E-5G2	342803116535501	6.23	6.90	2.23	9.56	6.93	3
4N/1E-5H1	342757116533401	−5.19	−3.49	−19.21	6.02	8.23	37
4N/1E-5P2	342736116540401	−9.09	−8.87	−17.49	−6.12	9.37	35
4N/1E-6H1	342757116543001	10.07	10.80	−10.10	23.32	14.86	7
4N/1E-6N2	342738116553101	5.49	5.49	3.54	7.43	5.82	2
4N/1E-7C1	342725116551501	1.16	1.16	1.16	1.16	1.16	1
4N/1E-7D1	342725116553101	13.58	13.58	13.58	13.58	13.58	1
4N/1E-7K1	342658116545901	−13.54	−13.54	−13.54	−13.54	13.54	1
4N/1E-7P2	342645116551501	−14.67	−14.67	−14.67	−14.67	14.67	1
4N/1E-7P3	342646116551401	−19.64	−19.64	−19.64	−19.64	19.64	1
4N/1E-7R1	342645116545001	−30.75	−29.66	−38.32	−24.09	30.99	33
4N/1E-8P3	342645116541101	0.45	1.08	−1.84	1.48	1.42	4
4N/1E-9C1	342725116531901	−4.86	−5.07	−6.43	−2.39	4.99	11
4N/1E-9D4	342728116531901	−12.27	−12.76	−20.34	−5.62	12.96	19
4N/1E-9N1	342645116532401	4.16	4.71	1.18	6.57	4.72	3
4N/1E-12R1	342638116492601	−31.82	−31.82	−31.82	−31.82	31.82	1
4N/1E-13E1	342613116502201	−37.36	−41.69	−50.49	−19.91	39.51	3
4N/1E-13M1	342603116501801	2.86	2.68	−4.24	8.39	4.03	68
4N/1E-14Q1	342545116503901	−3.55	−5.28	−8.19	7.04	5.62	9
4N/1E-15R1	342546116513901	−42.78	−43.03	−50.32	−31.82	43.15	11
4N/1E-16N1	342547116533101	−35.04	−35.63	−37.42	−32.40	35.08	124
4N/1E-16N2	342547116532701	−43.16	−43.64	−44.97	−40.98	43.18	124
4N/1E-17L1	342609116541001	−20.68	−21.26	−31.88	−11.01	21.45	10
4N/1E-17Q1	342553116535601	−29.11	−29.11	−29.11	−29.11	29.11	1
4N/1E-18C2	342627116551801	−3.72	−3.72	−3.79	−3.65	3.72	2
4N/1E-18C3	342633116551001	−32.88	−32.88	−44.09	−21.68	34.74	2
4N/1E-20E1	342530116542101	−17.12	−18.28	−22.56	−8.39	17.78	12
4N/1E-22A1	342542116513401	−60.28	−60.06	−63.93	−55.65	60.36	5
4N/1E-23K1	342518116505401	45.52	45.40	43.11	49.26	45.54	329
4N/1E-23K2	342518116505402	25.85	44.32	−6.49	58.47	36.09	324
4N/1E-23R1	342504116503801	56.85	56.77	56.47	57.15	56.85	11
4N/1E-27D1	342441116522201	5.21	5.21	3.85	6.57	5.39	2
4N/1E-28F1	342439116530801	1.82	1.84	1.52	2.21	1.83	5
4N/1W-1H1	342759116553801	−0.16	−4.47	−7.78	11.78	8.55	3
4N/1W-1P3	342738116561701	−10.04	−10.04	−10.04	−10.04	10.04	1
4N/1W-1P4	342737116561801	−17.60	−17.60	−17.60	−17.60	17.60	1
4N/1W-1R3	342744116553901	−1.78	−1.78	−1.78	−1.78	1.78	1
4N/1W-1R4	342738116553901	4.03	4.29	−2.17	7.44	4.49	205
4N/1W-1R5	342738116553902	−7.21	−7.25	−13.42	10.00	7.44	205
4N/1W-1R7	342738116553904	−2.53	−1.34	−32.69	−0.59	3.82	203
4N/1W-1R8	342738116553905	−8.27	−8.29	−13.95	2.20	8.40	296
4N/1W-1R9	342738116553906	−2.09	−0.96	−27.16	−0.21	3.39	212
4N/1W-2G2	342803116570501	−15.83	−15.83	−15.83	−15.83	15.83	1
4N/1W-3D1	342817116583901	−36.11	−40.88	−45.18	−10.23	37.93	9
4N/1W-11B4	342725116570601	−28.07	−28.07	−28.07	−28.07	28.07	1
4N/1W-11B5	342723116570301	−49.78	−49.78	−49.78	−49.78	49.78	1
4N/1W-11J1	342658116564901	−15.26	−15.26	−15.85	−14.67	15.27	2
4N/1W-12H1	342711116554601	5.28	3.64	−12.77	26.63	15.08	4
4N/1W-13C2	342637116562001	−23.39	−23.39	−30.80	−15.98	24.53	2
4N/1W-13R1	342544116555001	−0.55	−0.62	−3.88	1.96	1.55	322
4N/1W-13R2	342544116555002	−3.03	−2.79	−10.74	8.52	5.35	341
4N/1W-13R3	342544116555003	−8.70	−6.85	−24.60	−4.73	9.74	104
4N/1W-14A2	342629116564401	17.40	18.57	3.64	22.09	18.15	10
4N/2E-17H1	342615116472301	14.17	14.17	14.17	14.17	14.17	1
4N/2E-17H2	342612116471601	3.55	1.76	−1.07	13.46	5.78	11
5N/1E-6N1	343253116552801	−3.45	−3.21	−16.67	2.03	5.31	18
5N/1E-7C2	343239116551401	−11.86	−11.86	−12.23	−11.50	11.87	2
5N/1E-8N3	343155116543401	−5.33	−5.41	−7.91	−1.82	5.48	327
5N/1E-8N4	343155116543402	−5.12	−5.16	−7.81	−1.70	5.28	282
5N/1E-10H2	343223116513301	4.00	3.54	−0.53	9.17	4.84	20
5N/1E-15C1	343148116521301	0.79	0.79	0.50	1.08	0.84	2
5N/1E-17D1	343153116542301	−9.93	−10.52	−20.07	−4.27	10.46	202
5N/1E-17Q1	343107116535501	−0.71	0.42	−5.11	2.54	3.30	3
5N/1E-20N1	343009116542801	−0.54	−2.42	−3.92	4.71	3.80	3
5N/1E-21R1	343014116523701	−3.34	−3.11	−6.09	−0.65	3.75	6
5N/1E-22H1	343030116513701	40.07	38.25	−12.20	95.98	65.24	4
5N/1E-27D1	342959116522101	−8.81	−10.39	−10.46	−5.58	9.10	3
5N/1E-27M1	342933116522101	−12.29	−12.29	−14.10	−10.49	12.43	2
5N/1E-28D1	342955116533401	−4.30	−4.30	−4.33	−4.26	4.30	2
5N/1E-32B2	342910116534801	0.45	1.50	−6.56	8.09	5.09	11
5N/1E-32D1	342906116543701	18.46	17.21	14.61	22.03	18.69	7
5N/1E-32L1	342837116540401	−5.08	−6.20	−16.29	12.52	8.69	25
5N/1E-35F1	342854116510201	−11.70	−11.70	−12.09	−11.32	11.71	2
5N/1E-35F3	342851116510701	−17.12	−17.12	−17.80	−16.44	17.13	2
5N/1W-1R3	343247116555001	−6.65	−6.64	−11.96	0.34	7.03	35
5N/1W-25G1	342943116555201	−3.09	−4.11	−27.22	20.52	6.74	222
5N/1W-25H1	342949116554501	−8.29	−8.29	−26.59	10.01	20.09	2
5N/1W-35L1	342844116571701	−18.50	−18.50	−18.50	−18.50	18.50	1
5N/1W-36F1	342850116562301	−1.02	−1.03	−4.28	1.94	1.78	189
5N/1W-36F2	342850116562302	−0.94	−0.87	−4.10	2.74	1.68	189
5N/1W-36F3	342850116562303	−1.92	−1.72	−6.91	1.04	2.29	189
5N/1W-36F4	342850116562304	−2.10	−1.89	−7.11	0.82	2.45	189
5N/1W-36F5	342850116562305	−5.35	−5.16	−13.38	−1.79	5.80	189
5N/1W-36P1	342826116561001	−5.55	1.85	−29.90	8.00	14.70	5
5N/1W-36R1	342826116553901	6.06	5.04	−4.59	22.52	11.05	9
6N/1W-36J1	343338116553801	3.27	2.04	−5.33	18.65	8.99	5

Another measure of model fit compares the simulated hydraulic head to the measured groundwater levels on a graph; an exact match would cause all observations to lie on the one-to-one (1:1) correlation line. The simulated hydraulic heads for the LVHM are plotted against the measured groundwater levels shown on figure 27. Simulated hydraulic heads that are above the 1:1 correlation line indicate that the model overestimated the measured groundwater levels; conversely, simulated hydraulic heads that are below the line indicate the model underestimated the measured groundwater levels. Overall, the simulated hydraulic heads and measured groundwater levels generally followed a 1:1 correlation line (fig. 27); however, 84 percent of the residuals were below the 1:1 correlation line, indicating that the model tended to underestimate groundwater levels. In addition, based on simulated residuals calculated from the difference between simulated hydraulic heads and measured groundwater levels (fig. 27), about 12-percent uncertainty was estimated for hydraulic-head results.

The location of a subset of observation wells and their corresponding hydrographs showing the simulated hydraulic heads and measured groundwater levels used for the transient-state model calibration are shown on figures 26 and 28, respectively. The simulated heads from layers corresponding to the measured groundwater levels from a well are shown on figure 28; the simulated heads from other layers are not shown. The wells were selected to represent temporal variations (groundwater-level trends, seasonal fluctuations, and vertical gradients) and optimize spatial coverage. Hydrographs that show measured data from multiple wells were compared with the average simulated hydraulic head between the wells’ locations.

The simulated hydraulic heads indicated that the model reasonably simulated the timing and magnitude of groundwater-level changes in Lucerne Valley. The greatest difference between simulated hydraulic heads and measured groundwater levels was in the southern part of the basin (figs. 26, 28). The discrepancies in this southern part of the basin could be attributed to uncertainties in the spatial discretization and distribution of geologic heterogeneity, particularly interbedded fine-grained materials with low hydraulic conductivity values, and the presence of unknown barriers to flow, such as faults. Mismatches between simulated hydraulic heads and measured groundwater levels can be attributed, at least in part, to uncertainty in the estimated distribution and magnitude of annual basin-wide pumpage for 1942–95. Estimates of the spatial and temporal distribution of pumpage from limited available data could result in excess or insufficient pumpage in some locations, causing an overestimation or an underestimation, respectively, of measured groundwater levels. Local variabilities in the hydraulic properties of the aquifer materials may not be adequately represented within the parameter zones of equal values, which also would contribute to the overestimation or underestimation of the measured groundwater levels.

The long-term measured groundwater-level data were used to evaluate the drawdowns, either decreases or increases, in hydraulic heads from the LVHM in all areas of the basin. To determine the drawdown in a well, the first groundwater-level measurement was assumed as the reference value and subsequent drawdowns were calculated as deviations from that first value. For the LVHM, drawdown was defined as the difference between the hydraulic heads from the steady-state simulation (before January 1942) and the end of the transient-state simulation (December 2016). Negative drawdown values represent increases in hydraulic heads compared to the reference value from the steady-state simulation. The drawdown data were evaluated for the three different periods to emphasize the importance of relative changes in groundwater levels at different points in time. The areal distribution of drawdown data for all model layers for 1942–2016 are shown on figure 29.

At the end of the model simulation period (2016), the drawdowns in all model layers generally were largest near the center of the basin and were less near the basin margins (fig. 29), which followed the trends observed by the measured data (figs. 12–14). In model layer 1, domestic and rural residential pumpage contributed to the simulated drawdown, but the amount of drawdown in layer 1 was less than that in the deeper layers because of the predominance of pumpage from the deeper layers and subsequent agricultural-return flows, as well as the artificial recharge from the BBARWA simulated effluent in the southern part of the basin (fig. 29A). The simulated drawdowns in model layer 1 ranged from about 110 ft near the center of the basin to an increase of about 35 ft near the BBARWA. The simulated drawdown pattern in layer 2 was similar to layer 1, with a decline of approximately 110 ft in the center of the basin and an increase of about 27 ft south of the Cushenbury Fault and Helendale Fault splay (figs. 29A, 29B). The simulated drawdown in model layer 3, which represented the middle aquifer, where most groundwater was withdrawn by agricultural pumpage, indicated widespread groundwater-level declines throughout most of the basin (fig. 29C). The simulated drawdown in layer 3 was spatially variable; about 125 ft of drawdown was simulated near the center of the basin, and hydraulic head increased about 14 ft south of the Helendale Fault splays and increased about 5 ft east of the Cougar Buttes Fault. Simulated drawdown in layer 4 also was spatially variable; about 130 ft of drawdown was simulated near the central part of the basin, hydraulic head increased about 10 ft south of the Helendale Fault splays and increased about 60 ft near the southeast margin of the model (fig. 29D).

Evaluating simulated drawdown for different periods can show how variations in hydraulic stresses cause groundwater levels to fluctuate temporally, areally, and with depth. Maps of the simulated drawdown for two additional periods (figs. 30, 31) were constructed to demonstrate the reversal of the groundwater-level trends observed in the measured data starting in the mid-1990s (figs. 13, 14). Figure 30 shows the drawdown by layer during 1942–95, which corresponds to the period with the largest pumpage volumes (fig. 10). The simulated drawdown for this period showed that the maximum decrease in hydraulic heads occurred in the center of the basin; the maximum simulated drawdowns from layers 3 and 4 were about 135 and 140 ft, respectively (figs. 30C, 30D).

As discussed in the “Long-Term Trends in Groundwater Levels” section, the amount of annual pumpage decreased after about 1996 (fig. 10), lessening the hydraulic stress on the aquifer system, and consequently, resulting in less drawdown and increases in groundwater-levels in the center of the basin (figs. 13, 14, 31). After the decline in the agricultural pumpage rates, groundwater in storage near the edges of the basin continued to migrate toward the center where the largest drawdowns were observed in the mid-1990s (figs. 15, 30, 31). Drawdown was evaluated for the later part of the simulation (1996–2016) to show how the decrease in the pumpage rates since the mid-1990s affected the simulated hydraulic heads across the basin. During 1996–2016, there were increases in the simulated hydraulic heads in all layers, particularly in layer 3 where most pumpage had occurred (fig. 31C). The pattern observed in the simulated drawdown for 1996–2016 is a reversal of the pattern that occurred for the entire simulation period of 1942–2016 (fig. 29) and for the early part of the simulation period of 1942–95 (fig. 30) and corresponds to the increase in the measured groundwater levels and simulated hydraulic heads discussed previously (figs. 14, 28). Simulated drawdowns in layer 3, for 1996–2016, showed increases in hydraulic heads as much as 15 ft in the center of the basin (fig. 31C). The migration of groundwater from the edges of the model to the center resulted in continued declines in the simulated hydraulic heads in the northern, eastern, and southern edges of the basin that ranged from about 5 to 15 ft between 1996 and 2016, whereas the simulated hydraulic heads increased in the center.

Subsidence Observations

Subsidence observation points were selected based on (1) monument locations from Sneed and others (2003) and (2) the location of maximum subsidence reported by the InSAR data for Lucerne Valley (Brandt and Sneed, 2017; fig. 32). Simulated subsidence and compaction for all observation points during July 2004 through November 2009, are shown on figure 33. Data were not available to determine when subsidence began, so the observed and simulated values began at 0 ft. Sneed and others (2003) estimated about 2 ft±5 ft of subsidence had occurred near the southern margin of Lucerne Lake from 1969 to 1998, and very little subsidence had occurred at the other observation points shown on figure 32. Subsidence likely began before 1969 because of the long history of pumpage, declining groundwater levels, and extensive clays and fine-grained materials at depth in Lucerne Valley. Simulated relative subsidence (relative to the beginning of the simulation period) ranged from almost no subsidence at observation point 13LC to about 7.5 ft (90 in.) at observation point InSAR_1 (fig. 33). Relative subsidence was slightly overestimated at observation points InSAR_1, RAIN, HOLM, and 49WB from Sneed and others (2003; fig. 33).

The maximum amount of subsidence in the Lucerne Valley that was estimated by the InSAR data, about 7.5 ft (2,286 mm), was simulated near observation point InSAR_1, which consists of the extensive fine-grained deposits near Lucerne Lake (figs. 16, 17). In the southern part of the basin, simulated subsidence ranged from greater than 1 ft at observation point R710 to less than 0.5 ft (152 mm) at observation point D720 (figs. 32, 33). Southeast of Lucerne Lake, simulated subsidence ranged from more than 4 ft (1,219 mm) of subsidence near the lake (observation point GOBR) to less than 0.5 ft (152 mm) near the Cougar Buttes Fault (observation point W710; figs. 32, 33). Simulated subsidence at station 49WB, located in the northeastern part of Lucerne Lake, indicated about 2.0 ft (610 mm) of subsidence and followed a pattern similar to the observed InSAR data in this area (figs. 32, 33).

The InSAR estimated land-surface deformation is plotted against simulated subsidence for all subsidence observations on figure 34. Simulated subsidence observations above the 1:1 trend line indicated that the model overestimated subsidence; conversely, simulated subsidence observations below the 1:1 trend line indicated that the model underestimated subsidence. Overall, the simulated subsidence skews above the 1:1 trend line, indicating some overprediction by the model. The total residual error for all simulated subsidence observations was 0.378 ft (115 mm), with a mean error in each observation of 0.008 ft (2 mm). The residual bias shows a tendency for the model to overestimate subsidence because there were a larger number of residuals greater than zero (88 percent) than there were less than zero (12 percent).

Mismatches between the observed and simulated subsidence values primarily can be attributed to the uncertainty in the thickness and lateral heterogeneity of the fine-grained sediments in model layer 2, which represents the confining unit. Uncertainties associated with the interpretation of the interferograms described in the “Land Subsidence” section of this report also can contribute to mismatches between measured and simulated subsidence.

Groundwater Budgets

The components of the groundwater budget for the 75-year transient-state simulation period (1942–2016) are shown on figure 35 and table 9. For the 1942–2016 simulation period, about 227,000 acre-ft of water was recharged into the Lucerne Valley groundwater basin, and about 692,000 acre-ft of groundwater was discharged from the basin, resulting in groundwater storage depletion of about 465,000 acre-ft (table 9). Almost all water discharged from the basin for the 1942–2016 simulation period was through pumpage (almost 100 percent of total discharge), most of which was for agricultural use (about 641,000 acre-ft or about 93 percent of total pumpage and total discharge; table 9). The total amount of municipal, industrial, and domestic pumpage accounted for a about 50,500 acre-ft or about 7 percent of the total pumpage and total discharge from the basin.

For the 1942–2016 simulation period, natural recharge was about 48,000 acre-ft, or about 21 percent of total recharge, and artificial recharge was about 35,000 acre-ft, or about 15 percent of total recharge (fig. 35B; table 9). Recharge from irrigation-return water, the amount of groundwater that returns to the aquifer system after being removed by pumpage, was a much larger component of total recharge than natural and artificial recharge for the 1942–2016 simulation period, accounting for a total of about 144,000 acre-ft, or about 63 percent of total recharge (fig. 35B; table 9). The volume of irrigation-return water was variable on an annual basis and averaged about 1,900 acre-ft/yr (fig. 35B; table 9). The simulated annual average recharge of treated wastewater that was applied to alfalfa fields and infiltration ponds between 1996–2016 (started in 1980; see the “Anthropogenic Recharge” section) was about 900 acre-ft/yr (fig. 35B; table 9).

Reduced pumping (beginning in the mid-1990s) decreased the rate of groundwater storage depletion. The annual average depletion in groundwater storage was about 7,700 acre-ft/yr for 1942–95 and about 2,900 acre-ft/yr for 1996–2016, which was a decrease of about 4,800 acre-ft or about 62-percent less depletion for the later period (table 9).