Content uploaded by Thomas J. Mathis
Author content
All content in this area was uploaded by Thomas J. Mathis on Jun 19, 2019
Content may be subject to copyright.
Estuarine Salinity Intrusion and Implications for Aquatic Habitat:
A Case Study of the Lower St. Johns River Estuary, Florida
Thomas J. Mathis1, Szu-Ting Lee1, Paul M. Craig, P.E.1, Nghiem Tien Lam, Ph.D.2, Kester
Scandrett2, Anurag Mishra, Ph.D.1, Rumana R. Arifin, Ph.D.1, and Jeffrey Y. Jung, Ph.D.1
1DSI, LLC, P.O. Box 1916, Edmonds, WA 98020, USA; e-mail: tmathis@ds-intl.biz
2DSI, LLC, No 24, Lane 535, Lac Long Quan, Tay Ho District, Hanoi, Vietnam; e-mail:
kscandrett@ds-intl.biz
ABSTRACT
The Lower St. Johns River estuary in Florida has experienced both urban and exurban
development along its banks over the course of many decades. The extent of salinity intrusion
into the estuary is controlled by a combination of factors which may be altered through natural or
anthropogenic forces. To better understand the factors controlling salinity intrusion, the
Environmental Fluid Dynamics Code Plus (EFDC+) was applied to model the hydrodynamics of
the estuary. This study discusses the calibration and validation of the model, analysis of the
extent of salinity intrusion, and application of habitat analysis using the EFDC_Explorer
Modeling System (EEMS). The results of this study showed that the extent of salinity intrusion
in the LSJR estuary is primarily controlled through varying combinations of upstream flow,
long-term variations in the open ocean, and wind drag. Results of the habitat analysis concerning
submerged aquatic vegetation and wetlands constitute a baseline for comparison with alternative
management and future change scenarios for the basin.
INTRODUCTION
The Lower St Johns River (LSJR) estuary is dominated by wetlands (greater than 50% of the
floodplain area) which host a diverse range of ecological communities. Both the composition and
structure of these wetlands exhibit a strong geographical pattern which correspond to gradients in
tidal amplitude, salinity, and soil texture and composition (Kinser, et al., 2012).
Increasing urban and exurban development over the last 60 years have already shown impacts
throughout the LSJR estuary and floodplain. For example, dredging in the Intracoastal
Waterway, deepening of the main river channel, and other navigational changes combined with
sea level rise appear to have driven a retreat of the tree line in many tributaries near the river
mouth. Other areas have been observed to have increased salinity throughout the year and have
seen a shift to more salinity-tolerant plant species (Kinser, et al., 2012).
The Jacksonville Harbor feasibility study (USACE, 2014) demonstrated that the partially-
stratified LSJR estuary required a three-dimensional hydrodynamic model to accurately simulate
salinity variations with depth. The purpose of this study is to evaluate potential impacts of
salinity intrusion into the estuary on the aquatic habitat and create a basis for evaluating multiple
management scenarios to support decision-making.
To model the LSJR estuary, the Environmental Fluid Dynamics Code Plus (EFDC+; DSI, 2018)
hydrodynamic and salinity model was utilized. The EFDC+ vertical layering option used for this
study was the Sigma-Zed (SGZ) approach (Craig et al., 2014), which allows for the number of
layers to be specified at every horizontal cell. The SGZ approach in combination with the
EFDC+ capability of using multiple CPU cores resulted in an up to 75% reduction in model run
times compared with single thread applications.
Five separate simulation periods were used for model calibration and validation. Data collected
by USACE, NOAA, USGS, and the St Johns River (SJR) Water Management District were used
to calibrate the EFDC+ model for water surface elevation and salinity. This paper summarizes
the model calibration and validation and provides analyses of the habitat suitability for various
aquatic plant species which are sensitive to long-term changes in salinity concentration.
Following model calibration, an 8.5-year Production Run (PRP) from January 2008 to June 2016
was run in order to evaluate the long-term trends in salinity intrusion and analyze the long-term
behavior of the estuary in relation to salinity stress on locations in the estuary with submerged
aquatic vegetation beds (SAV), and wetlands.
METHODOLOGY
Study Site
The domain selected for the EFDC+ model extends from the open ocean, approximately 25 miles
offshore, to Lake George and Crescent Lake which are approximately 120 miles upstream on the
Lower St Johns River (Figure 1). Tidal marshes surround the lower part of the estuary, where the
river is mostly channelized due to urban development around Jacksonville, Florida (riverine
segment). Upstream of Jacksonville (Acosta Bridge), the river opens into a lacustrine transitional
zone (lacustrine segment) between the saline and freshwater parts of the estuary. Approximately
60 miles upstream near Palatka, Florida the river again becomes more channelized through low
lying marsh areas surrounding Lake George and Crescent Lake.
Measured Data
Water surface elevation and salinity concentration were measured at stations around the model
domain which were maintained through a cooperation between USACE, NOAA, USGS, and the
St Johns River (SJR) Water Management District. The locations of the gaging stations and the
parameters being measured at each station have been illustrated in Figure 1. Measured data was
used to assess model performance through a combination of statistical performance metrics (i.e.
R2, and Index of Agreement) and qualitative assessment. To describe model fit qualitatively
using these metrics, we modified the performance rating guidelines developed by ASABE
(2017).
Bathymetry
Four separate bathymetric data sets have been collected for the study area over the last 20 years;
1998, 2003, 2010 and 2012. The 2012 data represents the “Existing Conditions” bathymetry.
The harbor specific data sets were merged with topo-bathymetric LiDAR collected in 2003 to
create a complete model bathymetry. The average depth of all model cells was 5 meters, and the
maximum depth was 23 meters. Wetland cells are typically between 0 and 5 meters, while the
main stem stations tend to be between 5- and 15-meters depth. For the current study, all
bathymetry and water surface elevations were converted to NAVD88 for a consistent vertical
datum. These data were then interpolated onto the horizontal grid discussed earlier to produce
the model bathymetry used by EFDC.
Figure 1. Geographical overview of the Lower St. Johns River model domain, with data
and wind stations indicated by callout boxes.
Predicted Tributary Inflows
The Feasibility Study used the Hydrologic Simulation Program Fortran (HSPF, Bicknell et al.
2001), as implemented by SJRWMD’s Water Supply Impact Study (WSIS) to compute tributary
inflows for the EFDC+ model (Taylor, 2017). There are 149 flow input locations to EFDC+ that
correspond to hydrologic basin outflows along the LSJR and tributaries. There are an additional
16 flows, for a total of 165 flow inputs to EFDC+, from springs and waste water treatment
plants.
Model Description
The Environmental Fluid Dynamics Code Plus (EFDC+) was selected as the framework for the
LSJR hydrodynamic and salinity model. The EFDC+ model provides general-purpose modeling
packages for simulating three-dimensional (3D) flow, transport, and biogeochemical processes in
surface water systems including rivers, lakes, estuaries, reservoirs, wetlands, and near shore to
continental shelf scale coastal regions (Hamrick, 1996; DSI, 2018). EFDC+ solves the Reynolds-
Averaged Navier-Stokes equations in 3-dimensions using an explicit finite difference numerical
solution on an Arakawa C-grid (Hamrick, 1996; Arakawa and Lamb, 1977). For this study, the
method of Smagorinsky (1963) method was used for horizontal momentum diffusion, and the
method of Galperin et al. (1988) was used for turbulence closure.
The version of EFDC used for this project, EFDC+ (formerly known as EFDC_DSI),
incorporates several enhancements to the base EPA EFDC code. DSI has developed a new
vertical layering approach for the EFDC model (Craig, et al., 2014). This new vertical layering
scheme, referred to as Sigma-Zed (or SGZ), allows for varying number of active layers from one
to KC (maximum number of layers for the run) in each cell and computes density metrics on
each face, matching the number of active layers to the adjacent cell. SGZ was found to be more
computationally efficient than a similarly configured Sigma Stretch grid, allowing models with
multiple layers to be developed and utilized for large natural systems (DSI, 2018).
Horizontal and Vertical Grid Development
The horizontal grid was developed with the CVLGrid 2-D grid generation tool (DSI, 2017). A
combination of aerial imagery and GIS data were used to generate curvilinear grids along the
main stem and tributaries of the Lower St. Johns River, from the open ocean to Lake George and
Crescent Lake. In total, the horizontal grid contains 18,525 grid cells.
The vertical grid was generated using the SGZ layering method for each bathymetric model used
in the calibration and validation. The main stem of the model used four layers, the wetlands used
one layer and the tributaries used a mix of 1 to 4 layers. This produced a nominal cell depth of 6
meters. Due to the SGZ layering method, the bottom active cell can be associated with any layer
in the model, depending on the bathymetry. Through sensitivity testing, this layering
configuration was found to provide adequate vertical resolution, while minimizing model run
times.
Initial Conditions
To apply appropriate initial conditions for water level and salinity for each model, the models
were configured to start up to six months prior to the actual model simulation start (listed in
Table 1), otherwise referred to as a spin-up run. This allows the model domain to become
representative of the actual conditions at the start of the simulation.
Boundary Conditions
Tributary discharge was based on the results of the HSPF model for the LSJR basin (Taylor,
2017). In total, 165 separate discharge boundary conditions were configured in the model.
For the open ocean boundary, there was a lack of appropriate measured data to specify the water
level at the boundary. Additionally, there are multiple model boundaries which connect to the
open ocean or other embayments which must be considered. To address this issue, water surface
elevation data measured at Mayport was lagged and scaled to account for the variations in timing
expected due to the varying position of each of the 6 open boundaries in the model.
The salinity boundary conditions were constant in time but varied by type of boundary. The
salinity for rivers used a constant of 0.5 ppt. The open boundaries used two different constant
salinities, 35 ppt for the ocean and 30 ppt for the intracoastal waterway.
Five wind stations were used to specify surface boundary conditions for the LSJR EFDC model.
The EFDC model used station weighting for each cell depending on its location and its data
availability. The station weighting was computed by EFDC_Explorer’s station weighting tool
using an inverse distance algorithm (power 2).
CALIBRATION AND VALIDATION
Five short-term model run periods were utilized to calibrate and validate the LSJR EFDC model.
Calibration periods were selected for years that exhibited wet and dry conditions, and one year
which exhibited average conditions. Wet and dry conditions were determined by examining the
period of record for discharge at Acosta Bridge. Calibration and validation periods selected for
this study are shown in Table 1.
Table 1. Selected calibration and validation periods for the LSJR study.
Period
Calendar Dates
Wet/Dry Year
Run Type
Bathymetry
Start
End
1998 - 1999
1998/11/30
1999/03/31
Dry
Calibration
1998
2003
2003/04/02
2003/08/01
Wet
Calibration
2002/ 2003
2015 - 2016
2015/12/01
2016/06/30
Average
Calibration
2012
2010 - 2011
2010/11/30
2011/04/01
Dry
Validation
2010
2014
2014/08/01
2014/12/01
Wet
Validation
2012
The parameters of interest in calibration include the bottom roughness (z0), horizontal eddy
viscosity (AHO), horizontal eddy diffusivity (AHD), vertical eddy viscosity (AVO) and
molecular diffusivity (ABO). These values were adjusted iteratively throughout the calibration
process, and values were selected based on overall model performance using the coefficient of
determination (R2) and index of agreement (IOA). A single set of parameter values were selected
based on the model’s performance during the calibration periods. The generality of these
calibrated parameter values is shown by the model’s performance during the validation periods.
Bottom roughness (z0) varied spatially with the principal values in the deeper parts of the main
stem, and the shallower wetland cells using a z0 value of 1E-4 m, and 0.025 m respectively.
Through the lower reach of the main stem from the mouth to Acosta Bridge, the model used a
combination of values with z0 in the harbor area having a value of 0.01 m up to Dames Point,
and a value of 1E-3 m from Dames Point to Acosta Bridge. The z0 value of 0.025 m in the
wetlands was intended to reflect the bulk hydrodynamic behavior of vegetation.
AHO varied spatially in the model as a function of cell area with a minimum background value
of 1E-3 m2 s-1. AHD used a single uniform value of 0.05 (dimensionless). AVO and ABO used
values of 1E-4 and 1E-5 m2 s-1 respectively.
For all the calibration and validation runs, the model achieved very good fit for water surface
elevation (R2 > 0.9, IOA > 0.8; Table 2). The model performance for the 1998-1999 and 2003
runs deteriorated slightly near the upstream extent of the model domain (Buckman Bridge and
upstream) with IOA dropping below 0.8, however the later calibration and validation years (2010
onward) showed very good fit at all stations. The model performance in 1998-1999 and 2003
may have been affected by the lack of high-quality bathymetric data collected prior to these
simulation periods.
Table 2. Index of Agreement (IOA) for water surface elevation for calibration and
validation periods. R2 values are not provided for conciseness but are generally closer to 1
than corresponding IOA values.
Station
Calibration
Validation
1998-1999
2003
2015-2016
2010-2011
2014
Mayport
0.956
0.882
0.960
0.961
0.966
Dames Point
0.895
-
0.920
-
0.949
Acosta Bridge
-
-
0.899
0.873
0.887
Main Street Bridge
0.849
0.846
0.900
-
0.901
Buckman Bridge
-
0.770
0.893
0.877
0.905
Red Bay Point
0.767
0.765
0.895
-
0.868
Racy Point
-
0.757
0.860
-
0.878
Modeled salinity concentrations showed a good fit with observed conditions (R2 > 0.7, IOA >
0.6) at most main stem stations. In terms of the characteristics of the salinity intrusion, the
salinity concentration at Buckman Bridge represents the best location by which to measure
model performance. For all calibration and validation runs, model fit at Buckman Bridge shows
it to provide a satisfactory reflection of observed conditions (R2 > 0.6, IOA > 0.6).
Table 3. Index of Agreement (IOA) for salinity for calibration and validation periods. R2
values are not provided for conciseness but are generally closer to 1 than corresponding
IOA values.
Station
Calibration
Validation
1998-1999
2003
2015-2016
2010-2011
2014
Mayport
-
0.614
0.300
-
0.553
Dames Pt - Top
-
-
0.727
0.659
0.603
Dames Pt - Bot
-
-
0.530
0.506
0.682
Jacksonville University
-
-
0.606
-
0.701
Acosta Bridge
0.754
0.696
0.704
-
-
Buckman - Top
-
0.543
0.677
0.588
0.593
Buckman - Bot
-
0.565
0.744
0.105
0.667
Red Bay
-
-
0.241
-
0.368
Racy Point
-
-
0.350
-
0.369
Dancy Point - Top
-
0.375
-
-
Dancy Point - Bottom
-
0.317
-
-
Dancy Point
-
-
0.353
-
0.326
PRODUCTION RUN
The model results were compared to observed data to demonstrate model performance over the
entire Production Run Period (PRP) from January 2008 to June 2016. The observation dataset
from Mayport and Buckman Bridge provide the most complete record over the PRP. The
simulation under predicts WSE at Mayport (-0.028 m model error), but slightly overpredicts
WSE at Buckman Bridge (0.009 m model error). Among WSE stations, a very good level of
agreement was shown between modeled and observed values (RMSE < 0.1 m, IOA and R2 >
0.9).
For the salinity stations, simulation results are in good agreement with the observed data. Model
fit along the main stem of the Lower St. Johns River was variable, with model errors ranging
between -1.15 (Dames Point – Top) and 6.98 ppt (Mayport). Note that there is considerable
uncertainty regarding the quality of the data from the Mayport salinity monitoring station, as the
concentrations are lower than was observed upstream at Dames Point. Model IOA was
unsatisfactory upstream of Buckman due to the low salinities (less than 2 ppt) even though the
RMSE was below 0.2 ppt. This was considered to be an acceptable level of error since the
overall salinity concentration and relative magnitude of the errors were within the expected range
of instrument error and uncertainty.
The production run simulations form the basis for the long-term analysis of salinity intrusion and
its various impacts due to salinity stress on submerged aquatic vegetation, and wetland
communities.
Table 4. RMSE, IOA, and R2 values for the production run period for Water Surface
Elevation.
Station
RMSE (m)
IOA
R²
Mayport
0.048
0.959
0.995
Dames Point
0.055
0.936
0.988
Acosta Bridge
0.047
0.906
0.965
Main Street Bridge
0.045
0.913
0.972
Buckman Bridge
0.044
0.901
0.959
Red Bay Point
0.033
0.905
0.968
Racy Point
0.050
0.876
0.946
Table 5. RMSE, IOA, and R2 values for the production run period for Salinity.
Station
RMSE (ppt)
IOA
R²
Mayport
8.40
0.377
0.334
Dames Point - Top
3.10
0.735
0.767
Dames Point - Bot
3.98
0.637
0.769
Jacksonville University
3.42
0.701
0.743
Acosta Bridge
2.65
0.742
0.812
Buckman - Top
1.85
0.785
0.886
Buckman - Bot
1.61
0.841
0.870
Buckman Bridge
0.84
0.783
0.739
Red Bay Point
0.17
0.168
0.003
Racy Point
0.13
0.103
0.009
Dancy Point
0.14
0.376
0.002
ANALYSIS
The results of the 8.5-year PRP were used to analyze the extent of salinity intrusion upstream,
the impacts of salinity stress on submerged aquatic vegetation (SAV) beds, and various salinity-
tolerant wetlands.
Salinity Intrusion
The extent of salinity intrusion into the freshwater estuary was analyzed by determining the
distance upstream of the mouth of the river of the 2 ppt isohaline (a line of equal salinity
concentration) as a function of time. The drivers of salinity intrusion can be examined through
comparison with major drivers of variation throughout the estuary. Among the possible forcing
mechanisms, tidal forcing, atmospheric pressure variations (both regional forcing mechanisms),
tributary discharge, wind speed and direction (both local forcing mechanisms) could potentially
play a role in explaining the extent of the salinity intrusion (Figure 2). Among the potential
forcing mechanisms listed, tidal forcing, tributary discharge, and wind speed and direction were
examined in this study directly.
Figure 2. The upstream distance (in miles) from the mouth of the 2 ppt isohaline (Blue, top
panel), compared with the tidal forcing (Red, with diurnal trends filtered out). Middle
panel represents the combined flow from the two largest tributary inflows to the LSJR.
Bottom panel shows wind speed and direction plotted together for Mayport (Red) and
Jacksonville NAS (Blue). Tropical storms which tracked over or near the LSJR model
domain are shown with grey boxes along the time series, including the storms’ name.
During the PRP, the 2 ppt isohaline typically ranged between 20 and 80 miles upstream of the
LSJR mouth. Tide levels varied between ± 0.4 meters but was typically corresponded to the
mean tide level of -0.18 m NAVD88. Tributary inflow to the system is reflected by the time
series of discharge of the Upper St. Johns River and Ocklawaha Rivers which combined account
for over 70% of the total volumetric loading to the system. Wind speed and direction at Mayport
and Jacksonville NAS reflect winds at the coast and further inland.
A clear relationship between tributary inflow and the extent of salinity intrusion can be seen in
Figure 2 and was reflected in coefficient of determination (R2 = 0.5). High discharge events in
the upstream tributaries are frequency correlated with strong decreases in the extent of salinity
intrusion. This can be seen clearly with the example of Tropical Storm Fay in 2008, which
dropped heavy rain over the SJR basin for almost one week, and subsequently reduced the extent
of salinity intrusion by approximately 50 miles. By comparison, the other forcing mechanisms
examined in this study showed much weaker correlations with the extent of salinity intrusion.
Based on statistical measures, wind speed and regional tidal forcing are not highly correlated
with the extent of salinity intrusion (R2 < 0.1).
Submerged Aquatic Vegetation (SAV) Salinity Stress
Dobberfuhl et al. (2012) determined the two principal factors impacting SAV communities were
changes in water levels and salinity. Due to the long-time scales on which water surface
elevation is expected to change, the more immediate concern is changes in the salinity
concentration around submerged aquatic vegetation beds.
For this analysis, a set of littoral cells that represent existing and potential SAV beds in the
lacustrine segment of the LSJR was used to compute the acres of SAV per day experiencing
average salinity concentrations in excess of 3, 5, 10, 15, and 25 ppt. To provide a conservative
estimate of stress levels for SAV, the 90-day running average was utilized to compute the acres
of SAV per day experiencing average salinity concentrations of 3, 5, 10, 15, and 25 ppt (Table
6).
Table 6. Comparison of SAV stress levels (90-day running average) under existing
conditions.
SAV Salinity Effect
SAV Salinity Stress (ppt)
90-day running average (ac/day)
None
3
1683
Low
5
885
Low
10
140
Moderate
15
4.3
Extreme
25
0
Wetlands Analysis
To assess the impacts on wetlands habitat, a similar analysis to the SAV analysis was conducted.
For this analysis, the list of littoral cells was extended upstream to the Shands Bridge. Using
these cells, the depth averaged hourly salinities were extracted and saved to a separate file. Data
were then post processed to extract daily high tide salinities. These were then used to determine
percentages of time these cells experience salinity concentrations greater than 1 ppt. The percent
of time exceeding 1 ppt was classified into 3 categories: Tidal swamps are classified as cells with
less than 12% of daily values exceeding 1 ppt, Transitional zones are those having between 12
and 25% of daily values exceeding 1 ppt, and tidal marshes are cells where salinity
concentrations exceed 1 ppt for more than 25% of daily values (Figure 3).
Conditions were found to be ideal for tidal marshes in the lacustrine segment downstream of the
confluence with Julington and Durbin Creek. A short transition zone was found along Julington
and Durbin Creek and upstream to the confluence with Black Creek. Upstream of Black Creek,
conditions were found to be ideal for tidal swamps. The relatively short transitional zone
between tidal swamp and tidal marsh habitat illustrates the tenuous balance supported by
upstream discharge which, in turn, facilitates the dominance of specific vegetative communities.
Based on these results and those of the salinity intrusion analysis in the preceding section,
variations in discharge in the long term would likely impact the dominance of vegetative
communities within the lacustrine segment of the LSJR.
Figure 3. Percent of time exceeding 1 ppt salinity under existing conditions for wetland
shoreline cells.
CONCLUSIONS
The calibration and validation of the LSJR model showed that salinity intrusion can be
appropriately modeled using EFDC+. The generality of the calibration parameters used was
confirmed by the results of the PRP model, which generally showed better fit at observation
stations than in the calibration and validation runs. This can be attributed to the various boundary
conditions (particularly bathymetry) used in each calibration and validation period.
While model fit upstream of Buckman appears to be poor for salinity concentration in both the
calibration, validation, and PRP model, these statistics are somewhat deceiving. This is because
these stations typically have salinity concentrations below 1 ppt. This combined with the model-
data RMSE of less than 0.2 ppt, and the reporting standard of USGS instruments makes these
errors mostly negligible (USGS, 2006). The USGS (2006) standard is to report salinity values
lower than 1 ppt to the nearest 0.1 ppt. Additionally, instrument placement may play an
important role as sensors can often be affected by local conditions or organic fouling. Overall,
the calibration, validation, and PRP model results indicate that the model is well calibrated and
provides results which are generally reliable for reproducing salinity intrusion dynamics
throughout the main stem of the LSJR.
While variations in extent and duration of salinity intrusion are expected over natural cycles (i.e.
seasonal or tidal), the extent to which anthropogenic forces can have an impact are less clear.
Among the natural forcing mechanisms examined in this study, the magnitude of tributary inflow
appears to be the strongest controlling factor in determining the extent of salinity intrusion.
Tributary inflow appears to correlate most strongly with the extent of salinity intrusion (R2 =
0.5), while wind speed and regional tidal forcing are more likely to play a lesser role (R2 < 0.1).
Although not borne out by the statistics, it appears by inspection that both wind speed and tidal
forcing do play some role in determining the extent of salinity intrusion, but their effects may not
be as well captured by the model due to the limitations in both factors. The primary limitation
with the analysis of the role of wind forcing is that the observation data is derived entirely from
land-based stations. Furthermore, several of the stations are clustered around Jacksonville, with
only two stations being used south of Craig Airport. Using the inverse distance weighting
method, any locations which are far enough away from all the stations will weight every station
almost equally, which may not be appropriate.
For the regional tidal forcing, since there was a lack of appropriate high frequency water level
data offshore, data from Mayport was transformed using harmonic equations to approximate
water levels far from this station. However, the location of the Mayport gage may still contain
some portion of the overall discharge signal from the larger drainage basin which would not be
observable in a similar series of water level collected far from the rivers mouth. Without
removing this signal from the Mayport data, the regional tidal forcing may not accurately reflect
the offshore water levels. This could be remedied in the future by coupling the EFDC+ model
with a larger scale model like ADCIRC (Luettich et al., 1992), or by performing wavelet analysis
on the Mayport data to determine the appropriate frequency ranges to include farther offshore
(Jay and Flinchem, 1997).
While submerged aquatic vegetation beds (SAV) in the lacustrine segment of the LSJR
experiences generally low stress levels over the PRP, the relationship between the increasing
frequency of extreme weather events and sea level rise could cause stress levels to increase in the
coming decades (Field, 2014). Similarly, wetlands in the lacustrine segment are dominated by
tidal swamps, with only a short transition zone to buffer the freshwater wetlands from increase in
salinity over longer periods of time. As development continues to encroach on wetland habitat
around Jacksonville, changes in the distribution of runoff or the contents of that runoff could
have important effects on the long-term sustainability of these aquatic ecosystems. Further study
of the impact of extreme weather events or future sea level will be required to determine the
potential risk posed to aquatic habitat.
The extent to which various anthropogenic factors could impact the extent and duration of
salinity intrusion is an area of greater uncertainty, but such factors may constitute a more direct
potential impacts to SAV and wetland communities. To better account for these impacts, more
work will be required to better understand the role that regional and local forcing mechanisms
play in determining the extent and duration, and whether and to what extent human impacts can
enhance or curtail those impacts.
The current model forms a basis by which future conditions can be studied and their impacts
more precisely understood. Future work will be needed to address the various anthropogenic
factors which could drive change in the LSJR estuary into the future. Likewise, similar work will
be needed to address the potential impacts of sea level rise and an increasing frequency of
extreme weather events expected under climate change scenarios (Field, 2014).
FUTURE WORK
The development of the current modeling study identified several areas which can be improved
upon in future work. For example, many of the tributaries contain large swaths of vegetation
which was reflected using a higher z0 value. However, the use of z0 does not fully reflect the
hydrodynamic behavior of the vegetation itself in the process of mixing and stratification of the
water column. Furthermore, many of the upper tributaries are very small relative to the size of
the grid. Although it is unclear if a more refined grid would improve model performance.
The role of temperature was not considered in this study but is expected to be a minor factor in
determining the extent and duration of salinity intrusion. Inclusion of temperature would
facilitate the implementation of a water quality and eutrophication model which would permit
more detailed habitat analyses to be conducted.
While some efforts have already been undertaken to examine its impacts, the role of bathymetric
change as a result of channel deepening is another important factor which could influence the
extent of salinity intrusion in the estuary. Preliminary work in this area has shown bathymetric
change to play an insignificant role in determining the extent of salinity intrusion.
REFERENCES
Arakawa, A. and Lamb, V.R., (1977) “Computational design of the basic dynamical processes of
the UCLA general circulation model.” General circulation models of the atmosphere, 17,
pp.173-265.
Bicknell, B.R., Imhoff, J.C., Kittle, J.L., Jobes, T.H. and Donigian, A.S., (2001) “HSPF version
12 user’s manual.” National Exposure Research Laboratory. Office of Research and
Development US Environmental Protection Agency.
Craig, P., Chung, D., Lam, N., Son, P., and Tinh, N. (2014). “A Computationally Efficient
Approach to Reduce the Horizontal Gradient Error in the EFDC’s Vertical Sigma Grid.”
Proceedings of the 11th International Conference on Hydrodynamics (ICHD, Singapore,
October 2014)
Dobberfuhl, D., Chamberlain, B., Hall, S., Jacoby, C., Mattson, R., Morris, L., Slater, J., Moore,
K., Virnstein, B. (2012) “Chapter 9. Submerged Aquatic Vegetation. St. Johns River Water
Supply Impact Study.” Technical Publication No. SJ2012-1. SJRWMD, Palatka, FL.
DSI, LLC (2017) “CVLGrid User Guide” DSI, LLC, Edmonds, WA, 2017
DSI, LLC (2018) “EFDC_Explorer Knowledge Base”, <
https://eemodelingsystem.atlassian.net/wiki/spaces/EEREF/overview> (Dec. 28, 2018)
Field, C. B. (Ed.). (2014). Climate change 2014–Impacts, adaptation and vulnerability: Regional
aspects. Cambridge University Press.
Galperin, B., Kantha, L.H., Hassid, S. and Rosati, A., (1988) “A quasi-equilibrium turbulent
energy model for geophysical flows.” Journal of the Atmospheric Sciences, 45(1), pp.55-62.
Hamrick, J.M. (1996) “User’s Manual for the Environmental Fluid Dynamics Computer Code.”
Special Report No. 331 in Applied Marine Science and Ocean Engineering, Virginia Institute
of Marine Science, Gloucester Point, VA.
Jay, D. A., & Flinchem, E. P. (1997). “Interaction of fluctuating river flow with a barotropic tide:
A demonstration of wavelet tidal analysis methods.” Journal of Geophysical Research:
Oceans, 102(C3), 5705-5720.
Luettich Jr, R. A., Westerink, J. J., & Scheffner, N. W. (1992). ADCIRC: An Advanced Three-
Dimensional Circulation Model for Shelves, Coasts, and Estuaries. Report 1. Theory and
Methodology of ADCIRC-2DDI and ADCIRC-3DL (No. CERC-TR-DRP-92-6). COASTAL
ENGINEERING RESEARCH CENTER VICKSBURG MS.
Smagorinsky, J. (1963) “General circulation experiments with the primitive equations: I. The
basic experiment.” Monthly weather review, 91(3), pp.99-164.
St. Johns River Water Management District (2012) “St. Johns River Water Supply Impact Study
Final Report – Chapter 10: Wetland Vegetation.” Retrieved from:
https://www.sjrwmd.com/documents/water-supply/#wsis-final-report
Taylor Engineering (2017) “HSPF Modeling for Jacksonville Harbor Corrective Action Plan –
Navigation Project.”
United States Army Corps of Engineers (2014) “Final Integrated General Reevaluation Report II
and Supplemental Environmental Impact Statement” Duval County, Florida, Jacksonville
Harbor Navigation Study.
United State Geological Survey (2006) “Guidelines and Standard Procedures for Continuous
Water-Quality Monitors: Station Operation, Record Computation, and Data Reporting”
Techniques and Methods 1-D3.
