In estuaries suspended sediment concentration (SSC) is one of the most important contributors to turbidity, which influences habitat conditions and ecological functions of the system. Sediment dynamics differs depending on sediment supply and hydrodynamic forcing conditions that vary over space and over time. A robust sediment transport model is a first step in developing a chain of models enabling simulations of contaminants, phytoplankton and habitat conditions.
This works aims to determine turbidity levels in the complex-geometry delta of the San Francisco estuary using a process-based approach (Delft3D Flexible Mesh software). Our approach includes a detailed calibration against measured SSC levels, a sensitivity analysis on model parameters and the determination of a yearly sediment budget as well as an assessment of model results in terms of turbidity levels for a single year, water year (WY) 2011.
Model results show that our process-based approach is a valuable tool in assessing sediment dynamics and their related ecological parameters over a range of spatial and temporal scales. The model may act as the base model for a chain of ecological models assessing the impact of climate change and management scenarios. Here we present a modeling approach that, with limited data, produces reliable predictions and can be useful for estuaries without a large amount of processes data.
Rivers transport water and sediments to estuaries and oceans. Sediment dynamics will differ depending on sediment supply and hydrodynamic forcing conditions, both of which vary over space and time. The human impact on sediment production dates from 3000 years ago, and has been accelerating over the past 1000 years due to considerable engineering works (Syvitski and Kettner, 2011). Milliman and Syvitski (1992) estimated that the budget of sediment delivered to the coastal zone varies between 9.3 and 58 Gt per year. Estimating the world sediment budget is still a challenge because of the lack of data and detailed modeling studies (Vörösmarty et al., 2003). In addition, there is considerable uncertainty in hydraulic forcing conditions and sediment supply dynamics due to variable adaptation timescales over seasons and years (such as varying precipitation and river flow), decades (such as engineering works) and centuries to millennia (sea level rise and climate change).
Examples of anthropogenic changes influencing sediment dynamics in river basins and estuaries are manifold, e.g., the San Francisco Bay and the inland Sacramento–San Joaquin delta (Bay–Delta system) (Schoellhamer, 2011) and Yangtze estuaries (Yahg, 1998), and the Mekong Delta (Manh et al., 2014). These three systems are similar in how anthropogenic changes altered sediment supply. After an increase in sediment supply (due to hydraulic mining and deforestation) each had a steep drop in sediment discharge (30 % or more) due to reservoir building and further estuarine clearing after depletion of available sediment in the bed. This implies (a) continuous change in sediment dynamics and hence sediment budget in the estuary, and (b) change in sediment availability leading to change in turbidity levels.
Turbidity is a measurement of light attenuation in water and is a key ecological parameter. Fine sediment is the main contributor to turbidity. Therefore, suspended sediment concentration (SSC) can be translated into turbidity by applying empirical formulations. Besides SSC, algae, plankton, microbes and other substances may also contribute to turbidity levels (ASTM International, 2002). High turbidity levels limit photosynthesis activity by phytoplankton and microalgae, therefore decreasing associated primary production (Cole et al., 1986). Turbidity levels also define habitat conditions for endemic species (Davidson-Arnott et al., 2002). For example, in the San Francisco Bay–Delta estuary the delta smelt seeks regions where the turbidity is between 12 and 18 NTU to hide from predators (Baskerville and Lindberg, 2004; Brown et al., 2013). Examples of other ecological impacts related to SSC are vegetation stabilization (Morris et al., 2002; Whitcraft and Levin, 2007), and salt marsh survival under sea level rise scenarios (Kirwan et al., 2010; Reed, 2002).
To assess the aforementioned issues, the goal of this work is to provide a
detailed analysis of sediment dynamics including (a) SSC levels in the
Sacramento–San Joaquin delta (Delta), (b) sediment budget and (c) translation
of SSC to turbidity levels using a two-dimensional horizontal, averaged in
the vertical (2DH), process-based, numerical model. The 2DH model solves the
2-D vertically integrated shallow-water equations coupled with
advective–diffusive transport. This process-based model will be able to
quantify high-resolution sediment budgets and SSC, both in time
(
The Delta and Bay are covered by a large survey network with freely
available data on river stage, discharge and
SSC and other parameters from the US Geological Survey (USGS)
(nwis.waterdata.usgs.gov), Californian Department of Water Resources
(
Regarding ecological value, starting from the bottom of the food web, the Delta is the most important area for primary production in the San Francisco estuary. The Delta is 1 order of magnitude more productive than the rest of the estuary (Jassby et al., 2002; Kimmerer, 2004). It is an area for spawning, breeding and feeding for many endemic species of fishes and invertebrates, including some endangered species like delta smelt (Brown et al., 2013), chinook salmon, spring run salmon and steelhead. Additionally, several projects for marsh restoration in the Delta are planned and the success of these projects depends on sediment availability (Brown, 2003).
SSC spatial distribution and temporal variability is important information for the ecology of estuaries. However, observations including both high spatial and temporal resolution of SSC are difficult to make, so we revert to using coupled hydrodynamic–sediment transport models to make predictions at any place and time.
For the first time, a detailed, process-based model is developed for the San Francisco Bay–Delta, to focus on the complex delta sediment dynamics. From this model it is possible to describe the spatial sediment (turbidity) distribution and deposition patterns that are important indicators to assess habitat conditions. Seasonal and yearly variations in sediment dynamics and turbidity levels can be used as indicators for ecological modeling (Janauer, 2000). This work fills the gap between the physical aspects (hydrodynamic and sediment modeling) and ecology modeling. Previous work focused on understanding the San Francisco Bay–delta system through data analysis (Barnard et al., 2013; Manning and Schoellhamer, 2013; McKee et al., 2006, 2013; Morgan-King and Schoellhamer, 2013; Schoellhamer, 2002, 2011; Wright and Schoellhamer, 2004, 2005), while similar work in other estuaries around the world does not provide a direct link to ecology (Manh et al., 2014).
San Francisco estuary is the largest estuary on the US west coast. The
estuary comprises San Francisco Bay and the inland Sacramento–San Joaquin Delta (Bay–Delta system), which together cover a total area of 2900 km
Tides propagate from the Golden Gate into the bay and most of the Delta up
to the Sacramento River at Freeport (FPT) and the San Joaquin River near Vernalis (VNS) when river discharge is low. Suisun
Bay experiences mixed diurnal and semidiurnal tide that ranges from about
0.6 m during the weakest neap tides to 1.8 m during the strongest spring
tides. During high river discharge the 2 psu isohaline is located in San
Pablo Bay while during low river discharge it can be landwards of Chipps
Island (westernmost reach of the black rectangle; Fig. 1). The topography greatly influences the wind climate
in the Bay–Delta system. Wind velocities are strongest during spring and
summer with afternoon north-westerly gusts of about 9 m s
San Francisco estuary collects 40 % of the total Californian fresh water discharge. It has a Mediterranean climate, with 70 % of rainfall concentrated between October and April (winter) decreasing until the driest month, September (summer) (Conomos et al., 1985). The orographic lift of the Pacific moist air linked to the winter storms and the snowmelt in early spring govern this wet (winter) and dry (summer) season variability. This system leads to a local hydrological water year (WY) defined as 1 October to 30 September, including a full wet season in 1 WY.
The Sacramento and San Joaquin rivers, together, account for 90 % of the
total fresh water discharge to the estuary (Kimmerer, 2004). The daily
inflow to the Delta follows the rain and snowmelt seasonality, with average
dry summer discharges of 50–150 m
The hydrological cycle in the Bay–Delta determines the sediment input to the system, and thus biota behavior. McKee et al. (2006) and Ganju and Schoellhamer (2006) observed that a large volume of sediment passes through the Delta and arrives to the bay in pulses. They estimated that in 1 day approximately 10 % of the total annual sediment volume could be delivered and in extremely wet years up to 40 % of the annual total sediment volume can be delivered in 7 days. During wet months more than 90 % of the total annual sediment inflow is supplied to the Delta.
The Delta's recent history is dominated by anthropogenic impacts. In the
1850s hydraulic mining started after placer mining in rivers became
unproductive. Hydraulic mining remobilized a huge amount of sediment
upstream of Sacramento. By the end of the nineteenth century the hydraulic
mining was outlawed leaving approximately 1.1
Typical SSC in the Delta ranges from 10 to 50 mg L
Several studies have been carried out to determine sediment pathways and to estimate sediment budgets in the Delta area (Schoellhamer et al., 2012; Jaffe et al., 2007; Gilbert, 1917; McKee et al., 2006, 2013; Wright and Schoellhamer, 2005). These studies were based on data analysis and conceptual hindcast models. Although the region has a unique network of surveying stations, there are many channels without measuring stations. This might lead to incomplete system understanding and knowledge deficits for the development of water and ecosystem management plans. The monitoring stations are located in discrete points hampering spatial analysis. Also, the impact of future scenarios related to climate change (i.e., sea level rise and changing hydrographs) or different pumping strategies remains uncertain.
Structured grid models such as Delft3D and ROMS (Regional Oceanic Modeling System) have been widely used and accepted in estuarine hydrodynamics and morphodynamics modeling including studies of the San Francisco estuary (Ganju and Schoellhamer, 2009; Ganju et al., 2009; van der Wegen et al., 2011). In all of these studies the Delta was schematized as two long channels because the grid is not flexible, which would have allowed for efficient 2-D modeling of the rivers, channels and flooded island of the system together with the bay.
In cases with complex geometry, unstructured grids or a finite volume model is more suitable. There are three widely known unstructured grid models: (1) the TELEMAC-MASCARET (Hervouet, 2007), (2) the Unstructured Tidal, residual, intertidal mudflat model (UnTRIM) (Casulli and Walters, 2000; Bever and MacWilliams, 2013) and (3) Delft3D Flexible Mesh (D3D FM) (Kernkamp et al., 2010). The first two models are purely triangle based and are not directly coupled (yet) with sediment transport and/or water quality and ecology models.
Location of the San Francisco Bay–Delta. The black rectangle highlights the Delta, and the red squares indicate measurement stations.
Numerical mesh for the D3D FM model. Red dots indicate the
calibration stations (
The numerical model applied in this work is D3D FM. D3D FM allows for straightforward coupling of its hydrodynamic modules with a water quality model, Delft-WAQ (DELWAQ), which gives flexibility to couple with a habitat (ecological) model. D3D FM is a process-based unstructured grid model developed by Deltares (Deltares, 2014). It is a package for hydro- and morphodynamic simulation based on a finite volume approach solving shallow-water equations applying a Gaussian solver. The grid can be defined in terms of triangles, (curvilinear) quadrilaterals, pentagons and hexagons, or any combination of these shapes. Orthogonal quadrilaterals are the most computationally efficient cells and are used whenever the geometry allows. Kernkamp et al. (2010) and the D3D FM manual (Deltares, 2014) describe in detail the grid aspects and numerical solvers.
The bay area and river channels are defined by consecutive curvilinear grids
(quadrilateral) of different resolution. Rivers discharging in the bay, and
channel junctions are connected by triangles
(Fig. 2). The average cell size ranges from
1200 m
We assume that the main flow dynamics in the Delta is 2-D, which does not
account for vertical stratification. The Delta does not experience
salt–fresh water interactions due to the pumping operations and we assume
that temperature differences between the top and bottom of the water column
do not govern flow characteristics. D3D FM generates hydrodynamic output for
off-line coupling with water quality model DELWAQ (Deltares, 2014). Off-line
coupling enables faster calibration and sensitivity analysis. D3D FM
generates time series of the following variables: cell link area; boundary
definition; water flow through cell link; pointers that give information
about neighbors' cells; cell surface area; cell volume; and shear stress
file, which is parameterized in D3D FM using Manning's coefficient. Given a
network of water levels and flow velocities (varying over time) DELWAQ can
solve the advection–diffusion–reaction equation for a wide range of
substances including fine sediment, the focus of this study. DELWAQ solves
sediment source and sink terms by applying the Krone–Parteniades formulation
for cohesive sediment transport (Krone, 1962; Ariathurai and Arulanandan,
1978) (Eqs. 1 and 2).
The Bay–Delta is a well-measured system; therefore, all the input data to the
model are in situ data. Initial bathymetry has 10 m grid resolution, which
is based on an earlier grid (Foxgrover et al., 2012,
The hydrodynamic model has been calibrated for the entire Bay–Delta system
(see Appendix A and
In this study we applied five open boundaries. Water levels at the seaward
boundary are based on hourly measurements from the Point Reyes station
(
The river water flow hourly input data at the FPT, the VNS and YOLO were
obtained from the California Data Exchange Center website (
Input boundary conditions. The top panel is water level at Point Reyes. The lower three panels show discharge in a dashed blue line and SSC in a solid green line for Sacramento River at FPT, San Joaquin River at VNS and Mokelumne River at Woodbridge, respectively.
Calibration station locations (top panel) and comparison of model outputs and measured data. Left panels show SSC calibration and right panels the show discharge. Data are dashed red lines and model results are solid blue lines. Note that in the discharge plots of RVB and STK the data line is behind the model line.
Since 1998, USGS has continuous measuring stations for sediment concentration which is derived from backscatter sensors (OBS) measurements every 15 min, and are calibrated approximately monthly with bottle samples (Wright and Schoellhamer, 2005). This type of sensor converts scattered light from the particles to photocurrent, which is proportional to SSC. To define the rating curve it is necessary to sample water, filter it and weigh the filter. However, in some locations the cloud of points when correlating photocurrent and filtered weight shows a large scatter. Large scatter leads to errors in converting photocurrent to SSC. The causes for errors include variation in particle size, particle desegregation (cohesiveness, flocculation, organic-rich estuarine mud), particle shape effects and sediment-concentration effects (Kineke and Sternberg, 1992; Downing, 2006; Sutherland et al., 2000; Gibbs and Wolanski, 1992; Ludwig and Hanes, 1990). Wright and Schoellhamer (2005) showed that for the Sacramento–San Joaquin delta these errors can sum up to 39 %, when calculating sediment fluxes through Rio Vista.
In this work we modeled the 2011 WY – 1 October 2010 to 30 September 2011. First, we ran D3D FM for this year to calculate water level, velocities, cell volume and shear stresses. Then, the 1 year hydrodynamic results were imported in DELWAQ which calculated SSC levels.
The SSC model results are compared to in situ measured SSC data. The
calibration process assesses the sensitivity of sediment characteristics
such as fall velocity (ws), critical shear stress (
Our focus is to represent realistic SSC levels capturing the peaks, timing
and duration, and to develop a sediment budget to assess sediment trapping
in the Delta (Fig. 1, highlighted by the black rectangle). Throughout the
following sections the results are analyzed in terms of tide-averaged
quantities by filtering data and model results to frequencies lower than
2 days. We applied a Butterworth filter with a cutoff frequency of
1
The results shown below are the derived from an extensive calibration
process where the different sediment fractions parameters (ws,
The best fit of the calibration process (uRMS
The 2011 WY simulation reproduces the SSC seasonal variation in the main Delta regions such as the north (Sacramento River) represented by Rio Vista station (RVB), the south (San Joaquin River) represented by Stockton (STK), central-east delta represented by Mokelumne station (MOK) and delta output represented by Mallard Island (MAL) (Fig. 4).
All stations clearly reproduce SSC peaks during high river flow from November to July, and lower concentrations during the remainder of the year (apart from MAL during the July–August period). The good representation of the peak timing indicates that the main Delta discharge event is reproduced by the model as well as the periods of delta clearance. These two periods are critical for ecological models, and a good representation generates robust input to ecological models. The differences found between the model and data are further discussed in Appendix B.
We considered one fraction for simplicity and because it reproduces more than 90 % of the sediment budget throughout the Delta as well as the seasonal variability of SSC levels. Although more mud fractions considerably increase running time, several tests with multiple fractions were done to explore possibilities for improving the model results.
Including heavier fractions changes the peaks timing and also lowers the SSC
curve. Comparing the standard run (ws
To be able to find a single best parameter setting a sensitivity analysis was done varying the main parameters in the Krone–Parteniades formulation (Table 1). Regarding sediment flux, these tests show that RVB and MAL are more sensitive to parameter change than STK (Fig. 5). The model results are most sensitive to the critical shear stress for erosion and least sensitive to the erosion coefficient. Analyzing the time series, one concludes that in stations where the fluxes are higher, the change in critical shear stress is less important, since during most of the time the shear stress is already higher than any given critical shear stress.
Parameters set of sensitivity analysis.
Sensitivity analysis for sediment flux at RVB on the Sacramento River (green squares), at STK on the San Joaquin River (red triangles) and at MAL where the Delta meets the Bay (blue circles). The colored lines indicate the data values.
Statistical metrics for sensitivity runs.
We quantify error using two metrics, the unbiased root mean square error
(uRMSE; Fig. 6) and skill (skill;
Fig. 6) (Bever and MacWilliams, 2013).
The uRMSE indicates the variability of the model relative to the data and
is zero
when the model and data have equal variability.
Skill is a single quantitative metric for model performance (Willmott,
1981). When skill equals 1 the model perfectly reproduces the data. The
two metrics where evaluated at RVB, STK and MAL, representing respectively
Sacramento River, San Joaquin River and delta output.
To study the importance of initial bottom sediment availability we considered two cases: one excluding sediment (no sediment available at the bed) and the other with mud at places shallower than 5 m b.m.s.l., the same setting as the standard run.
We did some tests varying the 5 m threshold. From 3 to 10 m the final results are all similar. However, allowing mud availability in the channels deeper than 10 m starts to affect the SSC levels. Time series of SSC comparing the two cases show that bottom composition has virtually no influence on SSC after the first couple of days. This result also applies for different mud fractions availability and suggests it may be possible to accurately model less-measured estuaries where virtually no bottom sediment data are available.
Another test shows that it is better to initialize the model with no sediment at the bed than with mud available in the entire domain. Initializing the channels with loose mud generates unrealistically high SSC levels through the years, which can take up to 5 years to be reworked.
In the previous section we presented the model calibration, a normal practice in the modeling process. In this section we discuss the new insights that were derived from the model results. Although these insights are specific to the San Francisco Bay–Delta system, the same approach can be applied to other estuaries and deltas. The model produces detailed sediment dynamics and the main paths in which sediment is transported in the Delta. Sediment flux calculations define the sediment dynamics, while gradients in sediment describe the sediment distribution and deposition pattern in the Delta. We also discuss daily and seasonal variation of turbidity levels.
Anomaly of a SSC (mg L
Water discharge
We start the analysis by exploring the general Delta behavior. During dry
periods SSC in the entire Delta is low (
Three Mile slough (TMS) and the delta cross channel (DCC) connect the
Sacramento River with the central and eastern Delta. Model results show that
together they carry 60 Kt per year of sediment southward. DCC operation
controls SSC levels in the eastern/central Delta to a large extent. To show
the importance of the DCC we run the model twice, once with the DCC always
open and once always closed. When the DCC is open, high SSC Sacramento River
water (
During peak river discharge, Sacramento River sediment reaches Mallard
Island in approximately 3 days, Carquinez straight in 5 days and the Golden
Gate Bridge in approximately 10 days. This timing is proportional to river
discharge. However, from Mallard Island seawards this estimate is inexact
due to the 2-D approximation. San Joaquin River sediment remains largely
trapped in the southern delta. The flooded islands, breached levees like
Franks Tract, present a different behavior. During the entire year the SSC
levels are below 15 mg L
Sediment flux is a useful tool for a quantitative and qualitative analysis
of the sediment pathways and its derivative gives sedimentation/erosion
patterns. Sediment flux is defined by the product of water velocity (
Seaward from MAL considerable salt–freshwater stratification takes place in the water column. These 3-D effects are not captured by our 2DH approach and model results in this region are inaccurate. Therefore, Fig. 8 shows preliminary sediment flux to the bay by a dashed line.
From the previous section one can see that more sediment enters
(
Besides the overall spatial trend, different parts of the Delta have
different trapping efficiencies: the northern Delta (the least efficient)
traps
Of the total Sacramento River sediment input 40 % stays in the northern Delta and about 40 % is exported to the Bay. The remaining 20 % deposits in the central/eastern delta and only 2 % travels all the way to the south Delta. About 70 % of San Joaquin sediment deposits in the southern Delta, 10 % goes to the central Delta, 15 % is exported via Clifton court pumping facilities and 5 % is exported to the bay. This transport is reflected in the bottom composition of the Delta. Sacramento River sediment dominates the northern and central Delta and San Joaquin River sediment dominates the southern delta bottom composition (Fig. 9).
It is enlightening to divide the sediment budget analysis into wet and the dry seasons, since the delta has different dynamics for each season. Water year 2011 was a wet year, with the wet season lasting from mid-January until the end of May. During the wet period 60 % of the yearly sediment input budget entered the delta through FPT and VNS and 70 % of the yearly budget was exported through MAL. In the wet season the high river water discharges and SSC pulses flush the entire delta with sediment. In this season high SSC gradients are observed in the plume fronts leading to rapid changes in habitat conditions for many species. After the front the high SSC level can last for more than 1 month, indicating changing in habitat conditions
During the dry season the delta experiences lower river discharges and SSC levels resulting in lower sediment transport rates. In the dry season SSC levels do not have peaks and are more uniform. During the dry season the water is clear and the advective flux is lower, which will be discussed in the next section.
SSC peaks at FPT can be tracked down the estuary. At the RVB station the SSC peak follows the same dynamic as that observed at FPT; however, this behavior does not apply for the entire delta. Schoellhamer and Wright (2005) observed that the river signal is attenuated through the estuary. This attenuation can be understood by analyzing changes in the dominant sediment flux component.
Sediment bottom composition after 1 year, starting with no bed sediment available. Red shades indicate dominance of Sacramento River sediments and white shades dominance of San Joaquin River sediments. The black line highlights where this separation occurs.
Sediment flux calculations for several stations within the Delta.
Dyer (1974) decomposed the tidally averaged fluxes in three main
components: tidal mean, the advective term; tidal fluctuation, the
dispersive term; and the Stokes drift. This decomposition was possible
considering that the measured valued is the sum of a tidally mean component
Sacramento River at FPT, the most landward station, experiences no tidal influence so the flux is purely advective. At RVB, which is seaward, there are tidal fluctuations and the dispersive flux is responsible for 22 % of the total flux; however, no Stokes drift flux is present (Fig. 10). In contrast, Stokes drift accounts for 33 % of the total flux in MAL station implying that tides have a bigger influence in this region.
An analogue can be drawn to the San Joaquin branch, where VNS and STK experience only advective terms. At MOK and SJJ dispersive (20 and 63 %, respectively) and Stokes flux (5 and 11 %) start to influence the total flux (Fig. 10). The analyses of the three different flux components in smaller Delta channels show that river and tidal signals are equally important. The river peak signal is less important inside smaller channels than in rivers. At TMS, the dispersive flow accounts for 60 % of the total flux.
The flux analyses show that there is no change in the Delta net circulation when comparing wet and dry seasons. There is not a major change in the flux direction when comparing the seasons. However, there is a change in importance of each flux component.
Modeled deposition in millimeters for 1 year period.
Figure 10 shows that dispersive flux and Stokes drift
relative contributions vary seasonally. When river discharge is high the
relative contribution of dispersive flux is lower than during low flow
conditions. This pattern is more apparent at stations where the river signal
is stronger. At RVB the dispersive flux contribution is about 15 % during
the wet season and 26 % in the dry season; the same applies for MAL and
STK. In smaller channels, like TMS and SJJ, the dispersive flux seasonal
variation is milder, varying about 10 %, from 55 % in the wet season to
65 % in the dry season. In the dry season the change in flux
contributions, from advective to dispersive and Stokes drift, leads to a
lower net export of sediment from the Delta, even though the concentrations
in the Delta are only about 30 mg L
The flux changes from completely advective to dispersive and Stokes drift sheds some light on the Delta deposition areas. The places where the dispersive flux starts to play a role, near RVB and MOK, are the same places where net deposition is observed (Fig. 11). Other locations where considerable sedimentation takes place are in flooded island areas, such as Franks Tract and the Clifton court. The 2-D model is sufficient for such areas (Fig. 11).
The San Joaquin River downstream of Stockton experiences high deposition. This finding is confirmed by constant dredging needed to maintain the Stockton navigation channel. The river discharge modulates the deposition pattern in the main channels. In the Sacramento deposited sediment is gradually washed away and transported to the mud flats at the channel margins, until the next peak. At the flooded islands the sedimentation process is gradual and steady, erosion is not observed in these areas.
Turbidity in each Delta region. For each region, the left bars indicates wet season and the right bars the dry season. The light gray bars indicate the mean turbidity over the region, the darker bars the spatial deviation and the lines the daily deviation. Each horizontal line represents 10 NTU.
Deposition is primarily observed during the wet and dry season. Some exceptions occur in small bends in the Sacramento River that are erosional during the wet season and depositional during the dry season. The deposition pattern provides insight into the best areas for marsh restoration.
So far the discussion presented is in terms of SSC levels for the standard
run, budgets and fluxes, while ecological analysis is often based on
turbidity levels. SSC and turbidity are correlated by rating curves as
In this section we present average values for turbidity within a specific
Delta region as well as its seasonal and daily variations (Fig. 12).
Generally, the mean turbidity levels and spatial variations are higher
during the wet season than during the dry season. During the wet season, the
southern area had the highest mean value (50 NTU), and deviation (15 NTU),
caused by a combination of large sediment supply and low flow velocities.
The Northern region is the second most turbid area (45
The DCC and Georgina Slough (GLS) channels that connect the Sacramento and San Joaquin rivers are important bridges to export sediment from the Sacramento River to the eastern Delta. The smaller channels of the network play a minor role in the Delta sediment budget because the discharges in these channels are considerably smaller than in the rivers.
As a well surveyed area that now has a complex process-based model, the Delta offers the opportunity to test how much data are necessary to develop a reliable sediment model. The model supports high temporal and spatial resolution and includes multiple physical processes such as bottom friction, sedimentation and erosion. The available data allow for calibration and validation of model results.
As presented above, with simple settings of one mud fraction and simple bed sediment availability the model is capable of representing the main sediment dynamics processes, the peak timing and duration, and results in a sediment budget. The data necessary for accurate modeling and forecasting are fine resolution bathymetry to correctly reproduce hydrodynamics, SSC and discharge at the inflow and outflow boundaries. It is necessary as well to have one to two stations in the domain in order to properly calibrate the model. The results from the calibrated model using these few data can be extrapolated for the entire domain, allowing for closing the sediment budget for the whole system.
The 2-D model results output is available in high temporal (
In this work we make a step towards understanding and simulating sediment dynamics from source to sink in a complex estuary. This work shows that it is possible to reproduce the main system sediment dynamics as well as construct an accurate detailed budget for complex areas such as the Delta using a 2-D process-based numerical model coupled with a water quality model.
Overall, the model reproduces the SSC peaks and event timing and duration (wet season) as well as the low concentration in dry season throughout the Delta, except at Mallard where the water column is stratified due to salt intrusion. Stratification issues are not solved in a 2-D model. For this reason we are working on a 3-D model in order to include the bay area, leading to a unique source to sink model.
The Delta has many observation stations. However, this work shows that
substantial sediment is exported trough the pumping stations (100 kt yr
We show that with simple sediment settings of one fraction at the input boundary and a simple distribution of bed sediment availability, it is possible to reproduce seasonal variations as well as construct a yearly sediment budget with more than 90 % accuracy when compared with a data derived budget. It also shows that it is extremely important to have discharge and SSC measurements at least in the input boundaries and close to the system output in order to be able to calibrate the model settings applied for hydrodynamics and suspended sediment. This methodology now can be applied in less-measured estuaries.
Sediment is a key factor in the water quality and ecology of an estuary. The D3D FM software allows for direct coupling to water quality, sediment transport and habitat modeling. Our work provides the basis to a chain of models, which goes from the hydrodynamics, to suspended sediment, to phytoplankton, to fish, clams and marshes. The turbidity and deposition pattern analysis may guide ecologists in future works to define areas of interest and/or venerable areas to be study, as well as guide data collecting efforts. The present model opens the possibility for forecast and operational modeling. Forecasting the time frame of high levels of SSC (turbidity) allows for planning of measurements campaigns for ecologists, as well as the possibility of tracking potentially contaminated sediment and be able to make a contingency plan as well as temporary barriers and pumping operations.
The Sacramento–San Joaquin delta is a typical case of a highly impacted estuary. Being able to numerically simulate and determine sediment transport, budget and turbidity levels in this type of environment open possibilities to better informed political, ecological and management decisions including how to respond to climate change and sea level rise. This type of model is an important management tool that is applicable to other impacted estuaries worldwide.
The hydrodynamic calibration was carried out for 3-month high river flow conditions (16 December 1999 to 16 March 2000) and a 3-month period of low river flow conditions (16 July 2001 to 16 October 2001). All data are in NAVD88 (vertical datum), UTM 10 (horizontal datum) and GMT (time reference).
Hourly measured water levels at Point Reyes (
Measured data for the bay area were obtained
from
Calibration was carried out by systematically varying the value of the Manning's coefficient for different sub-areas of the Bay–Delta system. The calibration data analysis includes (local and time varying) influence of air pressure and wind in the definition of the boundary condition as well as in the calibration data inside the modeling domain. These may account for (part of) the error between measurements and modeling results. Also, the NAVD88 reference is not known for all measurement stations, although tidal water fluctuations may be modeled properly. To avoid these problems, a better method to assess the model performance is to focus on water level amplitude and phasing of the different tidal constituents. Boundary conditions, calibration data and model results are thus decomposed by Fourier transformation into tidal components which are then compared. By far, the main tidal constituents at Golden Gate (GGT) are O1, K1, N2, M2 and S2, with M2 being the largest. The model represents their values quite well. The difference in amplitude is 1.3 % for M2, up to 14 % for O1, but the phasing shows a maximum of only 3 % (O1)).
Figure A1 gives calibration results for the high and low river flow. The
largest (extreme) deviations are explained by the fact that the measured
water levels did not have a known reference to NAVD88 (
Hydrodynamic calibration example.
All stations clearly reproduce SSC peaks during high river flow periods and lower concentrations during the remainder of the year (apart from MAL during the July–August period). The good representation of the peak timing means that the main delta event is reproduced by the model as well as the periods of delta clearance. These two periods are critical for ecological models, and a good representation generates robust input to ecological models. A closer look at Fig. 4 reveals differences between model results and data. These differences are discussed station by station in this Appendix.
Comparison between SSC levels in RVB station in situ data (dashed red) and model result (solid blue) and FPT station (dotted green).
At RVB, SSC levels are directly proportional to Sacramento River discharge (Fig. B3), and that the model properly represents the water discharge peak intensity and duration. However, in the model, the first peak, which occurs in October, remobilizes sediment faster than observed in the data. Analyzing the raw data, it is possible to observe a trend of SSC increase which the model overestimates. A probable explanation lies in the initial sediment composition of the bed. Defining the bottom sediment composition does not account for consolidation processes; therefore, the first peak comes after the dry season when the mud in the banks has consolidated. In the simulation case, when river discharge increases, it remobilizes non-consolidated bottom/bank sediment causing an earlier peak than in the data. Similar behavior is observed at STK in December. Sediment trapped in sub-aquatic vegetation and marshes could be another explanation for the slower increase of the first peak as the model discharges for both stations agree with data (Fig. 4).
Another difference between the data and the model results at RVB is the peak in May (second rectangle, Fig. B1), which is not observed in the data. SSC level at the RVB station is directly proportional to water discharge in FPT (Fig. B3, RVB). The May peak is observed in FPT and so should have been transported towards RVB just as the two preceding peaks. However, the data set does not reproduce this peak. One of the possible explanations is an error in measurements, since it comes after a major event and the equipment might be damaged. Other explanations could be a different composition of the suspended sediment properties and/or flocculation.
The model underestimates the first and second SSC peaks at MOK. However, the measured SSC signal is not consistent with the local water discharge signal. First, we checked that modeled water discharge is reproducing the local conditions, where data are available from mid-February onwards. The last peak in Fig. 4 (mid-March) shows that water discharge, in situ and modeled SSC have the same rage of variation. Therefore, the SSC levels are proportional to the local water discharge. Earlier, the January SSC data peak is much higher than the water discharge and the SSC level calculated in the model. The same happens in mid-February when no water discharge peak is observed but there is a peak in the SSC data. Again the peaks in SSC could be caused by an error in the measurements or local, diffuse input of sediment such as from local farm waste water or biological activity remobilizing the substrate.
Water discharge (model) and SSC level (data and model) in MOK station.
The model represents the wet season SSC peaks well at MAL; however, during
the three drier periods of the year the model underestimates SSC levels (Fig. B2).
From the scatter plots of water discharge versus SSC (Fig. B3), it is
possible to explain the weaker performance of the model during low river
flow at MAL. These graphs represent river water discharge in FPT lagged by
2 days to SSC in RVB and MAL. Several time lags were tested, as MAL does not
have a reasonable correlation with any of the time lags; it is presented
here with the same time lag as the one for RVB. RVB station reflects a
positive correlation between river discharge and SSC derived from in situ
data and model results. The correlation coefficient (
Scatter plot discharge versus SSC shown for MAL station on the left-hand (MAL) side and on the right-hand (RVB) side for RVB station. The red dots represent the data and the blue model results.
At the MAL station
The research is part of the US Geological Survey Computational Assessment of Scenarios of Change for the Delta Ecosystem (CASCaDE) climate change
project (CASCaDE contribution 60). The authors acknowledge the US Geological
Survey Priority Ecosystem Studies and California Federation Bay-Delta Program (CALFED) for making this research
financially possible. The data used in this work are freely available on the
USGS website (