Water Use and Management

The Salinas River watershed contains three reservoirs: Lake Nacimiento and Lake San Antonio, each with an area of about 22 mi², and Santa Margarita Lake, with an area of about 1.1 mi² (fig. 1). The reservoirs are used primarily to provide flood protection and are operated for a variety of uses that include municipal water supplies, agricultural irrigation, recreation, groundwater recharge, and protection of fish habitat. The dams impounding Lakes Nacimiento and San Antonio were constructed to control floodwaters and provide water for summer recharge of the Salinas Valley groundwater basin for urban and agricultural use (Brown and Caldwell, 2014). The dam impounding Lake Nacimiento, located in northern San Luis Obispo County approximately 20 mi from the coast, was completed in 1957 (fig. 1) and provides a maximum storage capacity of 377,900 acre-feet (acre-ft). The dam impounding Lake San Antonio, located in southern Monterey County about 16 mi northwest of Paso Robles, was completed in 1967 and provides a maximum storage capacity of 477,000 acre-ft. The management of Lakes Nacimiento and San Antonio, including operation of the dams, is under the jurisdiction of the California Department of Water Resources, Division of Safety of Dams (Brown and Caldwell, 2014).

Climate

The climate in the Salinas Valley study area is characterized by warm, dry summers and cool, moist winters (Baillie and others, 2015). Based on an 80-year average of climate records from 1931 to 2015, from the National Climatic Data Center station at the Salinas airport (station USW00023233 at https://www.ncdc.noaa.gov/cdo-web/datatools/findstation) located on the valley floor, the average annual temperature is 57 degrees Fahrenheit (°F), and the average annual precipitation is 13 inches (in.) falling as rain primarily during the winter and early spring. The distribution of precipitation across the study area is dependent on the topography, prevailing winds, and proximity to the coastline (Daly and others, 2004). Precipitation generally increases with increasing altitude due to orographic lifting and adiabatic cooling of moist air, but also decreases with increasing distance from the coastline, such that the higher land elevations for summit locations on the east side of the Salinas Valley receive less precipitation compared to equivalent land elevations on the west side of the valley (Tinsley, 1975; Montgomery–Watson Consulting Engineers, 1994; Daly and others, 2004).

Hydrography

For reasons discussed in the “Model Layout and Discretization” section, the SVWM was divided into two sub-model areas south and north of the U.S. Geological Survey (USGS) streamgage 11150500 (gage 19, Salinas River at Bradley; in this report, streamgage is synonymous with gage; the streamgage numbering sequence does not include numbers 9, 12, 26, 33, and 37; fig. 2); the upper Salinas Valley sub-model (USVS), south of the streamgage and the lower Salinas Valley sub-model (LSVS), north of the streamgage (fig. 2; tables 1, 2). In addition to the two sub-model areas, the Salinas Valley study area was further subdivided into 10 surface-water subbasins and 28 surface-water subdrainages (figs. 2, 3; tables 1, 3). The subbasins and subdrainages were used in this study to analyze and compare characteristics and model results for different parts and tributaries of the study area. The subbasin and subdrainage areas were defined based on hydrography, including the location of tributary junctions, drainage areas, streamgages, and reservoirs. The USVS and LSVS each include 5 subbasins and 14 subdrainages.

The Estella River (EST) subbasin drains the southeastern part of the SVWM and includes four subdrainages: the Cholame Creek (CHO), the upper and lower San Juan Creek (USJ and LSJ) subdrainages, and the lower Estrella River (LER) subdrainage. The Salinas River headwaters (SRH) subbasin drains the southwestern part of the SVWM and includes the Santa Margarita Lake (SML), Salinas River near Paso Robles (SRP), and Paso Robles Creek (PRC) subdrainages. The upper Salinas River (USR) subbasin drains the south-central part of the SVWM, upstream of gage 19 and downstream of the city of Paso Robles, and includes the Huerhuero Creek (HUE), Big Sandy Creek (BSC), and Salinas River near Bradley (SNB) subdrainages. The Arroyo Seco (ARR), San Antonio River (SAN), and Nacimiento River (NAC) subbasins drain the western part of the SVWM with each subbasin including upper and lower subdrainages; the upper and lower Arroyo Seco (UAS and LAS) in the ARR, the upper and lower San Antonio River (USA and LSA) in the SAN, and upper and lower Nacimiento River (UNR and LNR) in the NAC. The middle Salinas River (MSR) subbasin includes five subdrainages in the west-central part of the SVWM downstream of gage 19 and upstream of USGS streamgage 11151700 (gage 6, Salinas River at Soledad): the Sargent–Pancho Rico Creeks (SPR), Pine Creek drainage (PCD), upper Salinas Valley (USV), Chalone–Stonewall Creek (CSC), and Monroe Creek–Salinas River (MCS) subdrainages. The San Lorenzo Creek (SLC) subdrainage is coincident with the San Lorenzo Creek (LOR) subbasin and is a major tributary to the MSR subbasin, draining the eastern side of the Salinas River valley. The lower Salinas River (LSR) subbasin includes four subdrainages in the lower Salinas Valley between gage 6 and the mouth of the Salinas River: the Limekiln Creek–Salinas River (LCS), Quail–Chualar Creek (QCC), El Toro Creek (ETC), and Salinas River outflow (SRO) subdrainages. The Monterey Coastal Basin (MCB) subbasin includes two separate areas consisting of several small drainages emptying into Monterey Bay: the Elkhorn Slough (ELK) subdrainage adjacent to the northeastern side of the SRO subdrainage and the Monterey–Seaside basin (MSB) subdrainage adjacent to the southwestern side of the SRO.

Physiography

An important characteristic of the Salinas Valley study area is the substantial amount (more than 3,000 ft) of topographic relief on both sides of the valley, particularly the central valley, and between the headwaters and mouth of the Salinas River (fig. 1; table 1). Based on the National Elevation Dataset (U.S. Geological Survey, 2023), the mean land-surface elevation of the Salinas Valley study area is 1,425 ft above the North American Vertical Datum of 1988 (NAVD 88) and ranges from a minimum of 0 ft at the mouth of the Salinas River to a maximum of 5,872 ft at the summit of Junipero Serra Peak in the headwaters of the Arroyo Seco drainage along the western boundary of the SRW (fig. 1). Relief in the UAS and LAS subdrainages and the USA subdrainage on the west side of Salinas Valley is about 5,000 ft. The relief in the LCS and SLC subdrainages on the east side of the valley is more than 4,000 ft (figs. 1, 3; table 3).

The land-surface slope is generally greatest in the headwater drainages along the western and eastern boundaries of the SRW. Calculated as rise over run, land-surface slope is less than 0.l for most of the valley floor and locations along the coastal plain, and it increases to more than 0.3 for most of the more rugged terrain in the upland areas, reaching values as high as 1.0–2.37 for steeper slopes and canyons in the uplands on both sides of the valley (fig. 4; table 1). The high relief and comparatively steep slopes for many of the tributary drainages to the Salinas River result in rapid runoff response times and accumulation of channelized streamflow during storms, often referred to as flashiness in the characteristics of streamflow.

Streamflow

The characteristics of high relief and mountainous terrain combined with a focused distribution of annual precipitation from a limited number of winter (December–March) storms result in large variations in streamflow, seasonally and between peak and mean streamflow conditions (fig. 5). Mean monthly streamflow in February ranges from about 1,200 to 1,450 cubic feet per second (ft³/s) at three streamgages along the main branch of the Salinas River (USGS streamgages 11150500 [gage 19], 11152300 [gage 10], and 11151700 [gage 6]), in the middle and LSR subbasins, compared to mean monthly flows of about 220 ft³/s and less from October to December at these locations (fig. 5A). Controlled reservoir releases from May to September cause increased mean monthly streamflow of as much as about 500 ft³/s at USGS streamgage 11150500 (gage 19); however, mean monthly streamflow is less than 200 ft³/s at USGS streamgages 11152300 (gage 10) and 11151700 (gage 6), downstream from USGS streamgage 11150500 (gage 19). Maximum monthly streamflows are approximately an order of magnitude greater than mean monthly flows along the Salinas River and minor tributaries (figs. 5B, F) and about five times greater for major tributaries (fig. 5D). Mean monthly July–October streamflow of about 15 ft³/s and less for major tributaries and 2 ft³/s and less for minor tributaries are very low compared to January–March flows of more than 200 ft³/s for major tributaries and more than 35 ft³/s for minor tributaries (fig. 5A).

An important characteristic of streamflow in the Salinas River is that mean flows are not the highest at the mouth of the river. The long-term average streamflow close to the mouth of the Salinas River to Monterey Bay, as measured at the USGS streamgage 11152500 (gage 21, fig. 1), is approximately 333 ft³/s or about 241,000 acre-feet per year (acre-ft/yr). In comparison, the average flow at USGS streamgage 11152300 (gage 10) is 377 ft³/s, the average flow at USGS streamgage 11151700 (gage 6) is 343 ft³/s, and the average flow at USGS streamgage 11150500 (gage 19) is 487 ft³/s. The differences in streamflow between the streamgages on the main branch of the Salinas River indicate a loss of streamflow likely caused by seepage through the streambed and into the underlying unsaturated zone along a section of the channel where the water table is lower than the streambed elevation.

Characteristics of streamflow in the SRW also include managed streamflow conditions downstream from the reservoirs, particularly Lake Nacimiento. In addition to controlled releases from Lake Nacimiento and Lake San Antonio to augment crop irrigation during the dry summer months, a portion of the Salinas River outflow to Monterey Bay occurs in response to controlled releases for environmental purposes to promote the threatened anadromous steelhead run in the Central Coast (Brown and Caldwell, 2014). Controlled reservoir releases are indicated by mean monthly streamflows greater than 400 ft³/s for the months of July and August at USGS streamgage 11150500 (gage 19; Salinas River near Bradley, downstream from the NAC junction; figs. 1, 5A) and to a lesser degree at USGS streamgage 11151700 (gage 6; Salinas River at Soledad; fig. 1), with mean monthly streamflows of more than 150 ft³/s during July and August (fig. 5A).

Groundwater

Groundwater flow generally follows the topography of the Salinas Valley study area, going from the mountain ranges, down to the Salinas Valley floor and then flowing toward Monterey Bay (Hamlin, 1904; Jenkins, 1943; Simpson and others, 1946; Kennedy Jenks L.L.C., 2004). Sources of groundwater recharge include percolation of precipitation, streamflow infiltration, and return flows from agricultural irrigation. The primary source of groundwater discharge in the Salinas Valley is through the pumping of wells (Hamlin, 1904; Simpson and others, 1946; Monterey County Water Resources Agency, 2006; Burton and Wright, 2018). A significant portion of runoff that originates in the uplands and contributes to streamflow in the middle and lower reaches of the Salinas River channel does not reach Monterey Bay, but rather infiltrates the riverbed and recharges the valley-fill aquifers that are subsequently pumped for irrigation or municipal use (Montgomery–Watson Consulting Engineers, 1994; Fugro West, Inc., 1995; Harding ESE, 2001; Kennedy Jenks L.L.C., 2004). Some of the infiltrated streamflow also may contribute to riparian ET.

Natural groundwater recharge to the Salinas Valley occurs during the wet winter months (December–March) as infiltrated streamflow from the Salinas River; with major inflows from the Arroyo Seco, Nacimiento, and San Antonio Rivers; as infiltrated streamflow from smaller tributaries, and as direct percolation from precipitation (Brown and Caldwell, 2014). Evidence of infiltrated streamflow is indicated by the streamflow records along the main branch of the Salinas River, with decreasing streamflow occurring at downstream streamgages. The natural groundwater recharge in the Salinas Valley occurring in response to winter precipitation is augmented during the late spring and summer months by controlled releases from Lakes Nacimiento and San Antonio.

Land Cover

Land cover can have an important effect on surface hydrology in terms of interception and retention storage and surface roughness affecting infiltration and overland flow. In addition, differences in vegetation type can result in substantial differences in ET. As defined by the 30-m resolution National Land Cover Database (NLCD) in 2011 (U.S. Geological Survey, 2014), natural vegetation constituted about 82 percent of the total land cover in the Salinas Valley study area, consisting mostly of grasslands and shrublands for most subdrainages, especially the mid to lower elevation subdrainages in the eastern and southeastern parts of the Salinas Valley watershed (figs. 3, 6; table 4). In contrast, forestlands have the highest percentage of land cover for several subdrainages along the wetter western side of the Salinas Valley watershed, including the UAS, PRC, and LNR subdrainages. The percentage of shrubland is about the same for the USVS and LSVS sub-model areas, whereas the USVS area includes a higher percentage of grassland and a lower percentage of cultivated crop areas compared to the LSVS. The coastal subdrainages, including the LSV, ELK, and MSB, contain the highest percentages of combined low, medium, and high-density developed lands (about 11–33 percent). The lowest elevation subdrainages generally contain the highest percentage of developed open space (about 17 percent). Open water comprises only about 0.5 percent of the land cover in the Salinas Valley study area.

The 2011 NLCD 30-m resolution, percentage of forest canopy cover supplements the NLCD land cover types and provides an additional measure of the density of tree cover. As with land cover type, forest canopy cover is an important characteristic affecting interception storage and ET. The percentage of canopy cover based on NLCD averages about 11 percent in the Salinas Valley study area and ranges from 0 percent throughout the study area to a maximum of 93 percent in the Monterey Coastal subdrainage (fig. 7; table 5). In general, the wetter and cooler western side of the Salinas Valley has a higher percentage of canopy cover, with greater than 20 percent average canopy cover for many subdrainages, compared with the eastern side of the valley, with less than 10 percent average canopy cover for many subdrainages. Several subdrainages in the coastal region also have comparatively higher average canopy cover resulting from higher precipitation and lower PET demand compared to inland areas.

The 2011 NLCD 30-m resolution impervious land cover dataset indicates the percentage coverage of impervious developed lands such as parking lots, roads, and rooftops. Impervious land cover can have a significant localized effect on runoff generation and streamflow in terms of increased peak flows, increased flow volumes, and an overall increase in the flashiness of runoff. Impervious land cover, based on the 2011 data, averages only 1.2 percent of the total land area within the Salinas Valley study area (fig. 7; table 5). However, in subdrainages with population centers such as the cities of Salinas, Seaside, and Paso Robles, the percentage of impervious land cover can be as high as 100 percent locally. The average percentage of impervious cover is higher for the subdrainages in the coastal region, ranging from about 6 to 18 percent impervious land, compared to most of the more remote inland subdrainages with an average impervious cover of 0.2 percent and less.

Soils

Soils properties affecting hydrologic processes include the soil storage capacity and soil hydraulic conductivity, both of which can be highly variable because of differences in soil thickness, texture, and structure. Soil texture, including grain-size distribution and structure, are important properties affecting water infiltration and percolation. The available water capacity of soils is an indicator of the potential amount of water available for plant transpiration when precipitation (or irrigation) is not limited and is dependent on soil texture and thickness. Soils with high permeability and storage capacities tend to favor recharge over runoff, depending on climate and the hydraulic conductivity of the underlying bedrock or alluvium.

Soil texture classes, defined according to particle size distribution based on the Soil Survey Geographic database (SSURGO; U.S. Department of Agriculture [USDA], 2017), range from clayey to fine soils to coarse-loamy and sandy-skeletal soils in the Salinas Valley study area (fig. 8; table 6). Fine-loamy soils are the most prevalent basinwide soil texture for the USVS and LSVS areas. On the scale of subdrainages, however, soil texture is variable, with clayey-skeletal soils being the most prevalent in 4 of the 28 subdrainages, and loamy soils being most prevalent in 4 different subdrainages. Clayey to fine soils are most prevalent in the LNR and ELK subdrainages, whereas sandy soil is the most prevalent in the MSB subdrainage, and loamy-skeletal soil is the most prevalent in the UAS subdrainage.

The available water capacity for soils in the study area, based on SSURGO data for a maximum soil thickness of 150 centimeters (cm; about 5 ft), averages 4.5 in. and varies from 0 to 16.5 in. (fig. 9). High soil storage capacities of 8.5 in. and more are generally located in the larger valley bottoms where medium to fine-grained (clayey and loamy) soils overlay unconsolidated valley-fill sediments and tend to be the thickest compared to locations on hillsides. In comparison, low available water capacities of 2.5 in. and less are dominantly located at the higher elevations of headwater areas with steep slopes and relatively thinner coarser-grained soils. Locations with very low soil storage capacities of 0.5 in. and less occur throughout the upper parts of the ARR, SAN, and NAC subdrainages, coinciding with locations having the steepest terrain within the study area and locations having clayey-skeletal, loamy-skeletal, and undefined soils. Locations with intermediate available water capacities from 4.5 to 8.5 in. include the uplands on the eastern side of the valley and the central region of the upper Salinas Valley and tend to be coincident with more intermediate slopes and finer, well-structured soils underlain by sedimentary rocks.

Generalized Surface Geology

In addition to topography, land cover, and soils, surficial geology consisting of unconsolidated deposits and consolidated rock forming the land surface is an important characteristic affecting the hydrologic system of the Salinas Valley study area. Alluvium and porous sedimentary rocks with high permeability are more conducive to recharge rather than runoff generation, whereas shales, siltstones, and igneous and metamorphic rocks with low permeability are more conducive to runoff generation.

The Salinas Valley study area includes a wide variety of different rock types, from unconsolidated alluvium to consolidated sediments and igneous, metamorphic, and volcanic rock types (fig. 10; table 7). The dominant consolidated rock type outcropping in the Salinas Valley study area is conglomerate, mostly throughout the USR watershed, followed by unconsolidated alluvium consisting of valley-fill sediments and river wash deposits. Sandstone and siltstone are present throughout the western side and east–central side of the valley, respectively. Shale and siltstone are present throughout most of the Nacimiento subdrainage, and igneous and metamorphic rocks crop out in the upland areas of the LSVS, the UAS subdrainage, and the upper parts of the SRH subbasin. Volcanic rocks are the least common consolidated rock type outcropping in the SVWM, located mostly in the uplands adjacent to the northeastern boundary of the SVWM.

Model Domain

The SVWM model domain was defined mostly by surface-water drainage divides based on hydrologic unit code 12 boundaries (HUC-12, https://water.usgs.gov/GIS/huc.html; Seaber and others, 1987), with minor modifications using National Hydrography Dataset (NHD) flowlines and flow directions based on calculated land-surface slope using 10-m (98.4-ft) resolution digital elevation data. The 4,530 mi² SVWM area includes 35 complete or partial HUC-12 areas.

The SVWM domain encompasses the entire SRW and smaller coastal drainages along the Monterey Bay coastline adjacent to the Salinas River outflow (figs. 2, 3). The SVWM consists of two connected HSPF sub-model components: the 2,540 mi² USVS and the 1,990 mi² LSVS (fig. 2) that are connected in the upper Salinas Valley at the Salinas River USGS streamgage 11150500 (gage 19, near Bradley; fig. 2; table 2). All surface-water outflows from the USVS enter the LSVS at this streamgage location.

Conceptual Model

The components of the natural hydrologic system represented by the SVWM include climate (precipitation and maximum and minimum air temperature), PET, ET, soil moisture, runoff, interflow, baseflow, recharge, and streamflow for the Salinas Valley study area. The SVWM was developed using the BCM (Flint and others, 2021) and the HSPF (version 12.4; Bicknell and others, 1997, 2005). The daily climate inputs required by HSPF are provided by the BCM (precipitation maximum and minimum air temperature and PET).

The HSPF application is a widely used and well documented modeling program originally developed and supported by the U.S. Environmental Protection Agency and USGS in 1984 and based on the Stanford Watershed Model (Bicknell and others, 2001, 2005). The HSPF application has been used extensively for hydrologic studies covering a broad range of areas and for a variety of objectives (Atkins and others, 2005; Amirhossien and others, 2015; Stern and others, 2016). Documentation of HSPF version history and the use and theory that led to the development of HSPF is widely available (Donigian and Imhoff, 2009). The conceptual model of the natural hydrologic system represented by the SVWM is based on the HSPF algorithm.

The HSPF computer program is a semi-distributed, process-based, mostly deterministic, and continuous-simulation algorithm for simulating water flow and storage processes with options for simulating water quality and transport processes (Donigian and others, 1984; Bicknell and others, 2001). Applications of HSPF can be used to simulate basin response to normal and extreme precipitation (or lack of precipitation) and to simulate spatial and temporal variability and trends in climate; to evaluate water budgets and water quality; and to evaluate changes in the hydrologic system, such as flow regimes, flood peaks and volumes, soil-water relationships, and groundwater recharge. Through parameter optimization and sensitivity analysis, models developed using HSPF can be calibrated to multiple streamgages to reflect a variety of physiographic characteristics.

Hydrologic processes simulated by HSPF include pervious and impervious surface storage (including interception and retention storage), pervious and impervious surface runoff (overland flow), pervious land infiltration, soil water storage, percolation, ET, interflow, recharge, streamflow, stream losses to evaporation and seepage, and shallow (active) groundwater reservoir storage and discharge (baseflow) contributions to streamflow and riparian ET (fig. 11; Bicknell and others, 2005). The HSPF model provides a comprehensive simulation of rainfall-runoff and streamflow processes, allowing for analysis of surface and shallow subsurface-water budget components and processes such as soil moisture, ET, and recharge for inter-channel areas and components of streamflow (overland runoff, interflow, baseflow, and streamflow seepage) for intra-channel areas.

The dominant components of the HSPF conceptual model are the surface and shallow-subsurface (primarily the root zone), stream channels, and larger water bodies (lakes and reservoirs). The surface and shallow-subsurface systems include the plant canopy, land surface, soil zone, and groundwater reservoir supplying groundwater discharge to the stream channels. The plant canopy includes natural vegetation, crops, and landscaped urbanized areas. Two types of land surfaces are represented by HSPF: (1) pervious land areas (PERLNDs) and (2) impervious land areas (IMPLNDs), which are further broken down according to soil type, land cover type, vegetation density, geology, and topography. Pervious land areas represent natural and developed land areas with soil cover, and IMPLNDs represent impervious developed land surfaces, such as rooftops, roads, and parking lots.

Each PERLND is partitioned into several storage zones, including interception storage, surface detention (upper zone) storage, interflow storage, root-zone (lower-zone) storage representing soils and the upper subsurface of consolidated or unconsolidated rock underlying PERLNDs (for example, fractured or weathered bedrock), and active groundwater storage (Bicknell and others, 2005). The lower zone storage is conceptually defined as extending from the ground surface to the base of the root zone. Pervious land areas are used to simulate the storage and transfer of water to the atmosphere as ET, shallow subsurface flow through the soil to stream channels as interflow, and deeper percolation to groundwater reservoirs as recharge.

The HSPF layout for the SVWM was configured such that all runoff generated by the PERLND and IMPLND model components (including overland runoff, interflow runoff, and groundwater discharge) is routed to a single stream reach or reservoir (RCHRES) model component. In addition to receiving inflows from upstream PERLNDs and IMPLNDs, RCHRESs along the downstream sections of flow paths receive inflows from one or more tributary RCHRESs. Streamflow through the connected stream network is simulated using the kinematic wave approximation of the Saint-Venant equations of one-dimensional channel flow, where water accumulated in each RCHRES from contributing land areas and upstream RCHRESs is routed to the downstream RCHRES.

The Hydrologic Simulation Program—Fortran uses a simplified representation of the deeper subsurface below the root zone that includes the groundwater-flow system. Recharge to the groundwater system is partitioned between active and inactive groundwater reservoirs. Water that is stored in the active groundwater reservoir is available for ET and groundwater discharge to streams, whereas recharge to the inactive groundwater reservoir is an outflow from the hydrologic system simulated by HSPF. The simplified representation accounts for the component of total streamflow that originates as groundwater discharge to stream channels (also referred to as the baseflow component of streamflow), which is important for model calibration and the simulation of water budgets. The inactive groundwater reservoir is used to represent outflows from the hydrologic system that are not explicitly defined by the HSPF model, such as pumping, groundwater underflows across basin boundaries, or increases to deep aquifer storage (deep recharge).

In the conceptual model applied for the SVWM, irrigation and flow diversions were assumed to be minor compared to precipitation, especially in the upland areas of the Salinas Valley study area where the percentage of irrigated agricultural land is small. Furthermore, it was assumed in the conceptual model that most of the irrigation water is returned to the atmosphere by ET. Reservoir operations were not simulated by the SVWM. However, the daily water budgets in the subdrainages contributing water inflows to the reservoir areas in response to precipitation were simulated. Modules in HSPF used to simulate snowfall, snow storage, and snowmelt were not activated in the SVWM; all precipitation was simulated as rain. Although snowfall can occur at higher elevations in the SVWM study area, the accumulation and subsequent melting of snow was not considered to be a significant factor affecting the natural hydrologic system of the Salinas Valley study area.

Simulation Period and Initial Conditions

The SVWM was run using an hourly time step to simulate a continuous daily water balance for each HRU and RCHRES. Daily climate inputs developed by the BCM were partitioned into hourly increments for the HSPF simulation. The target simulation period used for the SVWM started with water year 1949 (October 1, 1948) and ended with water year 2018 (September 30, 2018). The target simulation period was intended to provide an estimate of the historical natural water balance and to allow for a sufficiently long period of time to analyze the spatial and temporal variability of water-balance components, including ET, surface- and subsurface-water storage, recharge, and streamflow.

Precipitation-runoff models generally require an initialization period to help minimize uncertainties associated with the assumed or estimated initial conditions (Markstrom and others, 2008; Hevesi and others, 2019). The SVWM simulations were run starting on October 1, 1947, to provide for a 1-year initialization period (water year 1948) before the target simulation period (water years 1949–2018) to reduce the effect of transients associated with initial conditions specified at start-up. The initial conditions include water contents for interception storage, retention storage, soil moisture (upper and lower zones), interflow storage, and active groundwater storage that were either set to zero or based on values defined in the HSPF manual (Bicknell and others, 2005).

Model Layout and Discretization

The SVWM discretization is used to account for the spatially varying physical characteristics of the Salinas Valley study area, including spatial heterogeneity in topography, vegetation and land cover, soils, geology, water bodies, and climate. The discretization also is used to define the various tributary surface-water drainages connected to the main channel of the Salinas River, the smaller drainages along the Monterey Bay coast, and drainages supplying boundary inflows from upland areas to the lower elevation developed land areas of the Salinas Valley.

The HRU is the basic homogeneous model segment used to spatially partition and discretize the SVWM domain. The SVWM was discretized such that each HRU contains a single PERLND connected to a single RCHRES, with the network of connected RCHRESs representing the natural surface-water drainage system of the Salinas Valley study area as a one-dimensional flow routing network.

Outflows from each HRU are ET, surface-water discharges to RCHREs, and inflows to the inactive groundwater reservoir. Surface-water discharges from HRUs to RCHRESs are overland runoff, interflow, and groundwater discharge from the active groundwater reservoir. Water storage, inflows, and outflows are simulated as a water-equivalent depth for the area of each HRU and groundwater reservoir associated with each HRU. Inflows to each RCHRES are hourly rainfall, outflows from the connected HRU, and surface-water inflows from upstream RCHRESs.

The SVWM discretization includes 690 HRUs, ranging in area from 65 to 25,376 acres (0.1–39.6 mi²), with mean HRU land-surface elevations ranging from 12 to 4,252 ft (fig. 12). The HRU delineation was developed using irregular-polygon areas representing surface-water sub-drainages that were defined using a combination of 10-m (98-ft) digital elevation models (DEMs), the NHD streamlines and sub-drainage boundaries (U.S. Geological Survey, 2019), and Calwater version 2.2.1 watershed areas (California Department of Forestry and Fire Protection, 2004).

The percentage of impervious developed land cover averaged over the area of each HRU ranged from 0 to 55 percent (fig. 13). Hydrologic response units with a substantial percentage of impervious cover include one IMPLND element as a fraction of the total HRU area, with a corresponding reduction in the PERLND area such that the combined IMPLND and PERLND area equals the total area for that HRU. A total of 181 HRUs include IMPLND elements, with impervious areas ranging from 11 to 2,216 acres.

The SVWM discretization was defined such that each HRU is connected to one RCHRES, resulting in a stream channel network of 690 linked RCHRESs, with a given RCHRES having zero to many tributary connections to upstream RCHRESs, depending on the position of the RCHRES in the routing network. All RCHRESs in the stream channel network discharge to a single downstream RCHRES, with exceptions for a small number of RCHRESs discharging to Monterey Bay from the SRO, ELK, and MSB subdrainages. First-order RCHRESs in the headwater drainages do not have a tributary connection to upstream RCHRESs and only receive inflows from the single connected HRU. The headwater RCHRESs represented by the RCHRES network have drainage areas ranging from 0.1 to 50 mi² with 46 headwater RCHRESs having drainage areas less than 1 mi² and 179 RCHRESs having drainage areas of 1–5 mi² (fig. 14). There are 231 RCHRESs that have contributing drainage areas between 5 and 20 mi², and there are 150 RCHRESs that have drainage areas greater than 50 mi².

The HSPF code version used to develop the SVWM is limited to a maximum number of 1,000 model elements, with each PERLND, IMPLND, and RCHRES defining separate model elements. The SVWM model discretization includes a total of 1,561 elements (the sum of 690 PERLNDs, 181 IMPLNDs, and 690 RCHRESs), exceeding the capacity of the HSPF code version used. To preserve the degree of spatial detail provided by the 1,561 model elements, the SVWM was divided into the 2 HSPF sub-model domains, the USVS and the LSVS (fig. 2). The USVS has an area of 2,536 mi² containing 387 HRUs with 36 HRUs including IMPLNDs. The LSVS has an area of 1,992 mi² containing 303 HRUs with 69 HRUs including IMPLNDs. The LSVS is connected to the USVS by the outflow of the Salinas River at USGS streamgage 11150500 (gage 19, Salinas River near Bradley).

The more detailed discretization provided by the two connected sub-models allowed for an improved representation of variability in climate and watershed characteristics such as slope, land cover, soil properties, and surficial geology, compared to a single model representing the entire study area. In addition, the higher level of discretization using two sub-models with 1,561 elements compared to a single model with 1,000 elements, resulted in a more precise delineation of all tributary drainages to the lower Salinas Valley study area (defined by the SVIHM boundary as mentioned earlier; fig. 15). The SVWM simulated surface-water tributary inflows at 148 locations along the SVIHM boundary from the tributary drainages. The simulated inflows can be used to define boundary conditions for integrated hydrologic modeling of the lower Salinas Valley study area. The drainage areas for the 148 inflow locations were grouped into 25 tributary drainages with areas ranging from 11 mi² for the Quail Creek (QUAI) tributary drainage in the northwest part of the SVWM (southeast of the city of Salinas) to 1,574 mi² for the Salinas River (SALI) tributary drainage supplying surface-water inflow from the upper Salinas Valley watershed (fig. 15; table 8). Average elevations for the tributary drainages range from 997 ft for the Cherry Canyon (CHER) tributary drainage to 3,393 ft for the Arroyo Seco (SECO) tributary drainage (table 8). The total 3,625 mi² area comprising all tributary drainages, referred to as the Salinas Valley upland area (SVU), has an average elevation of 1,671 ft and elevations ranging from −1 to 5,872 ft. In contrast, the 904 mi² land area of the SVIHM has an average elevation of 441 ft with elevations ranging from −1 to 3,003 ft (table 8).

Basin Characterization Model Climate Inputs

The BCM was used to develop climate inputs for the SVWM. The BCM consists of a set of computer codes for grid-based water balance simulations and includes preprocessing applications for spatially distributing and downscaling climate variables and for simulating PET for historical climate and future climate scenarios and projections (Flint and Flint, 2007, 2012; Flint and others, 2013, 2021; Stern and others, 2016). The BCM applications use monthly climate data from the Parameter-elevation Regression on Independent Slopes Model (PRISM; Daly and others, 1994, 2004), available daily climate records, and the Gradient-Inverse-Distance-Squared method (Nalder and Wein, 1998) to downscale and spatially interpolate climate data to a 270-m (886-ft) grid covering the SVWM study area. Inputs used by the BCM to develop the HSPF climate inputs consisted of daily precipitation records from 155 climate stations (fig. 16), daily maximum and minimum air temperature records from 113 climate stations (fig. 17), gridded PRISM maps of monthly precipitation, monthly maximum and minimum air temperature, and land-surface elevations for the 270-m grid of the SVWM. The precipitation and air temperature climate grids developed by the BCM were applied to simulate daily PET using the Priestley–Taylor method (Flint and others, 2021).

A listing of the 155 climate stations shown on figure 16 with daily precipitation records used to develop climate inputs is provided in appendix 1, table 1.1. A listing of the 113 climate stations shown on figure 17 with daily minimum and maximum air temperature records used to develop climate inputs is provided in appendix 1, tables 1.2 and 1.3, respectively. The number of years of data for daily precipitation ranged from 1.8 years for Salinas 6 SSW (station 77, app. 1, table 1) to 70.8 years for Paso Robles (station 68, app. 1, table 1). The number of years of data for daily minimum and maximum air temperature ranged from 1.9 years for King City Airport (station 16, app. 1, tables 1.2, 1.3) to 70.4 years for Santa Cruz (station 43, app. 1, tables 1.2, 1.3). Mean annual precipitation ranged from 5.7 in. for Orchard Sunflower Valley (station 61, app. 1, table 1.1) to 54.2 in. for Ben Lomond (station 107, app. 1, table 1.1). Mean minimum daily air temperature ranged from 37.7 °F for station Priest Valley (station 32, app. 1, table 1.2) to 57.0 °F for station Kettleman Hills (station 64, app. 1, table 1.2) and mean maximum daily air temperature ranged from 60.4 °F for San Simeon Point Piedras Blancas (station 31, app. 1, table 1.3) to 80.1 °F for Avenal 9 SSE (station 1, app. 1, table 1.3).

The four BCM-simulated 270-m gridded climate inputs (precipitation, maximum and minimum air temperature, and PET) were averaged over the area of each HRU. Using a uniform hourly distribution, the four daily climate time series developed for each HRU were disaggregated into hourly time series, starting 1 second after midnight on October 1, 1947, and ending at midnight on September 30, 2018. The 690 unique sets of hourly climate time-series inputs were compiled and stored in a single binary Watershed Data Management (WDM) file used by the HSPF code.

The annual (water year) basinwide mean precipitation estimated for the Salinas Valley study area using the BCM indicates high interannual variability, with annual precipitation greater than 32 in. for 6 water years and less than 12 in. for 10 water years, compared to a 70-year mean precipitation of 18.5 inches per year (in/yr; fig. 18A). The ARR subbasin had the highest annual precipitation for all water years, with a maximum annual precipitation of about 61 in. for water year 1998 (fig. 18B). In comparison, water year 1983 was the wettest year for the SRH and NAC subbasins, with about 51 in. of annual precipitation for both subbasins (fig. 18B). Water year 2014 was the driest year in the 70-year period with about 7 in. of precipitation basinwide for the SVWM (fig. 18A), varying from about 13 in. for the ARR subbasin to about 6 in. for the LOR and MSR subbasins (fig. 18B).

The 70-year (water years 1949–2018) mean precipitation simulated by the BCM and averaged over the 690 HRU areas indicates a high degree of spatial variability, with about 36–60 in/yr for the high-elevation HRUs along the western boundary to less than 12 in/yr for low-lying HRUs in the central part of the valley and the southeastern part of the study area, and values less than 10 in/yr along the southeastern boundary (fig. 19). Precipitation was less variable in the Salinas Valley lowlands and coastal basins, ranging from about 12 to 15 in/yr for most locations, with higher values of 15–21 in/yr for the coastal basins in the northwest part of the lower Salinas Valley.

Annual (water year) 70-year mean PET simulated by the BCM and averaged over the 690 HRU areas also indicates substantial spatial variability in climate (fig. 20). The mean PET for the SVWM was 58.1 in. and varied from a minimum of about 55.3 in. for water year 1998 to high values of more than 59.5 in. for several water years including 1959, 1984, and 2014 (fig. 20A). Water year 2014, the driest year, also had the highest PET of 60 in. The MCB subbasin had the lowest PET for all water years, ranging from about 48 in. for water years 1999 and 2011 to 54 in. for water year 1997 (fig. 20B). The EST and USR subbasins had the highest annual PET values of about 62 in. or higher for water years 1959, 1960, 1984, 1996–97, and 2014–15 (fig. 21).

The BCM-simulated 70-year mean PET varied from high values of about 61–63 in/yr for HRUs in the more inland, southeast part of the SVWM to low values of about 42–46 in/yr for HRUs closer to the coastline (fig. 21). North- to northeast-facing slopes on the west side of the Salinas Valley (particularly in northwestern part of the SVWM) also had lower mean annual PET values of about 49–52 in/yr compared to the basinwide mean of 58.1 in/yr. Mean annual PET within the lower elevation of the upper Salinas Valley varied from about 60 to 63 in/yr and overall had higher values compared to most locations in the Salinas Valley study area.

Seasonal variability in precipitation and PET is an important characteristic of the Salinas Valley study area. The basinwide mean monthly precipitation is the highest for January for the SVWM and all subbasins, ranging from about 3 in. for the EST and MSR subbasins to 6.4 in. for the ARR subbasin (fig. 22A). February is the second wettest month for the SVWM and all subbasins, varying from about 2.8 in. for the LSR, EST, and MSR subbasins to 5.8 in. for the ARR subbasin. For the dry-season months, June through August, mean monthly precipitation is approximately zero for all subbasins. In contrast to precipitation, mean monthly PET is highest during the dry season months of May through August (fig. 22B). Although the basinwide 70-year mean PET of 58.1 in/yr is more than three times the basinwide 70-year mean precipitation of 18.4 in/yr, the mean monthly precipitation for December through February exceeds mean monthly PET.

Model Parameters

The HSPF code requires a User Control Input (UCI) file to specify model control options, the simulation period, model parameters, and input and output filenames needed to run a simulation. The UCI file is an American Standard Code for Information Interchange text file with a formatted, column-specified input structure. The model parameters are organized according to the modular structure of HSPF, with separate input groups used for simulating water flow and storage within PERLND areas, IMPLND areas, and the connected RCHRES network. In addition to the UCI file, simulations require the WDM file containing the time series inputs and outputs, including hourly climate (precipitation, air temperature, and PET) and hourly and daily streamflow.

Parameters used to simulate water flow and storage in PERLNDs define spatially varying catchment properties throughout the SVWM and were the most important for representing the physical characteristics of the Salinas Valley study area (table 9). Initially, model parameters were defined using a combination of geospatial data and representative values from previous studies, including suggested values provided in the HSPF user’s manual (U.S. Environmental Protection Agency, 2000; Bicknell and others, 2001). During the model-calibration procedure, 15 of the 17 parameters listed in table 9 were then adjusted and refined, as discussed in more detail in the “Model Calibration” section. Parameters (defined in table 9) LZSN, INFILT, LSUR, KVARY, AGWRC, INFEXP, DEEPFR, BASETP, and AGWETP were scaled using geospatial data and calibrated as a set of unique values for each HRU. Monthly parameters INTERCEP, UZSN, MANNING, INTERFLW, IRC, and LZETPARM were scaled using geospatial data and calibrated for each HRU as a unique set of 12 monthly values.

Initial estimates of LZSN, the lower zone storage capacity, were varied as a function of the mean AWS150 value from SSURGO (fig. 9), the mean 30-m DEM slope (fig. 4), and the weighted average of DEM slope classes for each HRU. The model parameter INFILT affects the infiltration capacity of the soil and was varied for each HRU as the weighted average of the 13 SSURGO soil texture classes within the SVWM, where the weighting factors were defined by the area-fraction of each soil texture class for each HRU (fig. 8). The model parameter LSUR, the length of the overland flow plane used to simulate re-infiltration of surface runoff, was estimated based on the inverse of DEM-derived slope, with steeper slopes resulting in lower LSUR values, and then adjusted during calibration.

The model parameters KVARY and AGWRC were varied for each HRU using the surficial geologic map (fig. 10; Jennings, 1977). The KVARY parameter affects the simulation of groundwater recession flow, enabling the recession flow to be non-exponential in its decay with time, and was calculated as the weighted average of KVARY values defined for each of the 9 surficial geologic rock types (for this study all metamorphic rock types from Jennings [1977] were grouped as metamorphic rocks and all volcanic rock types from Jennings [1977] were grouped as volcanic rocks), with the weighting factors calculated as the area-fraction of the surficial geologic rock type within each HRU (fig. 10). The AGWRC parameter is the basic groundwater recession rate and also was calculated for each HRU as the area-fraction weighted average of values defined for each of 9 surficial geologic rock types. The INFEXP parameter controls the infiltration rate into the root zone and was estimated as a function of the mean DEM slope for each HRU. The DEEPFR parameter controls the rate of inflow to the inactive groundwater reservoir (groundwater that does not contribute to groundwater discharge in the HSPF simulation). As with KVARY and AGWRC, DEEPFR was varied for each HRU based on the surficial geologic map and the area fraction of the different surficial geologic rock types within each HRU (fig. 10). The AGWETP parameter controls groundwater ET losses and was estimated based on the area-fraction of cropland in each HRU.

Six parameters, INTERCEP, UZSN, MANNING, INTERFLW, IRC, and LZETPARM (table 9), were defined using the HSPF option of having a set of 12 monthly values for each HRU to represent seasonal variability. The INTERCEP parameter defines the interception storage capacity for vegetation and was varied for each HRU as the area-fraction weighted average of values defined for each NLCD land cover type (fig. 6). In addition to land cover type, the NLCD 30-m resolution, percentage of forest canopy cover data (fig. 7A) was used to increase the interception storage capacity based on increasing canopy cover. The INTERFLW parameter controls the interflow inflow rate and was varied using monthly scaling factors and the calculated mean 30-m DEM slope for each HRU. The IRC parameter is the interflow recession coefficient and was also varied monthly as a function of the 30-m DEM slope for each HRU.

The model parameters LZETPARM, MANNING, and UZSN were varied by month and by the area-weighted average of the different NLCD land cover types within each HRU. The LZETPARM parameter is used for simulating plant transpiration as a function of PET, vegetation type, and growing season. The MANNING parameter is the surface roughness coefficient used for simulating overland flow. The UZSN parameter is the upper zone storage capacity used to account for surface retention storage. In addition to NLCD land cover type, the NLCD percentage of canopy cover was used to scale the parameters LZETPARM and UZSN by increasing parameter values with increasing canopy cover percentage. Mean land-surface slope also was used to scale the MANNING and UZSN parameters, with an increase in slope resulting in a decrease in values for both parameters.

Several parameters were defined directly using the Geographic Information System (GIS) applications or by the suggested default values and were held constant during model calibration. The SLSUR parameter, the slope of the overland flow plane used by HSPF to simulate overland runoff and infiltration of overland flow, was defined directly as the mean calculated rise-over-run slope of all 30-m DEM grid cells within each HRU. The INFILD parameter was set to a constant value of 2.0, as recommended in the HSPF user’s manual and supporting documentation (U.S. Environmental Protection Agency, 2000; Bicknell and others, 2001), and was not adjusted during model calibration. Impervious land-area parameters also were defined based on suggested values in HSPF documentation and were not adjusted during model calibration. Impervious land-area parameters LSUR, SLSUR, and NSUR (NSUR is equivalent to the MANNING roughness coefficient used for pervious land areas) were set to 200, 0.117, and 0.05 ft, respectively. The impervious land-area parameter RETSC, the retention storage capacity, was set to 0.1 in.

Parameters for simulating surface-water flow through the 690 RCHRES elements comprising the SVWM drainage network include stream reach length, the change in elevation over the length of the stream reach, and a table defining the stage-area-volume-discharge relation for each stream reach, referred to as the Flow-table (Ftable). The Ftables used in the SVWM included two outlets, one for streamflow and a second outlet used to simulate seepage losses from the infiltration of surface water through the streambed. Discharge for the second outlet representing seepage loss was set to zero or near-zero outflow for stream reaches assumed to have none to negligible stream seepage.

The Ftables were estimated using rating curves, field data, and peak discharge data measured at the USGS streamgages in the Salinas Valley study area (table 2). The data were used to define the Ftables in the RCHRES segments where the streamgages were located. The Ftables for segments without a streamgage were estimated based on the proximity to the nearest streamgage and scaled according to the length and slope of the channel represented by the RCHRES, the channel width, the upstream drainage area, and the estimated permeability of the streambed based on the surficial geology and soil texture. Additionally, Ftables for many stream reaches were adjusted during calibration, particularly with respect to the seepage outflow rate.

Calibration Procedure

Streamflow records for 29 USGS streamgages within the SVWM (fig. 1; table 10) with observed streamflow between October 1, 1948, and October 1, 2015, (water years 1949–2015) were used for calibration. In addition to the USGS streamflow data, estimates of surface-water inflows to Lake Nacimiento from October 1, 1958, to September 30, 2013, and to Lake San Antonio from October 1, 1966, to September 30, 2013, were provided by Monterey County Water Resources Agency (MCWRA) and included as observations. All calibration runs were started on October 1, 1947, to allow for a minimum of a 1-year model initialization period before the target period used for calibration at each streamgage.

Calibration was done using an iterative trial-and-error approach of adjusting PERLND parameters identified in table 9 to improve the match between simulated and observed streamflow. Trial-and-error adjustments for spatially distributed parameters were done primarily by changing scaling factors that were defined using geospatial data, which included topography, soils, land cover, and geology. In the case of discrete geospatial data such as land cover type, soil texture, and geology, the scaling factors were calculated as the weighted mean value based on the area-fraction of discrete data within each HRU. In the case of continuous 30-m (98-ft) gridded data such as land elevation, slope, percentage of forest canopy, and percentage of imperviousness, parameter values were calculated as the average of the gridded data values within each HRU area. This method of parameter estimation and calibration used in the SVWM resulted in a unique set of pervious land parameters for each HRU.

Calibration Statistics

The calibration procedure used a quantitative analysis of the goodness-of-fit between simulated and observed time series based on the available period of record at each streamgage from October 1, 1948, to September 30, 2015, (table 10) and was done using the percent-average estimation error (PAE) and the Nash–Sutcliffe model efficiency (NSME) statistics (Nash and Sutcliffe, 1970; Donigian and Imhoff, 2009; Duda and others, 2012; Hevesi and Johnson, 2016; Hevesi and others, 2019). Calibration periods varied from a minimum of 5.2 years for USGS streamgage 11147040 (gage 11) to a maximum of 67 years at USGS streamgage 11152000 (gage 1), with a mean calibration period of 28 years for all streamgages (table 10). The PAE statistic provides a measure of model bias, where a value of 0.0 percent indicates an unbiased model fit. Values of PAE in the range of plus or minus 10 percent indicate an acceptable calibration (Donigian and Imhoff, 2009). The NSME statistic is a standardized mean squared-error statistic, similar to the coefficient of determination (R²) used in regression analysis, that is often used to compare goodness-of-fit results between different models (Nash and Sutcliffe, 1970; Markstrom and others, 2008; Hevesi and others, 2019). An NSME value greater than 0.0 indicates the model provides a better match to the observed values than the mean of the observed values (the mean has an NSME equal to 0.0). The closer the NSME is to 1.0, the better the match is between simulated and observed values, with a value of 1.0 indicating a perfect match to observed values.

The PAE and NSME goodness-of-fit statistics between simulated and observed streamflow were evaluated using daily mean, monthly mean, annual (water year) mean, and mean monthly streamflow at each streamgage for varying values of parameters selected for calibration (table 9). To summarize the calibration results for the 31 streamflow comparisons, weighted mean PAE, percent estimation error (PE), and NSME statistics were calculated for each parameter set tested. The weighting factors were defined for each streamgage using the average of two products consisting of (1) the length of the calibration period at each streamgage divided by the total length of record for all streamgages and (2) the square root of the contributing drainage area for each streamgage divided by the sum of the square root of the total drainage area for all streamgages. The weighted mean PAE was calculated using the absolute value of PAE calculated at each individual streamgage. Calibration criteria used for the SVWM are provided in table 11. To achieve a minimum level of calibration for the SVWM, the weighted mean goodness-of-fit statistics needed to meet the satisfactory criteria or better for daily, monthly, and annual streamflow, with the goal of having results meet the good to very good criteria.

Parameter Sensitivity

Before the process of trial-and-error model calibration, an initial set of parameter values was defined using the recommended range of values provided by the HSPF user’s manual and supporting documentation (U.S. Environmental Protection Agency, 2000; Bicknell and others, 2001, 2005). Six baseline models were defined using the range of typical and possible parameter values listed in BASINS Technical Note 6, Estimating Hydrology and Hydraulic Parameters for HSPF (U.S. Environmental Protection Agency, 2000). Typical and possible mean parameter values were calculated using the minimum and maximum values for the typical and possible ranges. The six baseline parameter sets consisted of (1) typical minimum values, (2) typical maximum values, (3) typical mean values, (4) possible minimum values, (5) possible maximum values, and (6) possible mean values, with all HRUs having uniform parameter values as listed in table 9.

Goodness-of-fit statistics were calculated for the 31 simulated daily streamflow comparisons using the six baseline models as a means of evaluating initial estimates for pervious land parameters. The weighted mean PAE, PE, and NSME results indicated that the possible mean and typical maximum baseline models provided better initial parameter values compared to the typical minimum, typical mean, possible minimum, and possible maximum baseline values (table 12). The possible mean baseline model with PAE of 39.2, PE of −15.2, and NSME of 0.355 provided the best overall goodness-of-fit results compared to the four other baseline models; however, all six baseline models failed to provide a satisfactory calibration result. The typical and possible minimum baseline models greatly overestimated the amount of streamflow, and the typical minimum, possible minimum, and mean baseline models provided a very poor fit to observed streamflow based on the negative NSME results. The possible mean, typical maximum, and possible maximum baseline models provided an improved model fit compared to the other three baseline models but underestimated the weighted average mean, maximum, and minimum streamflows. All six baseline models underestimated the weighted average median streamflow.

As part of parameter estimation, the sensitivity of the SVWM to pervious land parameters was evaluated using a simple approach of varying selected parameters across a range of possible values and comparing the weighted mean goodness-of-fit statistics for simulated daily streamflow. The analysis was used to help indicate which parameters would likely have the greatest effect on streamflow and therefore have the greatest effect on calibration results. To perform the sensitivity analysis, pervious land parameters were set to uniform values across all HRUs for each simulation. Minimum and maximum values in table 9 provided an approximate guide for defining the range of parameter values tested. Each test simulation was done using parameter values defined by the typical mean baseline model (table 9), whereas a single parameter was varied across a range of minimum to maximum values. Parameters having the option for monthly values (table 9) were set to a constant value for all months. For each parameter tested, all other parameters were set to constant uniform values for all HRUs as defined by the typical mean baseline model (table 9). The procedure was repeated for each parameter in table 9.

Results indicated a high degree of model sensitivity in terms of simulated daily streamflow to parameters LZSN, INFILT, and LZETPARM (table 13). The best goodness-of-fit results using the typical mean baseline model were obtained with values from 8.0 to 15.0 in. for LZSN, 0.05 in/hr for INFILT, and 0.99 for LZETPARM. The LZSN values less than 5.5 in. resulted in a substantial overestimation of streamflow because of insufficient water storage capacity in the root zone causing a reduction in ET. The LZETPARM values of 0.3 and less also resulted in a reduction in ET and a corresponding overestimation of streamflow. The LZSN values greater than 8 in. resulted in an underestimation of streamflow because of the high root zone storage capacity causing a high percentage of precipitation being lost to ET.

Model sensitivity was shown as being intermediate to variations in parameters affecting baseflow recession (KVARY and AGWRC), the parameter controlling deep recharge to groundwater not contributing to baseflow (DEEPFR), parameters defining the interception storage capacity of vegetation and the retention storage capacity of the land surface (INTERCEP and UZSN), and the parameter controlling interflow recession (IRC; table 13). Model sensitivity was shown to be low for parameters controlling the re-infiltration of overland flow (LSUR and SLSUR), the exponent in the infiltration equation (INFEXP), the parameters controlling groundwater ET (BASETP and AGWETP), the roughness coefficient for overland flow (MANNING), and the coefficient controlling interflow (INTERFLW).

Calibrated Parameters

Summarized parameter values calculated as the mean, median, maximum, minimum, and variance for the 690 calibrated HRU values for the SVWM are listed in tables 14 and 15. Summarized values for parameters with constant monthly values (values are the same for each month) are in table 14. Summarized values for parameters with monthly varying values are in table 15. The goodness-of-fit statistics for the set of parameters in tables 14 and 15 are provided in the “Calibration Results” section.

Calibrated values for the length of the overland flow plane, LSUR, were defined by using the equation 0.7 minus HRU slope times 600 ft, resulting in a range of values from 49.2 to 418.5, with a mean value of 275.5 for the SVWM. The soil zone water storage capacity, LZSN, was the most critical parameter adjusted during calibration, with values for the calibrated model ranging from 0.10 to 29.74 in. and a mean value of 8.69 in. for the SVWM. The mean calibrated value for INFILT is 0.24, ranging from 0.0005 to 13.3, with high values of 0.81 and greater mostly in the lowlands of Salinas Valley and locations with loamy soils. The KVARY parameter, adjusted for each geologic rock type, includes a minimum value of 0.2 for metamorphic rock types and a maximum value of 2.0 for alluvium and conglomerate surficial geology.

Calibrated values for AGWETP, used to empirically represent groundwater pumped for crop irrigation, include high values of 0.2 and greater in areas with substantial cropland in the middle and lower Salinas Valley, and intermediate values between 0.06 and 0.2 are in the upper Salinas Valley for HRUs containing vineyards. Calibrated values for AGWRC, scaled according to surficial geologic rock type, range from low values of 0.91 for alluvium and 0.937 for conglomerate to high values of 0.97 for rock types in upland areas, including a maximum of 0.999 for igneous and metamorphic rock types.

Calibrated values for INTERCEP are highest in May and lowest in October (table 15). Higher values for INTERCEP are assigned to HRUs having mostly forested land cover and a high percentage of canopy cover, compared to HRUs with mostly grassland. The INTERFLW parameter, controlling the rate of interflow to streams, is highest for February and March and lowest for July through October (table 15). The highest INTERFLW values for March and October are for HRUs in the UAS and along the western boundary of the LSR subbasins. Calibrated values for the IRC parameter have a similar spatial pattern as INTERFLW, with high values in the Arroyo Seco subdrainage (table 15).

Calibrated values for LZETPARM are highest in April and lowest in October (table 15). Calibrated values for MANNING are highest in March and April and lowest in September and October (table 15). Calibrated values for UZSN include maximum values for March through May and a minimum mean for the SVWM for October (table 15). Locations with the highest values in April include the lower Salinas Valley floor and the SRH.

Calibration Results

The overall goodness-of-fit statistics for the calibrated model included weighted mean PAE and NSME values of 2.0 and 0.64 percent for daily mean streamflow, 2.0 and 0.84 percent for monthly streamflow, and 2.1 and 0.88 percent for annual streamflow, respectively (table 16). The results were considered satisfactory for daily streamflow, very good for monthly streamflow, and good for annual streamflow.