Watershed Modeling for the Santa Clara River in Southern California
A.S. Donigian1, Jr., B.R. Bicknell1, and M. Bandurraga2
1AQUA TERRA Consultants, 2685 Marine Way, Suite 1314, Mountain View, CA
94043; PH (650) 962-1864; email:email@example.com, firstname.lastname@example.org
2 Ventura County Watershed Protection District, 800 S. Victoria Ave., M/S 1610,
Ventura, CA 93009; PH (805) 654-2015; email: email@example.com
A calibrated watershed hydrologic model of the Santa Clara River (SCR) Watershed
was developed for use as a tool for watershed planning, resource assessment, and
ultimately, water quality management purposes. The study was a joint effort of the
Ventura County Watershed Protection District (VCWPD), Los Angeles County
Department of Public Works (LACDPW), and U.S. Army Corp of Engineers Los
Angeles District. The modeling package used for the study was the U.S. EPA
Hydrologic Simulation Program – FORTRAN (HSPF).
The SCR main stem flows east-to-west from the San Gabriel Mountains of central
Los Angeles County to its mouth at the Pacific Ocean near the cities of Ventura and
Oxnard. The 4,263 square kilometer watershed includes rapidly urbanizing valleys
and extensive agriculture, and is subject to severe flooding and erosion. There are
four large reservoirs within the watershed. Although the Santa Clara River watershed
remains primarily in a natural physical state, the flow regime within the watershed is
highly engineered to optimize water deliveries and aquifer recharge.
The SCR Watershed was segmented into 209 subwatersheds and 192 stream reaches,
based on precipitation patterns, drainage boundaries, hydrography, stream gage
locations, and impaired water segments. Each land segment was further subdivided
into individual model segments based on land use/cover categories. A long-term data
base of 46 years of model input data (precipitation, evaporation, diversions, POTWs,
etc.) was developed. The model was calibrated and validated to observed flow at
multiple locations, using both graphical and statistical model-data comparisons.
Long-term simulations were run to assess the impacts of alternative conditions
(baseline, natural) on flow and flood frequencies. The model was also used to
generate design storm event hydrographs for selected return intervals with synthetic
input rainfall hyetographs for the corresponding rainfall return period developed from
available rain gage data. The design storm results were submitted to FEMA for the
Flood Insurance Study currently underway.
This study was undertaken to develop a comprehensive watershed hydrologic model
of the SCR Watershed for use as a tool for watershed planning, resource assessment,
and ultimately, water quality management purposes. This study was a joint effort of
the Ventura County Watershed Protection District (VCWPD), Los Angeles County
Department of Public Works (LACDPW), and U.S. Army Corp of Engineers
(USACE) Los Angeles District (USACE, 2003). The modeling package used for this
application was the Hydrological Simulation Program-FORTRAN (HSPF) (Bicknell
et al., 2005). HSPF is a comprehensive watershed model of hydrology and water
quality, that includes modeling of both land surface and subsurface hydrologic and
water quality processes, linked and closely integrated with corresponding stream and
Two previous HSPF studies in Calleguas Creek, adjacent to the SCR Watershed,
provided the foundation for this effort. In both studies, HSPF was set up and
calibrated to available flow records for recent hydrologic conditions, and customized
to include consideration of localized groundwater pumping impacts and
lawn/landscape irrigation practices on surface water flow levels. The Calleguas
model also included consideration of diversions and deep groundwater recharge
losses through the streambed. In this study, initial hydrologic parameters and the
procedures for representing groundwater pumping, irrigation, and channel losses were
initially based on these predecessor models but subjected to further review,
refinement, and revisions as needed for the SCR watershed conditions.
The SCR is the largest river system in southern California that remains in a relatively
natural state. The main stem flows east-to-west from the San Gabriel Mountains of
central Los Angeles County to its mouth at the Pacific Ocean between the towns of
Ventura and Oxnard (Figure 1). After descending from its mountainous headwaters,
the river passes through the northern Los Angeles suburb of Santa Clarita, across the
Los Angeles/Ventura County line, then transitions to the mostly agricultural valley
with a series of small towns, and finally discharges to the ocean. All major tributaries
flow from the north and include (from upstream to downstream) Bouquet Canyon,
San Francisquito Canyon, Castaic, Piru, and Sespe Creeks. Sespe Creek lies in a
relatively wet zone of the watershed and has a 100-yr peak that is about 60% of the
mainstem peak even though the Sespe area is only 16% of the total. There are four
major reservoirs within the tributary system. Pyramid and Castaic Reservoirs are part
of the State Water Project (SWP) system and are operated by the California
Department of Water Resources (CDWR). Pyramid is located on Piru Creek while
Castaic is located on Castaic Creek, but the two are hydraulically connected. State
water is sent through the William E. Warne Power plant into Pyramid Lake, through
the Angeles Tunnel into the Castaic Power plant, and then into Castaic Lake,
terminus of the West Branch of the SWP. Piru Reservoir is run by the United Water
Conservation District (UWCD) and is located on Piru Creek below Pyramid Lake.
UWCD’s primary operational goals are groundwater recharge, public recreation, and
The watershed drainage area is about 4,263 square km, ninety percent of which
consists of rugged mountains, ranging up to 2,700 meters high. Los Padres and
Angeles National Forests, home to most of the major northern tributaries, comprise
47% of the watershed area. The remaining ten percent of the drainage area lies on the
valley floor and coastal plain with the main stem of the Santa Clara River. The
watershed is surrounded to the north, east, and south by largely undeveloped hills and
canyons. The watershed is subject to severe flooding and erosion. The SCR
watershed areas to be modeled in this study are shown in Figure 1 along with major
waterbodies, municipalities, and other prominent features.
Figure 1. Santa Clara River Watershed, municipalities, and major waterbodies
One of the goals of this effort is to provide the capability to perform long-term
simulations in order to assess the impacts of alternative conditions – Baseline, Natural
(pre-development) Condition, Alternative Future Conditions (e.g., land use, facilities,
reservoir operations) - on flow and flood frequencies. A long-term data base of 46
years of model input data (precipitation, evaporation, diversions, POTWs, etc.) with
the most critical being precipitation and evaporation was developed. This data base
supported long-term model runs to analyze and compare flow and flood frequencies
of Natural, Current Baseline, and planned studies for Alternative Future conditions.
In addition, the model was used to generate storm event hydrographs for selected
return intervals with synthetic input rainfall hyetographs for the corresponding
rainfall return period developed from available rain gage data. This was performed
by VCWPD and LACDPW for selected tributary and mainstem sites. The results
were used by FEMA for the currently underway Flood Insurance Study update.
Segmentation. Subwatersheds were defined using GIS procedures and a number of
data sources, including NHDPlus, Digital Elevation Models of 10 meter resolution,
and GIS shapefiles of Los Angeles and Ventura Counties containing previously
delineated subbasin boundaries within the more urbanized and flatter areas of the
counties. The primary factors that produced the preliminary segmentation also
included rain gage locations, Thiessen network boundaries, isohyetal contours,
differences in slope and elevation, locations of stream gages, mainstem bridges, and
locations of debris basins. The final model segmentation (Figure 2) consists of 209
model subwatersheds and 192 stream reaches.
Each model reach segment was analyzed to define its hydraulic behavior by a
hydraulic function table, which defines the flow rate, surface area, and volume as a
function of water depth. The function tables for the streams and rivers within the SCR
Watershed model application were developed using HEC-RAS and DEMs of varying
resolution. The resolution of the DEMs ranged from 5 meters in Los Angeles County
to 30 meters in the mountainous northern parts of Ventura County, where less
resolution was required to define the channel. Function tables for the reservoirs and
lakes were developed using stage-storage and stage-surface area relationships
provided in tables and figures from a variety of sources.
Since land use affects the hydrologic response of a watershed by influencing
infiltration, surface runoff, and water losses from vegetal evapotranspiration, each
model segment was segmented further based on land use. The SCR Watershed is a
mix of urban and agricultural lowlands and upland open areas, with the latter
comprising approximately 87% of the total area. Agriculture covers 4% of the
watershed, concentrated along the river valley. The urban areas, including Santa
Clarita, Piru, Fillmore, Santa Paula, and Ventura, are comprised of
commercial/industrial, medium to high-density residential, and low-density
residential areas. The primary land use coverage used in the SCR Watershed model is
based on the Southern California Association of Governments (SCAG) land use
designations, with coverages corresponding to land use conditions for 1990, 1993,
2001, and 2005. Although the SCAG land use data provided a reasonable mix of
urban categories for the model, both the agriculture and the large upland open area
groups needed better definition in order to allow their representation and
contributions within the model. These areas were differentiated into categories that
can better define the actual vegetation types and their characteristics using the
LANDFIRE Rapid Assessment “Potential Natural Vegetation Group” coverage.
Using the SCAG and LANDFIRE coverages, the final set of land uses incorporated in
the model were forest/woodland, shrubland, open/grass, agriculture, low density
residential, medium density residential, high density residential, and
Reservoirs. The major reservoirs (Pyramid, Piru, and Castaic) were modeled by
simulating the natural runoff inputs, and defining daily interbasin transfers (i.e.,
imports) and reservoir releases based on observed data compiled by CADWR (2007)
and USGS. Most of the imported water is ultimately used for water supply, and does
not flow downstream. During calibration, the natural inflows were calibrated to
maintain observed storages or water levels in the reservoirs. For the natural scenario,
the reservoirs were removed and replaced with free-flowing reaches. For the Baseline
Scenario – which was prior to the compilation of transfers and outflows by CADWR
– a reservoir operation rule was used to compute the outflows based on simulated
Figure 2. SCR watershed model segmentation, showing major basins and
streamflow calibration locations
Rainfall and Evaporation Data. Rainfall data was developed from a database of
more than 100 stations operated by Ventura and Los Angeles counties and the
National Weather Service. The rainfall data development and correction task required
filling-in missing data, performing QA/QC, and disaggregating daily data to hourly
utilizing nearby hourly stations. The final model database consisted of 35 of these
stations, which were assigned to model segments using Thiessen analysis and
Evaporation data in the form of potential evapotranspiration (PET) were developed
from 27 stations in and near the watershed. Most of the data were monthly totals,
which were distributed to daily values using a long term daily station at Lake
Cachuma, in Santa Barbara County. The PET data were further disaggregated to
hourly values with a distribution function based on the pattern of daylight at the
latitude of the watershed. The database was extended back to 1956 for long term runs.
The final PET dataset consisted of 12 stations distributed across the watershed.
Snow. Due to the high elevations in the upper watershed, snow accumulation and
melt was modeled for selected model segments in the upper Sespe and Piru Creek
watersheds where significant snowfall is common. Since HSPF allows use of an
optional degree-day method to simulate snow, hourly air temperature time series were
required as inputs to the snow model. A database of air temperature was developed,
using observed data and adjustments for elevation, where necessary.
Irrigation. The Santa Clara River watershed includes significant areas of agricultural
and developed residential land, so the model considers both urban and agricultural
irrigation applications for a complete water balance accounting. The approach
assumed that water was applied to satisfy monthly crop and lawn evapotranspiration
(ET) demands exceeding available rainfall. ET demands were computed using the
“landscape coefficient” method described in the CADWR (2000) manual Water Use
Classifications of Landscape Species. Daily reference ET (ETo) is given by month
for each climate zone in the state. The SCR watershed is spread across a range of
ETo zones, transitioning from the South Coast Marine to inland desert climates.
Weighted crop coefficients were applied, and the amounts were adjusted to account
for application efficiency.
Diversions and Point Sources. Several diversions and point sources were included
in the model. Six diversions, including two sizable ones were defined; they are
primarily used for groundwater recharge and agricultural irrigation. Most irrigation is
derived from deep groundwater, and was not explicitly accounted for as diversions.
Of the nine WWTP’s in the watershed, only one discharges directly to the river.
Groundwater. The HSPF model represents groundwater as both a shallow, active
groundwater storage that can contribute directly to streamflow (baseflow), and a
deep, inactive storage that represents deep aquifers that do not contribute to
streamflow. The flux into the inactive storage is represented as deep recharge. Both
of these groundwater components are evaluated as part of the model calibration
process and the water balance assessment.
In the SCR watershed, groundwater provides much of the water for human use
through pumping from an extensive network of alluvial aquifers and the Saugus
Formation in the river valley, thereby transforming deep, inactive groundwater in the
HSPF model into surface water. Most of the extracted groundwater (> 90%) is used
for agricultural irrigation. In this model, it was assumed that most agricultural
irrigation water was derived from deep aquifers and/or local channel losses
recharging the shallow alluvial aquifers. The model explicitly includes deep recharge
and irrigation applications. These quantities were compared to available annual
estimates for these fluxes in order to improve the accuracy of the modeled water
Another groundwater-surface water interaction issue is the presence of recharge and
discharge zones along the SCR channel. Areas of rising groundwater and recharge
are observed at a number of locations in the watershed as shown in Figure 3.
Literature documenting recharge and discharge zones of the Santa Clara River,
including the model developed by McEachron (2005), was used as the basis for
modeling channel gains and losses in various reaches on the mainstem and tributaries.
CALIBRATION AND VALIDATION
Model calibration and validation followed a ‘weight-of-evidence’ approach involving
multiple graphical and statistical model-data comparisons for a comprehensive
assessment of overall model performance (see Study Report, AQUA TERRA
Consultants, 2009). Model parameterization was initially derived from the prior local
HSPF applications, with subsequent adjustments as part of the calibration process.
Figure 3. SCR ground water basins (from Luhdorff & Scalmanini, 2005)
Parameter adjustments focused primarily on lower zone storage and infiltration
parameters, as a function of soils, land use, and slope conditions, to obtain reasonable
overall water balances. Adjustments to the interflow and baseflow parameters were
made to improve agreements in the flow duration curves, daily time series, and storm
events. The groundwater recession and groundwater ET-related parameters are
usually watershed specific, as they are a function of local groundwater and riparian
conditions; therefore, they were calibrated to local conditions in each watershed.
Calibration and validation periods were based on the available meteorological,
streamflow, and land use data. The data supported model simulations spanning water
years (WY) 1987 through 2005; therefore, the calibration period was WY 1997-2005,
and the validation period was WY 1987-1996. The later time span was selected for
calibration because it covers a wider range of wet and dry years, it includes the most
extensive coverage for POTW and diversion data, and it provides a starting point for
future conditions. Land use coverages were available for 2001 and 1993, and so these
coverages were used for the calibration and validation periods, respectively.
The approach to calibrating the watershed hydrology initially focused on the
relatively natural, undeveloped areas in the upper portions of the watershed in both
counties in order to provide the best estimate of hydrologic parameters without the
complicating issues of irrigation, water regulations, importations, and channel losses.
The major calibration/validation sites and watersheds are shown in Figure 2. Figures
4a and 4b show flow duration plots for the mainstem gage at the Ventura-Los
Angeles County line for the calibration and validation periods, respectively. Figure 5
shows the hydrograph of a major storm event at the basin outlet gage near Montalvo.
Figure 4a. Flow duration curves for SCR at County Line – calibration period
Figure 4b. Flow duration curves for SCR at County Line – validation period
Table 1 shows the ‘Weight-of-Evidence’ summary of the model performance metrics
for the calibration and validation periods, providing the mean and range of the metrics
at the various gage sites. The last column provides the qualitative assessment of the
overall model performance based on how the statistical means and ranges compare to
the targets proposed in the Project Plan. The model results show a range of model
accuracy but the majority of the metrics show a Good-Very Good model
performance. The SCR HSPF model was judged to be a robust representation of the
hydrologic regime, and a viable tool for watershed management purposes.
Figure 5. Hydrograph of February 2005 storm for SCR at Montalvo
Table 1. Weight-of-Evidence for SCR Watershed Model Performance
Calibration Validation Overall Model
Mean Range Mean Range
Runoff Volume, % Δ 2.0 -7.8 - 11.8 2.7 -5.8 / 7.0 Good / Very Good
Correlation Coefficient, R:
- Daily R 0.91 0.74 - 0.96 0.89 0.85 - 0.97 Fair / Very Good
- Monthly R 0.97 0.91 - 0.99 0.97 0.96 - 0.99 Very Good
Coefficient of Determination, R2:
- Daily R2 0.82 0.55 - 0.92 0.80 0.72 - 0.94 Poor / Very Good
- Monthly R2 0.94 0.82 - 0.99 0.94 0.92 - 0.98 Very Good
Flow-Duration Good / Very Good Fair / Good Fair / Very Good
Water Balance Good / Very Good Good / Very Good Good / Very Good
- Daily Storm Peak, % Δ -6.6 -35.9 - 20.1 -7.6 -13.4 - 9.5 Fair / Very Good
SIMULATIONS OF BASELINE AND NATURAL CONDITIONS AND
The long term (46 year) database of meteorologic data and other inputs was used to
make long term simulations of the baseline/existing conditions and natural conditions.
For the existing run, the most recent land use dataset was used, and the reservoirs for
the time prior to the validation period was modeled using a set of operation rules
based on knowledge of the system. In the natural run, urban and agricultural land was
converted to the natural categories, and the reservoirs, irrigation, point sources and
diversions were removed from the model.
The calibrated model was also used as the basis for generating design storm peaks
and hydrographs for use in a hydraulic modeling study performed by the clients. The
approach involved identifying a storm where saturation levels were very high across
the watershed and then applying balanced design storm hyetographs for the 100-yr
storm for each rain gage used in the HSPF model. The gaged tributaries with long-
term records were used as calibration points in the modeling. The calibration was
done by adjusting the rainfall factors applied to the rain data for each subarea and
associated reach at the calibration points to establish corresponding rainfall factors
that could then be applied to ungaged tributaries. The model provided 100-yr peaks
at the calibration location that were within 10% or better of the results based on
analyses of historic annual peak maxima. The model was then run with the
appropriate rainfall distributions at 5-min timesteps for the storm of interest to
provide 100-year design storm peaks at the ungaged tributaries. The 100-year peaks
were converted to other return intervals of interest by using multipliers developed
from flow frequency analyses of long-term Ventura County and Los Angeles County
AQUA TERRA Consultants. (2009). Hydrologic Modeling of the Santa Clara River
Watershed with the US EPA Hydrologic Simulation Program – FORTRAN.
Final Report. Prepared for the Ventura County Watershed Protection District,
Ventura, CA. November 25, 2009.
Bicknell, B.R., J.C. Imhoff, J.L. Kittle Jr., T.H. Jobes, and A.S. Donigian, Jr. (2005).
Hydrological Simulation Program - FORTRAN. User's Manual for Release
12.2. U.S. EPA Ecosystem Research Division, Athens, GA. and U. S.
Geological Survey, Office of Surface Water, Reston, VA.
California Department of Water Resources. (2000). A Guide to Estimating Irrigation
Water Needs for Landscape Plantings in California; The Landscape
Coefficient Method and WUCOLS III. Sacramento, CA. August 2000.
California Department of Water Resources. (2007). State Water Project Monthly
Operations Data, wwwoco.water.ca.gov/monthly/monthly.menu.html.
McEachron, M. (2005). Description and Surface Water Modeling Capabilities of the
Piru Basin. Prepared by the United Water Conservation District.
Luhdorff and Scalmanini, Consulting Engineers. (2005). 2004 Santa Clarita Valley
Water Report. Castaic Lake Water Agency, Santa Clarita Water Division, Los
Angeles County Waterworks District 36, Newhall County Water District,
Valencia Water Company. May 2005.
USACE. (2003). Santa Clara River Watershed, Los Angeles and Ventura County,
California. Feasibility Study. Project Management Plan, Feasibility Phase.
Los Angeles District, South Pacific Division, 15 October 2003.