The landscape consumptive use for every model stress period is computed using (1) reference evapotranspiration (ETref) for each model cell that is approximated using PET, (2) a seasonally varying crop coefficient (Kc), which is a scalar value that is multiplied by ETref to estimate a landscape consumptive use, (3) specification of the fraction of the maximum leaf area in each cell (that is, fraction of transpiration; FTR) and fractions representing the remainder of the area subject to evaporation from precipitation (FEP) and irrigation (FEI), and (4) the land use area.

The ETref is a reference value that assumes a well-watered grass surface and is used for landscape consumptive use calculation. The Kc value represents a stage of crop growth and associated landscape consumptive use. The FTR represents the fraction of land use “leaf area” in each model cell where plant transpiration occurs and varies between 0 and 1. The FTR is assumed to be independent of whether the transpiratory portion of landscape consumptive use is satisfied by irrigation, precipitation, or groundwater uptake. The remaining fraction (1−FTR) of each cell is assumed to represent FEP and FEI. The FEP represents the fraction of the land use area over which evaporation occurs and is calculated internally by FMP as 1−FTR. The FEI can be conceptualized as the irrigated area that is not planted, such as the irrigated area between plants or planting beds. FEI is specified as a fraction less than or equal to FEP. The FTRs vary linearly with the respective area occupied by crops and the area open to soil evaporation (Schmid and others, 2006). FTR or FEP and FEI depends strongly on the type of land use and the associated crop growth stage. When the vegetation cover approaches 100 percent, FTR=1 and FEP and FEI=0. As a result, the fractions of transpiration and evaporation vary by land use type for different months of the year.

The landscape consumptive use is the sum of transpiration and evaporation consumptive use. The transpiration consumptive use by plants is estimated based on FTR, ETref, and Kc parameters. The transpiration consumptive use is computed as ETref multiplied by each land use area and Kc in each cell. The transpiration consumptive use is then prorated by the monthly FTR for each land use type. The evaporative consumptive use is computed using the FEP and FEI parameters that are multiplied by the volumes of precipitation and irrigation that are applied to each cell. The consumptive use due to transpiration and evaporation are summed to estimate a landscape consumptive use in each model cell.

In FMP, landscape consumptive uses can be satisfied from natural sources, such as precipitation and direct uptake from shallow groundwater above a specified rooting depth. If natural sources of water are not available to meet the calculated landscape consumptive use for a WBS, a TDR is computed. For each model time step, FMP determines a residual of total landscape consumptive use that cannot be satisfied by natural sources—the TDR. For irrigated land uses, this residual water demand is increased using an on-farm efficiency factor (OFE) to account for crop, WBS-specific, and irrigation-type inefficiency losses to estimate the amount that must be supplied to meet demand. Available supplies are used to meet the TDR for the WBSs. The TAW represents the amount of water applied to meet land use water demands. In the integrated hydrologic models, the deficit-irrigation scenario is used; when demand cannot be satisfied with available supplies for a WBS, demand is reduced to the supply, and the deficit is shared among all land uses in the WBS. If the irrigated water demand cannot be satisfied with available water supply, then the TAW will be less than the TDR. Due to the prevalence of groundwater wells with substantial pumping capacity throughout the study area, this deficit irrigation scenario is unlikely to occur under current conditions. If constraints were placed on well pumpage in future simulations, this deficit scenario would affect the calculation of demands. More details for how the FMP accounts for inflows and outflows for each WBS are available in the MF-OWHM documentation (Schmid and others, 2006; Schmid and Hanson, 2009; Hanson and others, 2014a; Boyce and others, 2020).

The TDR in irrigated lands can be satisfied by additional water supplies, such as semi-routed surface-water deliveries from streams or canals (diversions), reservoir releases, non-routed delivery from external sources (for example, wastewater reclamation or pipelines), or groundwater pumpage by wells. Where the conjunctive use of surface water and groundwater are major sources of water used for irrigation, FMP attempts to satisfy the TDR by using surface-water diversions or non-routed deliveries (such as pipelines and recycled water) first, with residual water demand satisfied by groundwater. Surface-water deliveries can be limited to a specified allocation (surface-water allotments) to the agricultural WBSs that use both surface water and groundwater (Hanson and others, 2014b, c, d). If diverted flows to a WBS through a semi-routed delivery are more than land use water demands, diverted surface water is returned to the surface-water drainage network for potential reuse downstream. For each WBS, the FMP computes the collective potential pumping capacity of all wells that can provide groundwater for supplemental irrigation water. The residual water demand is distributed to every well in the WBS, and all active wells are pumped. The volume of groundwater pumpage is only limited by well capacity and any imposed volumetric constraints specified using groundwater allotments.

Runoff and recharge from each WBS are partitioned based on a specified fraction of excess water after all demands have been satisfied. Runoff is routed on a segment-length weighted basis to all streams within a WBS. Recharge results from excess irrigation and excess precipitation, reduced by losses to surface-water runoff and evapotranspiration from groundwater (Schmid and others, 2006). The evapotranspiration from groundwater is subtracted from the potential net downward flux as deep percolation to the uppermost aquifer. Hence, recharge to groundwater can be affected both by user-specified and head-dependent processes. This definition of recharge requires the following assumptions: deep percolation below the active root zone is equal to groundwater recharge; evapotranspiration from groundwater equals an instantaneous outflow from aquifer storage in any time step; and the net change in soil-moisture storage for irrigated, well-managed agricultural areas for periods of weeks to months is negligible (Schmid and others, 2006). The recharge to the aquifers is applied on a cell-by-cell basis to the uppermost active model cell in each WBS. Recharge is computed after evapotranspiration consumption losses. Therefore, recharge can be negative if groundwater evapotranspiration is greater than deep percolation. When negative recharge occurs, it is constrained to specific areas. For example, in the model cells near the stream in the Riparian analysis region, the amount of groundwater evapotranspiration from riparian vegetation may be higher than the amount of deep percolation into those cells, resulting in a negative recharge value; however, stream leakage is also occurring in those cells. Therefore, the recharge may be negative, but the total recharge (stream leakage plus recharge) is positive.

The FMP also includes a surface-water operations (SWO) module that allows for simulation of large-scale surface-water storage and distribution systems—including simulation of surface-water storage, allocation, release, and distribution—to meet agricultural, municipal, and industrial water demand, maintain a minimum streamflow requirement, and reserve a portion of storage for flood protection. This additional functionality facilitates improved analysis of basin-scale conjunctive use and large-scale surface-water management. To simulate two-way interactions and feedback between surface-water and groundwater management and use, SWO interacts with the Stream Flow Routing Package (SFR2; Niswonger and Prudic, 2005) and with the FMP. Reservoir dynamics are simulated using SWO by simulating surface-water storage, management objectives, allocation, and reservoir release. Reservoir dynamics using SWO determines downstream diversion amounts and makes reservoir releases as inflow to SFR2 that routes flow through a surface-water drainage network to simulate distribution. Then SFR2 calculates the associated groundwater/surface-water interaction. The appropriate reservoir releases and downstream diversion amounts are determined by SWO based on the surface-water conveyance calculated from SFR2. Reservoirs simulated by SWO can provide supply to FMP WBSs. Each connected WBS has a computed water demand, and SWO can deliver the appropriate water based on allocation, storage, and SFR2 stream gains and losses.

SWO was developed to allow analysis of the complete feedback cycle between surface-water and groundwater management and use, including effects of groundwater management and use on reservoir storage, allocation, and releases and corresponding effects of surface-water management on groundwater recharge and demand. SWO can be used to evaluate how changes in groundwater pumping affect surface-water management, including storage, allocation, release, and distribution of surface-water supplies. Notably, SWO also allows for analysis of how changes in surface-water management affect groundwater recharge via stream leakage and deep percolation of applied water and thus how surface-water management affects groundwater storage and movement. SWO further allows for analysis of how changes in surface-water management affect groundwater demand and use due to changes in surface-water allocations and deliveries, thus providing for complete two-way interactions between groundwater and surface-water management and use.

There are two related measures that describe the potential for evapotranspiration, ETref and PET. The ETref is a reference value that assumes a well-watered grass surface and is used for crop demand calculation. The PET is the total potential for evaporation from a surface if evaporation is not limited by water availability (Allen and others, 1998). The two quantities are related, but PET values may be higher seasonally. For all locations within the integrated hydrologic model area, the average annual ETref rate, calculated by CIMIS, exceeds average annual precipitation (Baillie and others, 2015; California Irrigation Management Information System, 2020). The highest ETref rates of 53–62 in. per year occur in the lowlands that define the valley floor in the central and southern parts of the integrated hydrologic model domain (California Irrigation Management Information System, 2020). Mean monthly ETref rates for these locations vary from 1 to 1.5 in. for December and from 8 to 9 in. for July (California Irrigation Management Information System, 2020).

The ETref values in the integrated hydrologic models are estimated using adjusted PET, as described here. Measured field-scale ETref from California Irrigation Management Information System (2020) was compared to 270-m resolution PET (Hevesi and others, 2022) at CIMIS stations within the study area (figs. 3, 5). These 270-m climate data were mapped using an area-weighted approach to the uniform 530-ft model grid. Comparison of WY-averaged ETref and PET for stations in the study area that have long-term records (greater than 9 years) show reasonable correspondence, with an R² value of 0.897 (fig. 17A; Python Software Foundation, 2021). Regressions for all stations did not have good correlation, so only long-term stations with records greater than 9 years were evaluated. The time series of values at stations with long records were further evaluated at these stations: Castroville (CIMIS 019, fig. 17B), Arroyo Seco (CIMIS 114, fig. 17C), Salinas South (CIMIS 089, fig. 17D), King City-Oasis Rd. (CIMIS 113, fig. 17E), and Salinas North (CIMIS 116, fig. 17F). The ETref records and PET estimates at these sites show reasonable correspondence, although PET is slightly underestimated in the winter and slightly overestimated in the summer. Accordingly, seasonal bias adjustment factors are used in this study to adjust PET values to estimate ETref for crop demand calculations (further discussed in the “Crop Coefficients” section of this report).

To show the relation between physical land use acreage compiled for the study and harvested acres, time series for eight crops are compared for calendar years 1967–2014 (fig. 18). To better illustrate the representation of multi-cropping in the simulation, simulated acres for multi-cropped land uses were estimated for high frequency rotational and annually stable crops that can have multiple harvests. To estimate the simulated crop acres, the physical crop acres were multiplied by a multi-cropping factor of 1.97, as defined and used in the previous SVIGSM (Boyle Engineering Corporation, 1987; Yates, 1988; Montgomery Watson, 1997). The time series show the comparison of estimated harvested acres from county crop reports, estimated physical acres using CalPUR data and land use GIS data (Henson and Jachens, 2024), and simulated acres assuming a multi-cropping factor of 1.97 applied to the estimated physical acres (fig. 18). The data show that harvested and simulated acres have similar trends, variability, and magnitudes. The trends of increased strawberries and vineyard acreage starting in the 1980s, the reduction in root vegetables (for example, sugar beets), and the steady increase in lettuce production correspond well (fig. 18). Although some of the simulated acreage values are lower than the harvested acres (for example, crucifers), these comparisons are for demonstration. The model simulates a long growing season with multiple harvests that account for the water demands due to multi-cropping but does not simulate individual harvests. After 2014, the strawberry and vineyard physical acreages do not increase as much in the model as the harvested acres reported in crop reports. Some of this effect could be due to differences between multiple harvests that may be counted in harvested acres. Physical acreage area is always the area that the crop occupies regardless of harvest. Although some differences among crop group harvested and simulated acres are present, the land use captures much more of the distribution and variability in crop group acreage than the common practices of estimating a crop group area that remains the same between land use datasets (for example, step functions in crop group acreage or linearly interpolating crop group acreages between land use datasets).

Each of the land use categories in the study are represented by seasonally varying crop properties. The closer association between crop type and properties used in this study allows for better definition of individual crops of interest and supports future refinements of crop properties in the model as data become available. When available, published Kc values for similar coastal areas were used (Brouwer and others, 1985; Brouwer and Heibloem, 1986; Snyder and others, 1987a, b; Allen and others, 1998; Michael Cahn, University of California Agriculture and Natural Resources, Crop Manager, written communication, 2018). When no published Kc values for coastal areas were available, published Kc values for the western San Joaquin Valley compiled by Brush and others (2004) were used. Additional specific Kc values were used for greenhouse crops (Orgaz and others, 2005), turfgrass (Gibeault and others, 1989), and strawberries (Snyder and Schullbach, 1992; Hanson and Bendixen, 2004). The Kc values from the literature were adjusted to account for differences between field-measured CIMIS ETref and 530-ft resolution gridded PET used in this study using a seasonal bias adjustment factor. A seasonal bias adjustment factor for winter, spring, summer, and fall was computed to reduce the sum of squares error from the simulated PET and estimated ETref from CIMIS station data at five long-term stations throughout the Salinas Valley. The seasonal bias adjustment factors for PET were highest in the winter (1.00) and lowest in the summer (0.88), with a value of 0.94 for the spring and 0.97 for the fall. The seasonal bias adjustment factor for each season was multiplied by the Kc value to account for the difference between PET and ETref, and the Kc values for each of the land uses in each land use category was defined (fig. 19; Henson and Culling, 2025).

Superimposed on the consumptive-use estimates are additional climatic-stress scale factors applied to Kc values as seasonal wet, normal, or dry scale-factor parameters that were estimated during parameter estimation and analysis. These climate-stress scale factors are applied to adjust Kc values for the period before and after 1995 in the model input (further described in the “Landscape-Process Parameters” section and defined in the model archive by Henson and Culling, 2025). Climate year type is defined using methods defined by MCWRA for reservoir operations (Monterey County Water Resources Agency, 2005, 2018; Henson and Jachens, 2022). Each year was defined as wet, normal, or dry based on the minimum storage in Lakes San Antonio and Nacimiento and the percentile of April 1st streamflow in Arroyo Seco. These seasonal scale factors are used to reflect potential differences in agricultural practices more appropriately among defined WBSs embedded in the consumptive-use estimates and the year-to-year changes in surface-water allocations and deliveries during the 1967–2018 simulation period. This approach is consistent with several studies of the region (Hanson and others, 2004, 2014b, c, d; Faunt and others, 2009b, c).

The complete time series of FTR and FEI values are provided in the model archive (Henson and Culling, 2025); a summary is provided here. The FEP is calculated internally by FMP as 1−FTR. There are no specific data for FTR available in the literature. The FTR was developed based on expert knowledge, crop type and month, informed by other model applications in the region that use MF-OWHM (Hanson and others, 2004, 2014b, c, d; Faunt and others, 2009b, c), and adjusted during parameter estimation using reported withdrawal data. The FTR values for cropped land uses range from 0.39 to 0.73. Conceptually, the FTR means that the leaf area of each land use in each cell can vary between 39 and 73 percent of the land use area. For non-irrigated land uses, the FTR only affects the consumption of precipitation and shallow groundwater from the root zone. Thus, the FTR values mainly affect runoff to the streams generated from these land uses. For many of the non-irrigated land uses (for example, beach-dunes, barren-burned, quarries, semiagricultural, and idle-fallow), constant values of 0.3 are assumed for land use. This constant value of 0.3 assumes that grasses and weedy vegetation covers as much as 30 percent of these areas. Constant values of 0.08 are assumed for urban areas, which assumes that about 8 percent of the urban areas are subject to evapotranspiration from precipitation, and 92 percent are subject to evaporation of precipitation. It is assumed that the urban water uses that include irrigation are included in M & I pumpage. The FEI is computed by crop for each irrigated land use type and can be conceptualized as the non-vegetated area that is irrigated (for example, sprinkler overspray). Through the simulation period, harvested acres have increased (fig. 7), yet water use has been relatively stable (fig. 14). Thus, FEI was reduced through time alongside assumed changes in irrigation methods. This reduction was done to represent the effects of shifting from sprinkler to drip irrigation and implementation of irrigation water conservation practices to reduce wetted areas that are not planted. This calculation is supported by documented increases in irrigation efficiency between 2001 and 2010, as noted by Sandoval‐Solis and others (2013) and Tindula and others (2013). FEI values for crops range from 0.05 to 0.15 before the year 2000 and 0.01 to 0.05 after 2000.

In general, irrigation efficiencies and irrigation system types are poorly known (Williamson and others, 1989; California Department of Water Resources, 1994; Brush and others, 2004). Data describing the association between crops and irrigation system type, distribution of irrigation systems, and associated efficiencies in the integrated hydrologic model area are not available. We used expert input from agricultural producers, regional information, and a statewide report (Sandoval‐Solis and others, 2013) to estimate irrigation system type throughout the simulation period. Specification of OFE and their variation in time was informed by similarly constructed hydrologic and agricultural models in this region of California, the Central Valley Hydrologic Model (Faunt and others, 2009b, c, 2024), Cuyama Valley (Hanson and others, 2014b), and the Pajaro Valley Hydrologic Model (Hanson and others, 2014c, d) and were adjusted during model parameter estimation. Williamson and others (1989) report values averaging 59 percent and ranging from 38 to 92 percent for the 1961–77 period. California Department of Water Resources reports overall efficiencies of 60–70 percent for parts of the Central Valley (California Department of Water Resources, 1994). The irrigation types specified in the integrated hydrologic models are a sprinkler and drip irrigation type for each of the two climate regions (inland and coastal) and nursery, flood, and urban for a total of seven specified irrigation types. For each of those irrigation types, an OFE was specified for each WBS. The OFE in the integrated hydrologic models vary between 0.3 for flood irrigation and 0.95 for drip irrigation. The OFE increase in time alongside assumed changes in irrigation methods (for example, shifting from sprinkler to drip irrigation) and improvements in irrigation water conservation practices. All OFE values used in the model are provided by Henson and Culling (2025).

Surface-water flow data used as input to the integrated hydrologic models include available monthly records of the three surface-water diversions (Henson and others, 2023), simulated monthly average watershed inflows (Henson and others, 2025a; Hevesi and others, 2025a, 2025b), and monthly historical reservoir inflows and releases (Henson and others, 2023). Simulated daily watershed inflows from the watershed model are aggregated to monthly average inflow time series at 148 watershed inflow points along the boundary of the integrated hydrologic model domain (fig. 20; Henson and Jachens, 2022). These monthly watershed inflow time series are input to the streams within the integrated hydrologic model simulation (Henson and others, 2025a). There are two watershed inflow points coincident with the location of reservoir releases. At these inflow points, the watershed model simulated monthly average natural flows are summed with monthly average historical reservoir flows. For the historical model, the watershed model inflows are summed with historical monthly average reservoir releases. For the operational model, the watershed model inflows are summed with simulated reservoir releases.

The surface-water drainage network in the integrated hydrologic model area represents the Salinas River, major canals, diversion channels, drains, and tributaries that drain each surrounding upland watershed outside the integrated hydrologic model area (Henson and others, 2022b). The topology of the surface-water drainage network (Henson and others, 2022b) was developed using analysis of surficial geology and land-surface elevations from the SVGF, National Hydrography Dataset stream data (U.S. Geological Survey, 2019b), and aerial imagery (U.S. Department of Agriculture, 2016). The surface-water drainage network simulates the distribution and conveyance of surface water within the integrated hydrologic model area (fig. 20). This network is represented by a collection of stream cells (referred to as “reaches”), which are combined to form a collection of cells or reaches known as a “segment.” The total surface-water drainage network contains 524 segments, 9,008 reaches (model cells), 3 diversions, 148 watershed inflows, and 2 outflows to the ocean (fig. 20). There are a total of 93 collector segments that collect surface runoff but do not discharge to the Salinas River or its tributaries. These collector segments represent abandoned drainage canals, intermittent arroyo channels not present in National Hydrography Dataset stream data and were specifically delineated to facilitate simulation of intermittent recharge so that the magnitude of recharge from these features can be evaluated as part of future sustainability analyses. These collector segments have high streambed permeability to facilitate infiltration. Estimated watershed inflows from 42 of the 148 watershed inflows are routed into intermittent runoff from these ungaged watersheds at watershed inflow points connected to the surface-water drainage network. All streamflow that does not infiltrate into the underlying aquifer or flow into another stream is assumed to be lost to evapotranspiration. Riparian vegetation evapotranspiration in streams outside of the Salinas River is not directly simulated. The flows into collector segments are intermittent such that flow in any collector segment was greater than 10 ft³/s in only 24 of the 1,224 model time steps (Henson and others, 2025a). Within the surface-water drainage network, channel bed elevations are specified on a model cell-by-cell (reach) basis using 1-m horizontal resolution light detection and ranging (lidar) data (U.S. Geological Survey, 2019a), where available, or a 10-m horizontal resolution DEM (U.S. Geological Survey, 2013). Streambeds were specified to be 1 ft thick throughout the network.

The surface-water flow simulation and water balance calculation used in SFR2 allow for streamflow routing, streamflow infiltration into the aquifer (losing stream reaches), and any potential base flow as groundwater discharge to streams (gaining stream reaches). Hydraulic flows among segments of the surface-water drainage network were simulated using two approaches based on available data for each stream segment: (1) a rating table approach that relates channel depth, width, and flow for a range of flow values and (2) an approach that assumes Manning’s equation and a wide rectangular channel. The rating table approach was applied to seven stream segments in the surface-water drainage network that contain a USGS streamgage with long-term streamflow records (USGS 11150500, USGS 11151700, USGS 11152300, USGS 11152500, USGS 11152050, USGS 11152000, USGS 11152650; U.S. Geological Survey, 2018; Henson and others, 2022b). The Manning’s equation approach was used for the rest of the stream segments. These stream segments were grouped based on the channel type (tributary, main channel, canal, ditch, or segment) that collects surface runoff and facilitate recharge but does not have a downgradient connection. Manning’s roughness coefficient for each segment was specified using literature values for natural channels, developed channels, and canals for each segment (Arcement and Schneider, 1989). Roughness coefficient values varied within the range of 0.02–0.05. When supported by local conditions, for example, in the upper Salinas Valley where vegetation in the channel can increase roughness, a value of about 0.05 was specified. The channel type was evaluated using adjacent land use through visual inspection of aerial imagery (U.S. Department of Agriculture, 2016). Hydraulic properties for these groups of segments have been parameterized to help with the parameter estimation of surface-water flow parameters. Surface-water flows were output using the HYDMOD package (Hanson and Leake, 1999).

To illustrate the spatial distribution of net stream leakage into the aquifer, the long-term (WY 1968–2018) annual average stream leakage in each stream segment was computed using output from the historical model. The net stream leakage for each segment was divided by the segment length to provide a normalized value in acre-feet per year per foot (afyf) of stream segment length. The resulting long-term stream leakage map is shown on figure 21. Stream leakage varies along the length of the river, such that substantial leakage in the Salinas River (greater than 0.5 afyf) and even greater leakage (greater than 1.5 afyf) occurs in the center of the model domain, and much lower leakage (less than 0.25 afyf) occurs in the tributary segments and canals. This distribution of stream leakage through the model domain aligns with the streamflow analysis described in the “Surface Water and Watershed” section of this report, which showed substantial leakage in several segments of the Salinas River between Salinas River near Bradley (USGS 11150500), Salinas River near Soledad (USGS 11151700), and the gage farthest downstream, Salinas River near Spreckels (USGS 11152500).

Groundwater pumpage for M & I is specified based on reported and estimated values. Domestic pumpage from individual landowners is not explicitly estimated. The M & I pumpage estimates before 1995 are based on population and include potential domestic groundwater pumpage. After 1995, domestic pumpage is assumed to be less than 10 percent of M & I pumpage. For example, domestic pumpage only meets water demands for approximately 31,000 of the 402,000 people estimated to live in Monterey County in the year 2000 (Henson and others, 2023). Groundwater pumpage information for each M & I well was estimated for the model period before November 1994 and specified using observations for the remainder of the simulation. For the model period before November 1994, the volume of groundwater pumpage was estimated using census population data and estimated gallons per capita per day for years 1970, 1980, and 1990 (Henson and others, 2023). After November 1994, M & I pumpage was specified based on reported data. A complete description of the M & I estimation methods and pumpage data are provided by Henson and others (2023). The estimated monthly pumpage rate was divided among the wells in each area on a monthly time step and assigned to wells that were known or assumed to exist at that time.

No-flow boundaries in the integrated hydrologic models were used for the bottom of the basement hydrogeologic unit (layer 9) that represents the basement aquifer and for the lateral boundaries of the active model domain. The lateral no-flow boundaries are defined at the topographic ridges of ranges that bound the Salinas Valley and represent the contact between the low-permeability basement hydrogeologic unit and the aquifers at the edges of the groundwater basin (fig. 22A). No-flow boundaries were also specified for faults that bound parts of the foothills surrounding the Salinas Valley. The lower boundary of the model was limited to 500 ft below the top of the basement hydrogeologic unit, which is deeper than the deepest supply wells.

Lateral and vertical head-dependent flow boundaries are implemented to represent net regional groundwater flow from adjacent groundwater subbasins. Using the General Head Boundary Package of MODFLOW (Harbaugh, 2005), head-dependent flow boundaries were simulated in three locations in the model: a coastal, inland, and offshore general head boundary (GHB). The net coastal and inland GHB groundwater flow data were summed to create a category referred to as the “interbasin underflow” groundwater budget, and the offshore GHB net groundwater flow data were summed for all model layers to create a category referred to as the “seawater coastal inflow” groundwater budget. This seawater coastal inflow groundwater budget category is an analogue for seawater intrusion. Seawater intrusion is usually quantified as an area of the aquifer that has water quality degradation due to contamination by saline water, and the seawater coastal inflow budget category reflects the total volume of water that enters the landward portion of the model domain from the ocean. The hydraulic conductance for each group of GHB cells was based on the hydraulic conductivity of the aquifer sediments (described in the “Hydraulic Properties” section) and the cell geometry. Hydraulic conductance for each GHB boundary was adjusted during model parameter estimation.

For the coastal GHB, lateral head-dependent flow boundaries were specified in selected cells in layers 1, 3, 5, 7, and 8 near the northern boundary of the study area near the coast in the 180/400-Foot Aquifer and Langley Area groundwater subbasins (fig. 25). The coastal GHB represents interbasin underflow between the integrated hydrologic model domain and the Pajaro Valley groundwater subbasin to the north. For the coastal GHB, four GHBs were specified with spatially constant and time-varying boundary heads obtained from nearby monitoring well groundwater levels (figs. 26A–D; Henson and others, 2023). The hydraulic conductance for the group of cells associated with each monitoring well in each layer is considered constant.

For the inland GHB, lateral head-dependent flow boundaries also were specified in layer 7 near the Salinas River at the southern boundary of the integrated hydrologic model domain. This inland GHB represents interbasin underflow between the integrated hydrologic model domain and the Paso Robles Area groundwater subbasin to the south. For the inland GHB (fig. 25), a GHB is defined for seven cells in layer 7 perpendicular to the first river reach. The inland GHB time-series value was a constant 400 ft through time, estimated as the mean hydraulic head near the river from the Paso Robles hydrologic model of the upper Salinas Valley (GEOSCIENCE Support Services, Inc., 2016; Henson and others, 2023).

For the offshore GHB, vertical head-dependent flow boundaries were specified for the exposed offshore geologic units to estimate seawater coastal inflow in the aquifers along the coast. The offshore GHB is used to estimate coastal inflow exchanges among the onshore and offshore areas of each hydrogeologic unit. The integrated hydrologic models do not explicitly simulate the density of seawater for the simulation of exchanges between the onshore and offshore areas of the model domain. For each model cell exposed on the seafloor, a vertical GHB boundary was applied. A monthly time series of sea-level variation was estimated using the mean monthly sea level elevation data (fig. 26E; National Oceanic and Atmospheric Administration, 2019) in San Francisco Bay before 1974 (identifier 9414290) and in Monterey Bay after 1974 (station 9413450). An equivalent freshwater hydraulic head was used to account for the density of seawater in the offshore hydraulic heads using the methods described by Motz and Sedighi (2009). The elevation value was based on the North American Vertical Datum of 1988. This approximation of equivalent freshwater hydraulic head was determined to yield accurate values for hydraulic heads in a coastal aquifer based on experiments in three-dimensional groundwater flow models. The GHB along the coastline was computed using equation 1:hfw =ρρfh−ρ−ρfρf Zi(1)where

The equivalent freshwater head used to compute the GHB value was variable for each cell with an offshore GHB. The equivalent GHB value depends on the height of the column of water above the center of cell Z_i where the GHB was applied. If the top of the cell where the GHB was applied was closer to the ocean surface (for example, near the shoreline), the GHB value was lower. If the top of the cell where the GHB was applied was deep under the sea level (for example, far from shore), the GHB value was higher. The range of GHB values over all offshore GHBs ranged from 2.5 to 97.4 ft.

Irrigation, municipal, and industrial wells are simulated as multiple-aquifer wells that can extract water from more than one hydrogeologic unit (fig. 24A). All single and multiple-aquifer wells were simulated by the multi-node well package (MNW2; Konikow and others, 2009). The MNW2 simulates two processes: (1) the produced groundwater from single or multiple aquifers during pumping and (2) the flow of water between aquifers via boreholes when multiple aquifers are connected to the same well and have different hydraulic heads. For many of the wells, data for each well describing the operational history, open screen intervals, construction information, and radius were estimated where they were incomplete. This section summarizes the methods used for estimating the missing well data to create a complete set of well-related data and properties needed for model construction. The resulting complete set of well information was published by Henson and others (2023).

Estimated operational history information (drill dates, active pumping periods, and destruction dates, if applicable) was used to construct the monthly pumping time series for each well. The number of active wells for any given stress period varied through time based on reported drill dates and destruction dates. Well construction and destruction dates were used where available to specify when wells are active in the simulation. Specifying wells with undefined construction information as active for the simulation is warranted based on historical reports of extensive agricultural groundwater development (California Department of Water Resources, 1968). Available open-screen interval data were used to identify the model layers from which water was withdrawn, with the assumption that wells fully penetrate each layer they pump from. If a well contained multiple open-screen intervals, all layers from the top of the uppermost open interval to the bottom of the lowermost interval are assumed to be completely screened and fully penetrating in those model layers. When well-screen intervals span multiple model layers, the well is simulated as a multiple-aquifer well, allowing pumping to be dynamically distributed along with intra-wellbore flow between all corresponding layers. Thus, pumpage for each well was allocated dynamically to individual model layers based on the available construction information to determine which layers contribute to potential pumpage or intra-wellbore flow within a well. There is substantial uncertainty about the hydrogeologic unit to which each well is connected, which can contribute to uncertainty in the magnitude of simulated well drawdowns, distribution of pumpage, and intra-borehole exchanges of water among layers. To mitigate this uncertainty to the extent possible, wells with missing construction information were assigned open screen intervals based on their WBS, construction date, depth, and the nearby MCWRA-estimated 500 mg/L chloride concentration contour (fig. 24A; Henson and others, 2023; Monterey County Water Resources Agency, 2023). Where available, the well pumping capacity and casing diameter of each well was obtained from MCWRA. Missing well pumping capacity and the casing diameter of each MNW2 well was estimated based on properties of similar wells from the MCWRA well database (Henson and others, 2023). If the casing diameter was not available for a well, the well was either assigned a casing diameter from a nearby well or assumed to have the median casing diameter of all wells with the same well-use category. The diameter of the well skin, representing the region of disturbed aquifer material surrounding the well casing due to drilling activities, was assumed based on the drill date and casing diameter of the well.

Groundwater discharge to the land surface was simulated using the MODFLOW Drain Return Flow package (DRT; Banta, 2000; Hanson and others, 2014a). This drainage boundary condition was applied to each of the uppermost model cells in the integrated hydrologic models, with the drain elevation set to the land surface elevation plus 1 foot inside the riparian area and cells containing streams and the land surface elevation everywhere else. This model setup allows routing of groundwater discharge to streams when using the DRT. The DRT applies a drain boundary condition to compute groundwater-level rise above the land surface and routes this drain flow to adjacent streams. This drain flow becomes streamflow that is managed by the SFR2.

To better quantify the magnitude and timing of groundwater discharge to streams in different locations of the study area, the model cells specified in DRT were grouped into three different budget group categories in the model output. Each cell in the riparian WBS was defined as a riparian drain and assigned to a riparian drainage groundwater budget category in the analysis. Each cell coincident with or adjacent to a stream cell that is not a part of the riparian WBS was defined as a tributary drain and assigned to a tributary drainage groundwater outflow budget category. Every other cell in the model domain was defined as a surface drain and assigned to a surface drainage groundwater outflow budget category. These budget categories are used to evaluate the role of different locations of groundwater discharge to streams in hydrologic budgets throughout the basin.

The Layer Property Flow package (Harbaugh, 2005) was used to define storage and hydraulic conductivity properties in each of the aquifers represented in the integrated hydrologic models. The Layer Property Flow package, along with the Parameter Value (Harbaugh, 2005) and Multiplier (Harbaugh, 2005) packages, was used to calculate and specify the aquifer storage and hydraulic conductivity parameters. Lateral and vertical variations in sediment texture affect the direction and rate of groundwater flow by constraining the magnitude and distribution of aquifer-system permeability, porosity, and storativity. The hydrogeologic units defined in the SVGF (fig. 22A; table 2; Sweetkind, 2023) were used as surrogates to define the vertical and lateral hydraulic conductivity and storage property distributions within the integrated hydrologic models. Each hydrogeologic unit can be characterized by variations in hydraulic properties, which are based on the textural distribution of coarse- and fine-grained sediments in zones that represent subregions in which sediments accumulated in particular depositional environments, referred to as “facies” (Sweetkind, 2017). For this study, facies generally are represented by the Salinas Valley groundwater subbasins (fig. 11) and are implemented as parameter zones using the ZONE package (Harbaugh, 2005). Hydraulic properties for the three model layers of the most recent SVIGSM (Montgomery Watson, 1997) were used as initial values for the integrated hydrologic models (table 4). Hydraulic properties were then estimated as separate parameters that were adjusted by groundwater subbasin using regional-scale factors.

The parameter estimation includes observation data for the period from October 1967 to December 2014. Observations representing groundwater conditions include groundwater levels, temporal changes in groundwater levels (drawdowns), vertical gradients in groundwater levels between units, published changes in groundwater level contours, aggregated agricultural groundwater pumpage by WBS, and streamflow differences among upstream and downstream gages. Observations representing surface-water conditions include surface-water flows along the Salinas River and in rivers within the integrated hydrologic model domain at USGS streamgaging stations, canals/laterals, and drains where data were available, as well as flow-differences between gages (fig. 20; Henson and Culling, 2025).

During parameter estimation, observations were weighted. The sensitivity of a parameter is also dependent on the observation weights. Observation weights are used for a variety of purposes, including accounting for differences in measurement units and quantification of measurement error; they are sometimes imposed to help distribute the importance of observations across many different types of observations (for example, to remove the effects of spatial or temporal clustering of measurements or to emphasize areas where a model will be used to make predictions). The weighting of observations helps to determine how the contribution to the objective function is distributed among the various types of observations. This weighting procedure also helps ensure that the parameter estimation only considers observations that the model can reproduce given data and model limitations, including observations that the model cannot reproduce that can lead to parameter compensation and model structural error (White and others, 2021). Therefore, some observations were weighted near zero or scaled to focus parameter estimation on observations of interest. For example, observations for the first 2 years of the simulation had reduced weights to allow for the parameter estimation to focus on the period after the assumed 2-year model spin-up. There were many parameter estimation iterations that alternated between trial and error and PEST-HP. Through this process, observations were regularly reweighted so that the objective function was updated and the contribution of error from each observation group was equalized.

Each observation group type (surface-water, groundwater, and agricultural pumpage) was weighted to represent about one-third of the total error; subgroups (for example, low flows) within each observation group were weighted differently. Within the stream observation group, selected streamflow observations in the stream network were given relatively more importance to the parameter estimation if they affected pumpage and groundwater levels, specifically those within the main channel of the Salinas River, diversion locations, and places where reservoir operational target flows were defined. The low flows and all other flows were equally weighted, ensuring parameters that control high and low flows affected model parameter estimation. High flows are driven primarily by precipitation runoff, whereas low flows are typically driven by irrigation runoff and groundwater and surface-water exchange. Many of the groundwater level observations in the model were affected by pumping or occurred in wells that were used for groundwater supply. The observations associated with these agricultural supply wells were weighted lower than other wells, and the observations that represented annual minima of groundwater levels were assigned zero weights. The observations associated with annual mean groundwater levels and drawdowns were increased to twice the weight of other groundwater level and drawdown observations. The monthly and quarterly mean agricultural pumpage observations were weighted based on their fraction of total agricultural pumpage reported in the Groundwater Extraction Management System. These weights were doubled for annual average agricultural pumpage observations.

The number of adjustable parameters changed during parameter estimation. A total of 311 parameters were created initially to facilitate model parameter estimation, and after initial global sensitivity and parameter estimation, about 40 parameters were determined to be relatively sensitive and were included in the computer-assisted and trial-and-error parameter estimation process (fig. 31; table 6). All other parameters were less sensitive than the least sensitive parameter shown on figure 31. The full enumeration of parameters is provided by Henson and Culling (2025). These parameters included aquifer conductivity, aquifer storage, climate scale factors, drain conductance, runoff, and stream conductance parameters. As discussed, hydraulic properties were initially assigned values based on previous modeling studies, then adjusted during model parameter estimation. Model parameters were adjusted within ranges of reasonable values to best-fit historical hydrologic conditions (history matching) and observations in the groundwater, surface-water network, and landscape.

Computer-assisted parameter-estimation techniques using PEST-HP (Doherty and Hunt, 2010; Doherty, 2024) primarily were used to estimate selected model parameters and related sensitivities, but additional insight was provided by trial-and-error analyses. PEST-HP computes the sensitivity of simulated values to changes in model parameters at the locations of measurements. Sensitive parameters were identified (fig. 31), which helped guide which parameters were adjusted during the parameter estimation process (Hill and others, 2000). The measure of parameter sensitivity used to remove insensitive parameters was composite scaled sensitivities (Hill and Tiedeman, 2007). Composite scaled sensitivities indicate the information content of all observations for the estimation of a parameter and provide insight into parameter importance and sensitivity.

The most sensitive parameter was the power law exponent for distributing the texture distribution within layer 7 in the Pressure and Forebay analysis regions. This aquifer is closest to the surface in the Upper Valley analysis region (fig. 22A), so the textural distribution where layer 7 is deeper may be important for transmission of recharge in the Upper Valley analysis region toward the coast in deeper aquifers. The next set of important parameters were the climate scale factors that represent seasonal stress adjustments for Kc values. An additional sensitive parameter was the streambed conductance in the Upper Valley analysis region between the USGS 11150500 and USGS 11151700 gages, where substantial stream leakage has been documented (fig. 21; Monterey County Water Resources Agency, 1995). Similarly, the sensitivity of the streambed conductance between the USGS 11151700 and USGS 11152300 gages was substantial for the same reasons. Other sensitive parameters include hydraulic conductivity (kc_ly1, l1hkpressure), specific storage (l1ss), and specific yield (sy_ly1c) of the shallow aquifer. The sensitivity of the parameters that govern (1) recharge and storage in the shallow aquifer, (2) streambed conductance, and (3) climate factors that affect agricultural demand simulation are to be expected given that the primary recharge to the aquifer from precipitation and agricultural return flows is affected by the storage and hydraulic conductivity of layer 1 and the Salinas River, and agricultural demands are such a substantial portion of the total water demand.

Correlation plots show good correspondence among observed and simulated equivalent streamflow values (fig. 32A) across the range of highly variable streamflows within the Salinas Valley. There is more spread around the 1:1 line for high peak streamflows. Figures 32B–G show observed and simulated equivalent flows at selected observation gages of the Salinas River and its major tributary, Arroyo Seco. Streamflow differences for four pairs of gages are shown on figures 32H–K. These hydrographs illustrate a reasonable match of streamflows through time within the region from the uppermost to the lowermost gage in the system. Monthly peak streamflows are well characterized, but low flows are commonly overestimated. This effect of low-flow overestimation increases downstream along the Salinas River due to accumulation of simulation error.

Streamflow hydrographs indicate a good visual fit of monthly observed values and their simulated equivalents (figs. 32B–K) such that the absolute value of all mean residuals is less than or equal to 81 ft³/s, RMSE is less than or equal to 437 ft³/s, scaled RMSE is less than or equal to 5 percent, and NMSE values are greater than or equal to 0.87 (table 7). Streamflow difference plots (figs. 32H–K) show a good visual fit of observed values and their simulated equivalents, such that the absolute value of all mean residuals is less than or equal to 48 ft³/s, RMSE is less than or equal to 238 ft³/s, and scaled RMSE is less than or equal to 8 percent (table 7). The NMSE values for streamflow differences are lower (less than 0.5) because of the effect of extreme values that are not simulated well by the model. Stream differences are challenging to match because errors propagate downstream, leading to even larger differences in USGS gages lower in the Salinas Valley. Also, stream gains and losses can be affected by localized runoff or withdrawal from shallow wells adjacent to the river that may not be represented by the model.

Diversions are well matched where inflow data to the channel are sufficiently accurate. The historical model has a good visual fit for diversions at Clark Colony (fig. 32M); the mean residual is 3 ft³/s, RMSE is 8 units, scaled RMSE is 13 percent, and NMSE is 0.52. The higher scaled RMSE and lower NSME are because Clark Colony diversions were estimated, and the estimated diversion values are commonly higher than the observed flow in the upgradient channel at USGS 11152000. Nonetheless, the low mean residual and RMSE for Clark Colony diversions imply that the diversion is reasonably represented even though the estimated diversions were typically higher than the available flow. Diversions at the SRDF (fig. 32L) are well matched, with a mean residual and RMSE less than 1 ft³/s, scaled RMSE of about 2 percent, and NMSE equal to 0.99. The well-matched diversions at SRDF indicate that the FMP is computing the full amount of the surface diversions used to meet demands at CSIP. Overall, these results indicate that the surface-water flow system is well represented in the integrated hydrologic model.

Groundwater hydrographs that show both simulated and observed heads for selected wells help to illustrate the match of groundwater levels throughout the model subareas of the historical model. An analysis well subset of hydrographs for 10 wells (fig. 33) was selected as representative due to the wells’ long period of record, regular measurements, and representation of the important aquifers in each region. Table 9 summarizes the hydrogeologic units and model layers for each well in the analysis well subset. Analysis wells were selected in each analysis region to evaluate the 180-Foot Aquifer, 400-Foot Aquifer, and deeper hydrogeologic units (Paso Robles Formation, Purisima Formation, and basement). The four wells selected in the Pressure analysis region represent hydrogeologic units of the 180-Foot Aquifer (CSI239 and ZPN1529) and the 400-Foot Aquifer (ZPN441 and CSI239). The hydrographs for the East Side-Langley analysis region represent the composite of the 180-Foot Aquifer, 400-Foot Aquifer, and Paso Robles Formation hydrogeologic units (ZES871 and ZES1572). The hydrographs for the Forebay analysis region represent the 400-Foot Aquifer hydrogeologic unit (ZFS1001) and the composite Paso Robles Formation and Purisima Formation hydrogeologic units (ZNE1267). The hydrographs for the Upper Valley analysis region represent the Purisima Formation hydrogeologic unit (ZSE355 and ZSE733). The hydrographs for all observation wells can be obtained in the head observation output file in the model archive (Henson and Culling, 2025).

Correlation plots (fig. 33A) and examples of hydrographs from the East Side-Langley, Pressure, Forebay, and Upper Valley analysis regions (figs. 33B–K) are used to illustrate the temporal fit of groundwater-level observations and their simulated equivalents. The observed and simulated equivalent groundwater-level correlation is good across the range of groundwater levels (fig. 33A). The monthly to interannual fluctuations in observed groundwater levels indicate the effect of groundwater pumping, followed by climate variability (figs. 33B–K) and streamflow infiltration (for example, wells CSI239, ZPN1529, and ZSE733 near the Salinas River). Even though there are places where the groundwater levels are over- or underpredicted, the change in groundwater levels from the first measurement in each observation well (drawdown) have low mean residuals (tables 8, 9).

Groundwater levels and drawdowns generally show good agreement between observed and simulated equivalent values. There are some areas of the Forebay analysis region where groundwater levels are overpredicted by about 10 ft (fig. 33I), and in the Upper Valley analysis region, they are overpredicted by 20 ft (fig. 33K). The absolute value of the mean residual for all drawdown observations is less than 1 ft, with an RMSE of 15 ft and a scaled RMSE of 6 percent. There are drawdowns in the Pressure analysis region that are underpredicted by approximately 19 ft (fig. 33C; table 8) or 21 ft (fig. 33D; table 8) or overpredicted by 6 ft (fig. 33B; table 8) or less than 1 ft (fig. 33E; table 8). This under- or overprediction of mean drawdowns is affected by the capability of the model to simulate the seasonal oscillations in these observation wells where pumping is occurring at unknown rates. The magnitude of seasonal oscillations in some wells were not matched everywhere. For example, seasonal oscillations in groundwater levels in the East Side-Langley analysis region are commonly 40 ft or more (fig. 33F). This effect results in the spread of simulated and observed groundwater levels across the 1:1 correlation line in the correlation plot (fig. 33A). This poor representation of seasonal oscillations is likely because 340 of the 439 observation wells are also agricultural supply wells, and pumping rates in the models are not simulated using time series from individual well owners. The differences between simulated and observed seasonal oscillations are caused by using subregional distributed pumping rates to replicate the pumpage rates applied to each well. Observations that were assumed to be affected by pumping were delineated to help focus the parameter estimation and analysis. Pumping occurs year-round, so the assumptions for pumping effect did not capture all wells where large oscillations due to pumping can be observed. The effects of pumping and oscillations are still observed in some wells, especially in the Pressure and East Side-Langley analysis regions (fig. 33A). The effect of this difference between observed and simulated can be observed in hydrographs that have multiple measurements per year (figs. 33B, 33C, 33G, and 33J). Many of these wells are in the Pressure and East Side-Langley analysis regions, where mean residual drawdowns are low (from 1 ft to −1 ft, respectively), and RMSE values (15 ft and 18 ft, respectively; table 9) are higher in these regions. Nonetheless, the scaled RMSE for the Pressure and East Side-Langley analysis regions are 8 and 7 percent, respectively. Mean residual drawdowns in the Forebay analysis region were −3 ft with an RMSE of 10 ft and scaled RMSE of 9 percent. Mean residual drawdowns in the Upper Valley analysis region were 1 ft, with a RMSE of 7 ft and a scaled RMSE of 13 percent. The scaled RMSE is higher here because there are fewer wells in this region that have known properties added to the integrated hydrologic models. Thus, all observation wells in the Upper Valley analysis region were specified as agricultural supply wells, with pumping rates specified on analysis region demands that are likely to be different from pumping time series applied at those wells. The specification of simultaneous pumping and observation wells was unavoidable because there are fewer reported supply wells in the Upper Valley analysis region (figs. 24A, 24B) among which to distribute estimated agricultural demands. Defining concurrent pumping and observation wells resulted in oscillations in some wells that differ from measured values.

Model results in early time periods are sensitive to estimates of initial conditions. Although the rates of decline and the elevations are like those in historical records, some of the temporal changes are not reflected in the simulated values. The magnitude of substantial drawdowns and subsequent recovery in the dry period from WY 1984–94 is not captured in some wells (for example, figs. 33F–H), which could be due to an interaction between the time of the groundwater level measurement and residual pumping effects of the observation well and adjacent wells. Other places where simulated and observed groundwater levels diverge could be a function of changes in actual land use or irrigation practices that are not well represented in the available land use data, variability in Kc values that estimate crop demands, and divergence between the rates of actual groundwater pumpage at a specific well and the wells simulating groundwater pumpage. Although there are some places where seasonal oscillations are not well matched, or drawdowns are well matched and groundwater levels are elevated or depressed, overall historical model results show reasonable correspondence among simulated and observed groundwater levels throughout the integrated hydrologic model domain.

Groundwater observation well mean residuals also show reasonable correspondence across all analysis regions (fig. 34); 77 percent of observation wells have mean residual drawdowns from −30 to 30 ft, and 56 percent of wells have mean residual drawdowns between −15 and 15 ft. These residuals are affected substantially by the aforementioned challenges in representing seasonal oscillations in observation wells that are also pumping wells. In the Langley Area groundwater subbasin of the East Side-Langley analysis region, some water levels are not well matched, but several faults in the groundwater basin here cause water level offsets of more than 100 ft in adjacent wells. There are three isolated wells in this area that have absolute mean drawdown residuals greater than 50 ft (fig. 34), indicating that the model residual bias is limited spatially.

To allow for a spatial comparison of the simulated historical model values to observed data, groundwater-level maps were developed for fall of 1994, 2003, and 2011 (figs. 35A–F). The observed data in these plots are contours generated by MCWRA for a composite of shallow aquifers (depth less than 201 ft) and deep aquifers (depth greater than 201 ft and less than 420 ft). The shallow contours from the historical model are approximated using the December simulated equivalent groundwater levels in the 180-Foot Aquifer (model layer 3). The deep contours from the historical model are approximated by the December simulated equivalent groundwater levels in the 400-Foot Aquifer (model layer 5). These maps were used during historical model parameter estimation to provide additional information on the effects of internal flow boundaries along faults and to help adjust selected model hydraulic properties, such as vertical hydraulic conductivities.

The MCWRA and simulated groundwater level contour maps show good correspondence among the shallow and deep aquifers in 1994, 2003, and 2011 (fig. 35). The simulated groundwater levels have similar areas of over- and underprediction. The historical model data and contours both show that water level declines are concentrated in the Pressure and East Side-Langley analysis regions and increase in magnitude toward the City of Salinas. Additional declines in groundwater levels are observed in the East Side-Langley analysis region over time. Simulated groundwater levels in 1994, 2003, and 2011 overestimate the MCWRA contours in the Upper Valley analysis region of the historical model by about 20–30 ft, where additional refinement of aquifer properties, land use, or recharge may be required.

The reported monthly agricultural pumpage was aggregated by WBS, resulting in 3,630 observations spanning the period from November 1994 through December 2014 that were compared to FMP simulated equivalent agricultural pumpage values during calibration. Before November 1994 when monthly agricultural pumpage observations were not available, simulated equivalent annual agricultural pumpage was compared to published long-term estimates. The historical model matches the total reported annual agricultural pumpage from November 1994 through September 2018 for Salinas Valley within 99 percent and has general agreement among monthly reported and simulated agricultural pumpage throughout the model domain (figs. 36A–E). Simulated equivalent annual agricultural pumpage varies from year to year with an average of approximately 470 TAFY for the period between 1970 and 1994. This value is consistent with prior modeling efforts (Montgomery Watson, 1997) and MCWRA reports (Monterey County Water Resources Agency, 1995). After November 1994, annual reported agricultural pumpage has varied from 380 TAFY in 2001 to as much as 529 TAFY in 1997, with an average of 439 TAFY (fig. 36B). The annual mean residual (annual reported pumpage minus simulated equivalent annual pumpage) for the history matching period (WY 1995–2014) was −4.3 TAFY, which is approximately 1 percent of the mean annual pumpage. There is reasonable correspondence among monthly simulated equivalent and reported agricultural pumpage (fig. 36A). The monthly reported and simulated equivalent agricultural pumpage for each analysis region for the simulation with reported observations is shown on figures 36C–G. The absolute value of all monthly mean residuals for the integrated hydrologic model domain and the analysis regions is less than or equal to 181 acre-feet, with RMSE less than 1,350 acre-feet and all scaled RMSE less than 9 percent (table 10). Approximately 73 percent of all simulated equivalent monthly agricultural pumping is within 700 acre-feet of the reported monthly values. The close correspondence of reported and simulated equivalent annual total pumpage indicates that monthly errors tend to cancel themselves out over the growing season. There is reasonable correspondence between simulated and observed pumpage among areas in analysis regions exclusively irrigated by groundwater (East Side-Langley and Upper Valley analysis regions; figs. 36D and 36F, respectively) and among areas that have irrigation from surface-water diversions and recycled water deliveries (Pressure and Forebay analysis regions; figs. 36C and 36E, respectively).

No water storage is considered in the landscape budget. Streams exist within the landscape; however, stream inflows and outflows are counted in the surface-water budget. Groundwater entering and exiting streams (gains and losses) also is not accounted for in the landscape budget (to avoid double counting with the groundwater budget). Runoff here is defined as overland runoff that goes into streams and thus out of the reference landscape area. Pumpage and surface-water diversions are taken from groundwater and streams (outside of the container) and flow into the container. Deep percolation flows out of the container and into groundwater. Instead, the landscape budget represents the flows into and out of the landscape throughout the historical model. Inflows to the landscape are precipitation, shallow groundwater, agricultural pumpage, surface-water diversions, and recycled water. Outflows from the landscape are evapotranspiration of precipitation, groundwater, irrigation, deep percolation to groundwater, and runoff to streams. Irrigation TDR to meet agricultural demands is represented by the sum of agricultural pumpage, recycled water deliveries, and surface-water deliveries on figure 37. The TDR analysis allows for evaluation of the landscape components that support deliveries to meet agricultural demands.

The temporal distribution of inflows and outflows to the landscape and surface-water systems indicates a strong climatic effect, with higher values overall in wet periods. Precipitation generally aligns with reported data. However, 2013 shows lower precipitation than expected given that it was a normal climate year, which is likely an anomaly in the estimated climate data. The watershed inflows and other surface-water flows align with historical records and are consistent with periods before and after 2013. The climate year type that is used for reservoir operations (based on surface-water flow percentiles as described in the “Climate” subsection of the “Description of Study Area” section) is not always aligned with observed climate within the basin, which is especially true for normal and dry climate year types; for example, 1984 (normal) and 1988 (dry) have similar precipitation magnitudes for the integrated hydrologic model domain (fig. 37A). There is also subregional inconsistency where the climate year type does not align with analysis region conditions; for example, 1984 precipitation in the Pressure analysis region is less than the 1988 precipitation (fig. 37C). Generally, runoff is higher in normal and wet years than dry years. However, substantial runoff can occur in dry years from inefficient irrigation. Analysis of landscape budgets among all analysis regions show similar trends within the basin (figs. 37B–F), including a strong relationship between climate and TDR, with increased pumpage during dry periods.

The effect of climate on TDR is evident in the simulation results. Although some variability does occur among analysis regions, the TDR commonly exceeds precipitation over the basin, with 1988–90 and 2007 having annual TDR values more than 1.5 times the precipitation those years. The TDR is greater than 1.5 times precipitation for 15 years in the Pressure analysis region (fig. 37C), 7 years in the East Side-Langley analysis region (fig. 37D), 33 years in the Forebay analysis region (fig. 37E), and 18 years in the Upper Valley analysis region (fig. 37F). This spatial variability in the ratio of precipitation to TDR in each analysis region further illustrates the water management challenges in delivering water to meet demands throughout the basin (California Department of Public Works, 1946). Because local TDR commonly exceeds precipitation, groundwater is used extensively in the basin to meet demands. More instances of years with TDR greater than 1.5 times the precipitation in the analysis regions compared to the integrated hydrologic model domain suggest an interconnected water supply with water demands supported by flows from adjacent areas. Water demands in each analysis region are supported by recharge from precipitation in the uplands outside of the analysis regions (other areas, fig. 15) and potentially underflow from the Riparian analysis region and groundwater subbasins in adjacent analysis regions. This interbasin underflow is evaluated using the groundwater budgets in the next section.

Offsetting groundwater use, surface-water deliveries (recycled water deliveries and surface-water diversions) have increased steadily since 1998, reaching nearly 4 percent of the TDR by the end of the simulation period, with a maximum of as much as 23,000 acre-feet in 2013 (fig. 37A). Especially in the Pressure analysis region, the TDR supplied by groundwater has been substantially reduced since 1998 through increases in recycled water deliveries and the development of surface-water diversions supported by the Salinas Valley Reclamation Project at the SRDF. These new surface-water supplies provide a substantial portion of the TDR. The surface-water supplies have supported decreases in observed pumpage in the Pressure analysis region.

Some items in the groundwater budget bar graphs and summary tables are presented as net values where the sum of inflow and outflow for each budget component are added to compute a net gain or loss. Positive values equal a gain in flow, and negative values equal a loss in flow. These net budget components include recharge, stream leakage, interbasin underflow, subbasin underflow, riparian underflow, seawater coastal inflow, and aquifer storage change. Recharge is computed as the difference between direct groundwater uptake by vegetation and the amount of water that percolates into the subsurface. Recharge is generally positive in the groundwater budget but can be negative if groundwater evapotranspiration is greater than the amount of percolation. Stream leakage is the amount of water that infiltrates in all stream segments and is comprised of Salinas River infiltration and other channel infiltration. Interbasin underflow is the onshore flow into the model domain from adjacent groundwater basins outside of the active model domain that is simulated by the inland GHB (fig. 25). Subbasin underflow is the regional groundwater flow to the analysis region within the active model domain. Riparian underflow is the regional groundwater flow to an analysis region from the riparian area. Seawater coastal inflow is landward flow from the ocean simulated at the coastal GHB (fig. 25). Aquifer storage change is the difference between all outflows and inflows. The sign convention for aquifer storage can lead to positive and negative values that are counterintuitive; they are explained in the “Groundwater Budget Bar Plots” section.

The components of outflows in groundwater budgets include surface drainage, riparian drainage, tributary drainage, surface drainage, M & I pumpage, and agricultural pumpage. Surface drainage occurs when groundwater is above the land surface in a model cell that is not associated with a stream. Riparian drainage occurs when groundwater is above the land surface in a model cell in the riparian area. Tributary drainage occurs when groundwater is above the land surface in a model cell that contains a stream segment outside of the Salinas River. Municipal and industrial pumpage is specified using furnished and estimated data (Henson and others, 2023). Agricultural pumpage is simulated to meet demands by FMP. Important groundwater budget components for water managers include total recharge, stream recharge (recharge from the surface-water drainage network outside of the Salinas River), riparian underflow to each analysis region sourced from recharge in the Salinas River, total pumpage, pumpage-recharge fraction (the fraction of analysis region pumping to recharge), aquifer storage change, and groundwater depletion.

Groundwater budget bar plots (fig. 38) show the volumes of each groundwater budget component. In these plots, aquifer storage change is negative if there are more inflows than outflows. This sign convention for the groundwater budget bar plots ensures that the sum of inflows equals the sum of outflows. In the cumulative storage change line plots for each groundwater budget bar plot and in the groundwater budget summary tables (tables 11–16), the sign for aquifer storage change and cumulative storage change are reversed (multiplied by −1) so that the changes in storage are shown as expected, with positive values indicating an increase in storage and negative values indicating a decrease in storage. The cumulative storage curves plotted on each groundwater budget bar plot show the actual change in groundwater volume starting in 1970.

Groundwater budget summary tables provide quantities of interest averaged over analysis periods and for selected high and low precipitation years. Comparison of quantities of interest over analysis periods and for selected high and low precipitation years allows the evaluation of these quantities under different historical conditions. The quantities of interest distill information from the budgets to inform decision making and sustainability efforts. These quantities of interest include net average groundwater inflows and outflows, as described in the “Groundwater Budget Components” section, and summarized values, such as local stream recharge, total recharge, total pumpage, pumpage-recharge fraction, storage loss, and groundwater depletion. Analyses of recharge, pumping, and groundwater depletion are supported by these summarized quantities of interest.

The local stream recharge is the stream leakage from all other surface-water drainage features except the Salinas River minus the sum of riparian and tributary drainage. Total recharge is computed as the sum of components of groundwater recharge, consisting of the recharge and stream recharge. Total pumpage is the sum of M & I and agricultural pumpage. Total recharge, total pumpage, and their ratio (pumpage-recharge fraction) provide information about drivers of aquifer storage change and local groundwater sustainability. Storage loss is the absolute value of the aquifer storage change groundwater budget component if aquifer storage change is negative. Groundwater depletion is the sum of storage loss and the seawater coastal inflow groundwater budget component. Groundwater depletion is an important quantity for sustainability analyses and to quantify the effect of water management strategies on the undesirable effects of unsustainable groundwater use, such as seawater coastal inflow and storage loss.

The groundwater budget flow charts show the average annual budget components and summarized values for the integrated hydrologic model domain and each of the five analysis regions. Each flow chart has defined inflows and outflows and provides a concise view of the groundwater budgets. The groundwater budget terms and summarized values are represented using boxes with arrows that indicate the direction of flow (in or out). This simple view of the budget facilitates an easy comparison to budgets evaluated in prior Salinas Valley analyses.

The groundwater budget bar plots for the integrated hydrologic model domain show pumpage and cycles of wet and dry years with cumulative storage change increasing in period A, decreasing during the dry period (period B), recovering in period C, and declining again in the most recent dry period (period D; fig. 38A). Groundwater budget analysis for the integrated hydrologic model domain (table 11) shows that average aquifer storage changes over the four analysis periods range from 139 TAFY (out) to 158 TAFY (in). There is substantial variability within the high and low precipitation end-member years before and after the SRDF, with pre-SRDF values ranging from 451 TAFY (out) in 1989 to 214 TAFY (in) in 1983, and post-SRDF values ranging from 275 TAFY (out) in 2013 to 112 TAFY (in) in 2013. The average storage losses were high during dry periods—139 TAFY (out) in analysis period B and 83 TAFY (out) in analysis period D. The driest year in period B (1989) had a substantial storage depletion of 469 TAFY. The driest year in period D (2013) had substantial but lower storage depletion of 275 TAFY. The effects of this depletion can be observed in the lowering of groundwater levels throughout the basin (figs. 33A–E). The lowest recorded measurements of groundwater levels occurred during the dry years of analysis, period B. The long-term average groundwater depletion is about 15 TAFY, comprised of mostly seawater coastal inflow. The average seawater coastal inflow varied over the analysis periods and individual years from 12 to 18 TAFY. There are slightly lower average values after the implementation of the SRDF (analysis period D) even though the conditions were dry for much of that analysis period. The magnitude of seawater coastal inflow is higher in years with higher overdraft (fig. 38A), so the cumulative effect of multiple years of overdraft may have a more substantial effect on usable storage of the aquifer than the long-term average implies.

Over the simulation analysis period, the average total recharge was 597 TAFY, average recharge was 93 TAFY, and average stream recharge was 504 TAFY (table 11; fig. 39A). Total recharge varied from year to year over the individual wet and dry analysis years, ranging from 250 TAFY in 1989 to 913 TAFY in 1983. The average total recharge for analysis periods A and B (1970–1994), 732 and 501 TAFY, respectively, are higher than previous groundwater budget tabulations that were 454 TAFY for the period 1970–92 (Monterey County Water Resources Agency, 1995; Montgomery Watson, 1997). However, the model domain in this study is about 40 percent larger than the SVIGSM from which that estimate was determined.

For the integrated hydrologic model domain, interbasin underflow is minimal and did not vary among analysis subperiods; flow to adjacent groundwater basins outside of the study area was less than 1 percent of total pumpage (5 TAFY out). The interbasin underflow is different from the 1970–1994 estimates of 38 TAFY into the SVIGSM model domain (Monterey County Water Resources Agency, 1995; Montgomery Watson, 1997). The boundaries for interbasin underflow are defined using measured GHB well data in the SVIHM (figs. 25–26), which is constructed differently from the formulation in the SVIGSM. Additionally, the SVIHM has a greater extent that may capture some regional flow paths not within the SVIGSM boundary.

Over the simulation period, average annual total pumpage was 510 TAFY (table 11; fig. 39A), with average agricultural pumpage of 457 TAFY and average M & I pumpage of about 12 percent of agricultural pumpage (53 TAFY; table 11). The average total pumpage for analysis periods A and B (1970–1994), 469 and 578 TAFY, respectively, average to 524 TAFY, which is close to the previous groundwater budget tabulations of 519 TAFY for the period from 1970 to 1994 (Montgomery Watson, 1997). Total pumpage in the most recent analysis period D (480 TAFY) is similar to the earliest period (469 TAFY) even though total harvested area has increased from approximately 300,000 to 400,000 acres over that time (fig. 7). An increase in total harvest area while maintaining a similar pumpage illustrates the effect of water conservation efforts and more efficient irrigation technology.

The average pumpage-recharge fraction is an indicator of the relationship between the magnitude of pumping relative to the magnitude of recharge. For basin wide groundwater sustainability, values should be less than 1.0. The average pumpage-recharge fraction for the simulation period is 0.9 (table 11). The maximum average pumpage-recharge fraction among all analysis periods is 1.2 during the dry years of analysis, period B. The average pumpage-recharge fraction over the entire simulation period of 0.9 suggests that a substantial amount of total recharge for the integrated hydrologic model domain is used to meet current demands on an average basis. The average pumpage-recharge fraction for dry years greater than 1 highlights the challenges of interannual and seasonal variability and meeting water demands sustainably.

Over the simulation period, analysis regions other than the Riparian analysis region (Pressure, East Side-Langley, Forebay, and Upper Valley) had similar trends across analysis subperiods A through D, showing similar responses to the climate. Riparian underflow from the Riparian analysis region contributed substantially to inflows in each period throughout three of the four other analysis regions, all but the East Side-Langley analysis region (figs. 38C–F; tables 12–15). Generally, the sources of inflows to analysis regions in order of decreasing magnitude are riparian underflow (groundwater flow between the Riparian analysis region and other analysis regions), recharge, and local stream recharge (locally sourced recharge from surface-water network drainage within the analysis region). The sources of outflows from analysis regions in order of decreasing magnitude are agricultural pumpage, M & I pumpage, and surface drainage. The average pumpage-recharge fraction in each analysis region is an indicator of the relationship between the magnitude of pumping relative to the magnitude of locally sourced recharge. Simulation period and analysis period average pumpage-recharge fractions above 1.0 indicate that water demands are met using riparian underflow from the Salinas River, subbasin underflow from adjacent analysis regions, or reductions in groundwater storage. The pumpage-recharge fraction can be a useful indicator for understanding local versus exogenous sources of groundwater supply to each analysis region.