Scientific Investigations Report 2007–5009
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2007–5009
The aquifer system in the regional study area is composed of Tertiary and Quaternary alluvial deposits shed from the surrounding Sierra Nevada and Coastal mountain ranges. The basin-fill comprises coalescing alluvial fans, which tend to be coarse grained near the mountains and finer grained toward the center of the basin. The Corcoran Clay Member of the Tulare Formation is a lacustrine clay deposit that separates the basin-fill deposits into an upper unconfined to semiconfined aquifer and a lower confined aquifer throughout much of the regional study area (Burow and others, 2004). The local study area is underlain entirely by the Corcoran Clay.
Ground-water flow directions have changed and rates have increased because of extensive development. Under natural conditions, recharge occurred in the upper alluvial fans where streams enter the basin, and ground water discharged to the San Joaquin River and surrounding marshlands. Ground-water pumping for irrigation and public supply combined with delivery of surface water for irrigation has greatly altered the distribution of recharge and discharge (Burow and others, 2004). Ground-water flow has become more complex as increased surficial recharge and ground-water withdrawal have increased vertical flow in the system; water on a lateral flow path may be repeatedly removed by pumping and reapplied at the surface.
The Central Valley is a northwest-trending structural trough between the Sierra Nevada and the Coast Ranges (Bartow, 1991). The Sierra Nevada Range lies on the eastern side of the valley, comprises primarily pre-Tertiary granitic rocks, and is separated from the Central Valley by a foothill belt of marine and metavolcanic rocks. The Coast Ranges lie on the western side of the valley and are a complex assemblage of rocks, including marine and continental sediments of Cretaceous to Quaternary age (Gronberg and others, 1998).
The San Joaquin Valley can be divided generally into three physiographic regions (fig. 1): the western alluvial fans, the eastern alluvial fans, and basin deposits (Burow and others, 1998). Alluvial fan deposits on both sides of the valley are composed predominantly of coarse-grained sediments near the head of each fan that become finer grained toward the valley trough. The sediments in the eastern alluvial fan region generally are more permeable than sediments in the western alluvial fan region because sediment-source rocks and watershed characteristics are different between the two areas. The basin deposits within the regional study area are a combination of coarse-grained channel deposits and fine-grained deposits from flood events.
Consolidated rocks and deposits exposed along the margin of the valley floor include Tertiary and Quaternary continental deposits (fig. 4), Cretaceous and Tertiary marine sedimentary rocks, and the pre-Tertiary Sierra Nevada basement complex (Piper and others, 1939; Davis and others, 1959). The Mehrten Formation, the youngest of the Tertiary rocks, is composed of volcanic-derived mafic rocks. Unconsolidated deposits in the study areas generally comprise gravel, sand, silt, and clay primarily derived from granitic sources, and most are contained within the Pliocene-Pleistocene Laguna (not mapped at the surface in study areas), Turlock Lake, Riverbank, and Modesto Formations, including the Corcoran Clay. These are interspersed with minor amounts of Holocene dune, stream-channel, and flood-basin deposits (Arkley, 1962; 1964; Davis and Hall, 1959). The Turlock Lake, Riverbank, and Modesto Formations form a sequence of overlapping terrace and alluvial fan systems (Marchand and Allwardt, 1981) indicating cycles of alluviation, soil formation, and channel incision that were influenced by climatic fluctuations and resultant glacial stages in the Sierra Nevada (Bartow, 1991).
The Corcoran Clay, a lacustrine deposit that is a key subsurface feature in the San Joaquin Valley, is a Member of the Tulare Formation (Croft, 1972). The Corcoran Clay has been correlated with the E-clay (Page, 1986), the diatomaceous clay (Davis and others, 1959), and the lacustrine clay within the Turlock Lake Formation in the regional study area (Marchand and Allwardt, 1981). Page (1986) used results of previous work and a limited number of well logs and geophysical logs to map the areal extent of this regional confining unit. Additional lithologic data were used recently to modify the mapped extent of this prominent unit in the regional study area (Burow and others, 2004). The eastern extent of the Corcoran Clay roughly parallels the San Joaquin River valley axis (fig. 4). In the regional study area, the top of the Corcoran Clay occurs between 26 and 80 m below land surface and the unit has a maximum thickness of 57 m. In the local study area, the Corcoran Clay is encountered 40 m below land surface and is about 18 m thick.
The Mehrten Formation is a key subsurface feature tapped by wells in the eastern part of the study area (fig. 4). The Mehrten Formation marks a change in lithology and texture from overlying sediments of primarily unconsolidated coarse-grained sediments of arkosic composition to primarily consolidated sediments of volcanic-derived mafic materials (Burow and others, 2004). The Mehrten Formation outcrops in the eastern part of the regional study area and lies at least 120 m below land surface beneath Modesto.
Only limited information on the distribution and magnitude of aquifer hydraulic properties was available prior to this study. One available ground-water flow model was developed at a similar scale of investigation (Londquist, 1981), and it was highly generalized spatially.
Single-well hydraulic tests—slug tests—were done in 18 shallow wells in the bed and on the banks of the Merced River down-gradient from the local well transect as part of this study (Zamora, 2006). The sediments tested ranged from silty sand to well-sorted coarse sand, and computed hydraulic conductivity values ranged from 15 to 250 m/d and had a mean of 85 m/d. Slug tests done for a previous study in three wells above the Corcoran Clay along the eastern bank of the San Joaquin River within the regional study area were used to compute hydraulic conductivities of 2, 24, and 53 m/d (Phillips and others, 1992). The sediment types that these tests represent are unknown.
Permeameter tests of cores from the Corcoran Clay indicate vertical hydraulic conductivities ranging from 1 × 10–6 to 3 × 10–6 m/d (Page, 1977). Previous investigations, however, indicate numerous wells screened across the Corcoran Clay have locally short-circuited the impedance to flow across the clay layer and significantly enhanced intraborehole flow through the clay (Williamson and others, 1989; Belitz and Phillips, 1995). The vertical hydraulic conductivity estimated by Belitz and Phillips (1995) for an area southwest of the regional model, where the Corcoran Clay is about 100–240 m below land surface, is 1.6 × 10–4 m/d.
Because available data on the hydraulic properties of the aquifer system are scarce, material properties in this study were estimated on the basis of the distribution of sediment texture or hydrofacies derived from drillers’ logs, geologic logs, and geophysical logs. The regional texture distribution was estimated using the general approach of Laudon and Belitz (1991), which is easily applied to regional-scale problems. The local texture distribution was estimated using a geostatistical approach that incorporates factors related to the depositional environments (Carle, 1996). Although this method is more powerful than that of Laudon and Belitz (1991), it is difficult to apply to many regional-scale problems, owing to regional variability in depositional environments.
To facilitate this texture-based approach for the regional model, a database was constructed as part of a cooperative effort between the USGS and the Modesto Irrigation District to organize information on well construction and subsurface lithology in the study area (Burow and others, 2004). About 10,000 drillers’ logs were examined. Because sediment descriptions on drillers’ logs can be ambiguous and widely variable, a rating scheme was developed to select a subset of about 3,500 logs for further use. The database contains lithologic and well-construction information, which was used to vertically distribute ground-water pumping in the flow model.
A geostatistical approach was used to generate a heterogeneous hydraulic-conductivity field for the regional flow model using the primary texture of sediments from the database (Burow and others, 2004). Lithologic descriptions in the database were expressed as a percentage of coarse-grained sediment for 1-m depth intervals. These percentages were then interpolated using three-dimensional kriging, with no lateral anisotropy, over a grid equivalent to that of the regional model in the horizontal dimension for successive 10-m thick intervals in the vertical dimension. These values were then interpolated over the variable-thickness model grid to obtain an estimated distribution of sediment texture throughout the regional model domain. A more detailed description of kriging parameters is provided by Burow and others (2004). Estimated texture distributions for depth slices below the land surface are shown in figure 5. The estimated texture distributions were well constrained by sediment texture data to a depth of about 50 m, reasonably constrained to 100 m, and poorly constrained below 100 m. For the deeper parts of the aquifer system, for which no data were available, the texture value in the lowest layer estimated was duplicated in all underlying model cells. These cell-by-cell texture values were used to generate the distribution of hydraulic conductivity values as described in the section “Estimation of Aquifer Hydraulic Properties,” under “Ground-Water Flow Simulations.”
The spatial distribution of sedimentary hydrofacies in the local model area was simulated using Transition-Probability Geostatistical Software (TProGS) (Carle, 1996; Carle, 1997; Carle and Fogg, 1996; Carle and Fogg, 1997; Carle and others, 1998). Using TProGS, probabilities of encountering one defined hydrofacies next to another are determined from borehole and (or) other raw data. Markov-chain models approximating these transition probabilities are used in sequential indicator simulation, and these results are smoothed with simulated annealing to generate three-dimensional realiza-tions that honor the measured data points and transition probabilities. This method preserves cross-correlations between sediment types that can affect preferential flow through interconnected high-conductivity zones (Fogg and others, 2001).
A single realization generated using TProGS represents one of many possible distributions of the defined hydrofacies. A key advantage of this method is that the same input can be used to generate hundreds of realizations, which provides the opportunity to explore a wide range of equally probable hydrofacies distributions.
Geostatistical realizations of the local study area were bounded by or included three major geologic features of the unconfined to semiconfined aquifer: the Corcoran Clay, Pleistocene alluvial fan deposits, and incised Holocene river deposits (Marchand and Allwardt, 1981). The Corcoran Clay is a semi-permeable barrier to vertical flow between the unconfined to semiconfined aquifer and the confined aquifer, and was the lower boundary for the geostatistical realizations. Overlying the Corcoran Clay, Pleistocene alluvial fan deposits account for most of the permeable sediments. Adjacent to the Merced River, Holocene alluvium fills an incised river valley in the Pleistocene alluvial fan deposits. The lower boundary of Holocene alluvium along the Merced River was estimated using cores from the USGS well transect, and previous maps and reports on the surficial geology of the area (Marchand and Allwardt, 1978, 1981). These boundaries were used to specify the spatial extent of sedimentary sequences in geostatistical realizations.
Pleistocene alluvial fan deposits were characterized using geologic logs that describe the USGS transect wells. The logs had a maximum resolution of about 0.3 m for those boreholes drilled using an air hammer and about 1.5 m for those drilled using mud rotary. Clay, silt, silty sand, and sand were chosen as the representative hydrofacies on the basis of the frequency of observation of various sediment types in cores collected during drilling, Each lithologic description in the logs was assigned to one of these hydrofacies, and vertical transition probabilities (fig. 6), mean thicknesses, and volumetric proportions (table 1) were computed from this data set. Horizontal transition probabilities were estimated by multiplying the vertical mean lengths by factors of about 200 in the dip (X, toward the southwest) direction and about 100 in the strike (Y) direction (table 1). The scaling factors, which are greater along the depositional slope, were chosen on the basis of observations of horizontal continuity of sedimentary layers between wells and previous reports of the area’s sedimentary geology (Marchand and Allwardt, 1981).
Changes in stratigraphic orientation owing to the gradual downwarping of sediments after deposition, or areal variability in rates of deposition, were approximated by setting the dip parallel to the gently undulating surface of the Corcoran Clay at that interface, and to zero at the ground surface. In the interior of the alluvial fan sediments, dip values were assigned for each location (X, Y, and Z) by linear interpolation between the dip at the ground surface and at the surface of the Corcoran Clay. The dip direction was set parallel to the X axis.
For Holocene alluvium, sediments were assigned to hydrofacies containing silt, silty sand, or sand. Proportions and mean thicknesses were calculated from the upper part of a geologic log of a borehole at the lower end of the well transect (table 1). Available data were insufficient to characterize cross-correlations among sediment types; therefore, transition probabilities were derived using mean lengths and maximum-entropy Markov chain models (Carle and Fogg, 1997). Mean lengths (table 1) were estimated from the dimensions of the Merced River channel and local floodplain areas as mapped by Marchand and Allwardt (1978). The azimuth of the dip of the Holocene alluvium was set parallel to the centerline of the mapped alluvium and the dip was set to zero.
Two hundred realizations of the three-dimensional distribution of hydrofacies were generated for the local study area. For these realizations, drillers’ logs and USGS geologic logs were used as conditioning data. Conditioning refers to the inclusion of information that will affect the realization; the degree to which the realization is affected depends on the user-specified accuracy of the information. In this case, drillers’ logs contain a variety of terminology used to describe sediments and typically are not very detailed; therefore, most observations were treated as “soft” conditioning data by assigning a probability, p, where 0 ≤ p ≤ 1 for each of the simulated hydrofacies on the basis of the specific well log description (table 2). USGS geologic logs were used as “hard” conditioning data, where p = 1. Separate realizations were generated for the Holocene alluvium and Pleistocene alluvial fans. The simulated stratigraphic sequences were then merged into composite realizations that incorporated sequence boundaries, conditioning data, and simulated hydrofacies for the Holocene alluvium and Pleistocene alluvial fans above the surface of the Corcoran Clay.
Hydrofacies distributions for the local model improved the regional model in that area. Boreholes used to estimate regional sediment texture were sparse in the local model area, and did not include the three drilling sites along the USGS well transect. Regional kriging results, which depicted the local study area as being very coarse-grained throughout, compared poorly to data from sites along the transect, and the regional model did not closely simulate measured local conditions. Therefore, regional sediment texture estimates were replaced with TProGS-generated hydrofacies distributions in the local model area. The hydrofacies were translated crudely to percent coarse-grained materials as follows: sand and silty sand were considered 100 percent coarse-grained, and silt and clay zero percent. These percentages were averaged for all local cells within each regional cell.
Ground water occurs in the unconfined to semi-confined aquifer above and east of the Corcoran Clay, within the Corcoran Clay, and in the confined aquifer beneath the Corcoran Clay. The thickness of the unconfined to semiconfined aquifer above the Corcoran Clay ranges from about 26 to 80 m in the regional study area. East of the Corcoran Clay, this aquifer comprises primarily unconsolidated alluvial sediments, but includes the upper part of the more-consolidated Mehrten Formation. Irrigation and public-supply wells were completed in coarse-grained units in the upper Mehrten. The confined aquifer comprises alluvial sediments and upper Mehrten Formation sediments from beneath the Corcoran Clay to the deepest occurrence of fresh water. The contribution of ground water from the consolidated rocks beneath the primary aquifers was assumed negligible and was not considered.
Under natural conditions, ground water was primarily recharged at the upper parts of the alluvial fans where the major streams enter the valley (fig. 7). Ground-water flow followed the southwest slope of the basement complex and the dip of the overlying sedimentary deposits toward the southwest in the direction of the valley trough. Artesian conditions near the San Joaquin River indicated discharge to the river and surrounding marshlands (Davis and others, 1959).
Ground-water development in the basin changed ground-water flow patterns. Historically, water pumped for agricultural irrigation and irrigation return flows is much greater than natural discharge and recharge, which has caused an increase in vertical flow in the system (fig. 7B) (Page and Balding, 1973; Londquist, 1981). Ground water generally flows toward the southwest, as occurred prior to development (fig. 8). However, ground water moving along a lateral flow path may be extracted by wells and reapplied at the surface multiple times before reaching the valley trough (fig. 7B), at which point it may flow beneath the river toward pumping centers on the west side of the valley rather than discharge to the river. South of the Tuolumne River is a centrally located ground-water flow divide (fig. 8), east of which water flows eastward toward irrigation wells in an agricultural area with no surface-water supplies. West of the flow divide, water generally flows westward toward the valley trough.
The western part of the study area along the San Joaquin River is an area of ground-water discharge where the water table is shallow, often within 3 m of the land surface. Water is pumped from this area to lower the water table so crop roots are not damaged; this water is added to irrigation canals. Depth to the water table increases eastward, particularly south of the Tuolumne River, where depths can exceed 40 m within the large cone of depression (fig. 8).
Water levels near the city of Modesto generally decreased until the early 1990s (fig. 9). This decrease likely was caused by increased urban development and associated pumping for public supplies punctuated by drought conditions in 1976 and 1987–92. A series of wet years in the mid-1990s and completion of a surface-water treatment plant in 1994, which provided additional surface-water supplies for public and industrial uses, resulted in a recovery of water levels (fig. 9).
Data from USGS wells in the Modesto urban area show that vertical hydraulic gradients from the water table (about 10 m below land surface) to the deeper production zones of most municipal wells (about 40–100 m below land surface) are strongly downward and larger during spring and summer (about 0.2–0.3) than during fall and winter (fig. 10). Seasonal water-level fluctuations are about 6 m in the deeper wells and generally less than 1 m at the water table in the Modesto urban area.
Water levels measured in the transect wells in the local study area are shown in figure 11 (locations of wells shown in figure 3). These data show a weak downward gradient during the spring and summer below the almond orchard at the upper end of the well transect, and essentially no vertical gradient in this area the rest of the year. Measurements at the middle site show no vertical gradient year-round in the three deepest wells. The shallowest well at this site was screened in sandy silt and was not properly developed, but the hydrograph shows irrigation cycles in the adjacent corn field. Vertical gradients underlying natural vegetation near the Merced River at the lower end of the well transect were weakly upward, with the exception of several reversals when water levels were at their highest. Temporal variability of measured water levels was dominated by irrigation at the upper end of the well transect, the river stage at the lower end, and a combination of these influences in the middle. The hydrograph for the well below the Corcoran Clay at the lower end of the well transect (not shown in fig. 11) indicates a strong downward gradient across the unit during the irrigation season (average of about 0.3) and little or no vertical gradient the rest of the year.
Sources of recharge in the regional study area include agricultural return flow, infiltration of precipitation, reservoir leakage, and inflow from rivers. Ground-water discharges to pumping wells, rivers, transpiration, and evaporation. Information is available for some of these processes in the regional study area. Results from a recent 3-month study conducted by the Modesto Irrigation District, which manages the Modesto Reservoir, suggest a leakage rate of about 67,600 to 84,500 m3/d (Walter Ward, Modesto Irrigation District, oral commun., 2002). The interaction of ground water and surface water is largely unknown, but some data are available. A previous study of a 19-mile reach of the San Joaquin River within the regional study area showed predominantly gaining conditions during an 11-month period of drought conditions during 1988–89 (Phillips and others, 1991). Temperature and water-level measurements made within and below the Merced River as part of this study indicated predominantly gaining conditions within the local study area (Zamora, 2006). Additionally, a water-level map that includes the regional study area (fig. 8) suggests losing conditions in the upper reaches of the Stanislaus, Tuolumne, and Merced Rivers, and gaining conditions in the lower reaches.
Discharge to pumping wells was measured for urban wells and wells owned or used by irrigation districts. No data were available for pumpage from domestic wells. Although domestic wells are common in the study area, much of the water from these small-capacity wells is returned to the aquifer system by way of septic systems; therefore, the net use was assumed to be a small percentage of total pumping. Data on private agricultural pumping are not available; nor are data on evaporation from the shallow water table.
Estimates of many of the key components of recharge and discharge—agricultural return flow, infiltration of precipitation, transpiration, and private agricultural ground-water pumping—have been made using a water-budget approach (Burow and others, 2004) for water-year 2000 (October 1, 1999, through September 30, 2000). This approach and minor modifications are described generally herein; for details, the reader is referred to Burow and others (2004). Surface water and ground water are used for irrigation in the agricultural areas. Surface-water delivery data were available for most of the regional study area, although private pumpage records were not. Ground-water use in agricultural areas and areal recharge estimates were well constrained by the regional water budget.
The water budget was delineated geographically by dividing the regional study area into subareas for which surface-water deliveries could be obtained or estimated. A separate water budget was calculated for each of the resulting 47 subareas (fig. 12), which included agricultural and urban settings, foothill areas with natural vegetation, riparian areas having natural vegetation and (or) crops, and reservoirs.
Digital land-use data (California Department of Water Resources, 2001a,b) were used to estimate the water budget for subareas containing primarily non-urban land (Burow and others, 2004). The area of each crop or other vegetation type was determined using a geographic information system (GIS). A daily crop demand was calculated on the basis of crop or vegetation type and climate, and a total daily irrigation demand was estimated for each subarea. The total irrigation demand was met by a combination of surface-water deliveries, ground water pumped by irrigation districts, and private ground-water pumping. The total reported or estimated monthly surface-water and ground-water deliveries were subtracted from the estimated monthly irrigation demand to determine the monthly unmet irrigation demand. Private ground-water pumping was then assumed to equal this residual irrigation demand.
Recharge was estimated differently for areas containing native vegetation, irrigated crops, and urban land. The recharge for each of these areas estimated by Burow and others (2004) was reduced by 10 percent during calibration of the regional model. Recharge in areas with native vegetation was based on an assumed fraction of precipitation that did not run off and was not consumed (transpired). Because consumptive use for native vegetation was unknown, it was assumed that about 40 percent of precipitation runs off or is consumed in the foothill areas. This differs from the assumption in Burow and others (2004) of no runoff or consumption of precipitation in the foothills.
Recharge in irrigated agricultural areas was estimated as the portion of precipitation and irrigation not consumed by the crops (consumptive use). The consumptive use of applied water was estimated to be 63 percent for most of the regional study area on the basis of calculated values for two subareas receiving surface-water deliveries and having few known wells (among the 3,500 wells cataloged from drillers’ logs). This estimate was considered reasonable because crops are cultivated in sandy soils and irrigated by flooding and permanent sprinklers. Consumptive use of applied water was assumed to be 80 percent in the older fan deposits in the foothill areas, where sediments are more indurated and efficient irrigation methods such as micro-sprinkler and drip are used.
For the urban areas, recharge from delivered water was estimated using the minimum-month method to determine indoor and outdoor water use (California Department of Water Resources, 1994). Ten percent of the estimated outdoor use was subtracted from the total to account for leakage from water distribution lines (California Department of Water Resources, 1994). Fifty percent of the remaining outdoor water use was assumed to be consumptive use for landscape irrigation or runoff to streams, and the remainder was assumed to contribute to urban recharge (Burow and others, 2004). Recharge in the urban area from precipitation was assumed to be 50 percent of total precipitation because much of the runoff in Modesto and other urban areas is drained through large-diameter boreholes backfilled with rock. These “rock wells” are direct conduits to the water table.
The estimated average areal recharge rate for the regional study area was about 0.54 m/yr; the highest recharge rates were for the agricultural areas in the southwestern part of the study area (fig. 13) and along the rivers in the northeastern part. The lowest estimated recharge rates were for the foothills and the urban areas. Similarly, the highest estimated pumping rates were for the agricultural areas in the southwestern part of the regional study area (fig. 14). The relatively high rates of pumping and recharge in the southwestern agricultural areas were related, in part, to the irrigation efficiency and the supplemental pumping required to manage the shallow water table.
Samples collected from most of the USGS wells along the transect in the local study area were analyzed for sulfur hexafluoride (SF6), an environmental tracer that has been used to estimate ground-water age (Busenberg and Plummer, 2000). Atmospheric concentrations of SF6 were very low prior to 1965 and have increased since then, making it a good means of dating young ground water. Concentrations of SF6 in water samples from the USGS well transect (Larry Puckett, U.S. Geological Survey, written commun., 2005) are compared in this report to concentrations estimated using simulated travel times from the water table to these wells and the historical atmospheric concentration of SF6 (Böhlke, 2005).
U.S.
Department of the Interior | U.S. Geological Survey
Persistent URL: https://pubs.water.usgs.gov/sir20075009
Page Contact Information: Publications Team
Page Last Modified: Thursday, 01-Dec-2016 19:40:16 EST
