You are on page 1of 17

Regional Studies in Marine Science 43 (2021) 101674

Contents lists available at ScienceDirect

Regional Studies in Marine Science


journal homepage: www.elsevier.com/locate/rsma

Hydrodynamic modeling of salinity habitat changes with reduced


submarine groundwater discharges in a spring-fed estuary
XinJian Chen 1
Natural Systems and Restoration Bureau, Southwest Florida Water Management District, 7601 Highway 301 North Tampa, FL 33637, USA

article info a b s t r a c t

Article history: Effects of reduced submarine groundwater discharge (SGD) on salinity habitats were studied for a
Received 15 December 2020 narrow spring-fed estuary. A laterally averaged model was used to simulate circulations, salinity
Received in revised form 10 February 2021 transport processes, and thermal dynamics in the estuary. It was calibrated and verified against
Accepted 11 February 2021
measured water elevations, salinities, and temperatures at five stations during November 2014–August
Available online 16 February 2021
2017, before it was used to evaluate effects of SGDs on salinity habitats during a 125-month period.
Keywords: Salinity habitats considered and analyzed here included water volumes, bottom areas, and shoreline
SGD lengths for salinity ≤ 1, 2, 3, 5, 10, and 15 psu.
Spring-fed estuary It is found that the long-term (hourly) and short-term (monthly or longer) responses of salinity
Coastal springs habitats to the SGD reduction are different in the estuary. Long-term average salinity habitats have a
Hydrodynamic simulations linear relationship with the SGD reduction. Except those for salinity ≤ 1 psu, long-term oligohaline
Laterally averaged hydrodynamic model salinity habitats in the estuary are more sensitive to the SGD than salinity habitats with mesohaline
Salinity habitats
salinity zones being included. Long-term average water volumes and bottom areas of ≤2 psu are most
sensitive to the SGD and they can be reduced by 1.38% for every 1% of SGD reduction. The short-term
response of salinity habitats is greatly affected by special physical features in the estuary, and the
sensitivity of a salinity habitat to the SGD varies with the location of the isohaline in relationship with
these physical features. For example, when the 2 and 3 psu isohalines pass a narrow transect in the
upstream area of the estuary, ≤2 and 3 psu salinity habitats become much more sensitive to the SGD
than other locations of the isohalines. On the other hand, when the 5 and 10 psu isohalines enter the
downstream distributaries, ≤5 and 10 psu salinity habitats are most sensitive to the SGD.
Published by Elsevier B.V.

1. Introduction wetland is self-percolating and barely contributes any quantifi-


able surface water runoff, the estuary receives an insignificant
The Homosassa River is a small spring-fed estuary located on amount of freshwater runoff from its 145 km2 Watershed. Most
the Gulf coast of central Florida (Fig. 1). From its headspring to of the hydrologic loading to the Homosassa River comes from
the Gulf of Mexico (GOM), the estuary is about 13 km long and the UFA in the form of submarine groundwater discharges (SGDs)
has a width varying from about 60 m in the upstream reach to out of numerous spring vents located in the headspring area and
about 305 m near the mouth (Yobbi and Knochenmus, 1989). several tributaries, including the SE Fork and the Halls River. The
Except for the navigation channel and some deep holes, most Homosassa River has a springshed of about 700 km2 , with a slight
of the riverbed is less than 1 meter deep under the mean tide year-to-year variation as the contributing groundwater field for
level. The estuarine system includes several tributaries, among the spring discharge is subject to the shape of potentiometric
which the SE Folk and the Hall River enter to the upstream area contours in the region. Locations of the springs that discharge
of the Homosassa River, while the Salt River and Mason Creek are groundwater flows to the Homosassa River are shown as solid
connected to the downstream part of the estuary. The river also dots in Fig. 1. The headwaters of the river are the Homosassa
has some other smaller tributaries, including a channel system in Springs, and several other springs. Because nearly all flows en-
the town of Homosassa and the Hidden river. tering the Homosassa River estuary are SGDs, SGD and flow are
The Homosassa River flows through a coastal swamp region, exchangeable in this article.
where poorly drained and saturated organic soils overlie lime- On each side of the Homosassa River, there is an extensive
stone rocks of the Upper Florida Aquifer (UFA). As most of the wetland system. As plant species are sensitive to changes in salin-
ity, different plants are distributed in the wetland and along the
E-mail address: xinjian.chen@swfwmd.state.fl.us. shoreline of the estuary. Salinity in the Homosassa River estuarine
1 Chief Professional Engineer, Southwest Florida Water Management District. system exhibits a natural gradient from the headsprings to the

https://doi.org/10.1016/j.rsma.2021.101674
2352-4855/Published by Elsevier B.V.
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 1. Aerial view of the Homosassa River on the Gulf coast of Florida. Triangles mark locations of USGS real-time data stations along the river.

GOM. As a result, hydric hammocks can be found near the freshest Knochenmus (1989) tried to find relationships among salinities,
waters in the upstream areas, especially along the southeast fork tides, and SGDs in the Homosassa River by examining measured
of the Homosassa River, while salt marshes and coastal hydric data at fixed gage stations during 1984 and 1985 with some
hammocks are found in the downstream areas where salinity is sporadic data collected at several additional sites. They found that
high (Leeper et al., 2012). the water column was typically well-mixed, with the top-layer
The mouth of the Homosassa River is located near the center of salinity being generally less than 15% lower than the bottom-layer
the southern half of the Florida Big Bend Coast, which is roughly salinity. During the two-year period, salinity at the downstream
from the mouth of the Suwannee River, located about 21 km side of the confluence of the Homosassa and Halls Rivers typically
northwest of Cedar Key, Florida, to the mouth of the Anclote River fluctuated between one to two psu when the river mouth had a
near Tarpon Springs, Florida. Macroscopically, the coastline of this salinity ranging approximately 13–26 psu. Yobbi and Knochen-
part of the Florida Big Bend Coast looks like an arc of a circle. mus (1989) noticed that the 2 psu isohaline varied over a 2.8-km
The circular segment has a sagitta of about 32 km and a chord stretch between river km 7.2 to river km 10.0. On the other hand,
length of about 130 km. The central angle and the area of the the 25 psu isohaline varied over a stretch of 8.6 km, between
circular segment can be estimated to be about 105 degrees and river km -8 to river km 0.6. Although these observations represent
2,900 km2 , respectively. Several rivers drain a large amount of actual salinity variations in the estuary, Yobbi and Knochenmus
freshwater flows to this relatively shallow coastal water, where
(1989) did not give a physically reasonable explanation for what
most areas are less than 2 m deep. These rivers include the
they observed. HSW Engineering (2011) applied the Environmen-
Suwannee River, Withlacoochee River, and four major spring-
tal Fluid Dynamics Code (EFDC), originally developed by Hamrick
fed estuaries (Crystal, Homosassa, Chassahowitzka, and Weeki
(1992), to simulate circulations and salinity transport processes
Watchee Rivers). The Suwannee River alone drains a watershed of
in the Homosassa River. They had to expand and deepen the
approximately 25,770 square kilometers stretching from northern
downstream area of the river to form a large funnel to allow
Florida to southern Georgia, and the Withlacoochee River has a
salt water to be transported upstream in their EFDC application
watershed of about 5,440 square kilometers. Each of the four
spring-fed estuaries receives SGDs from first magnitude springs (HSW Engineering, 2011). Clearly, this artificial alteration of the
or spring groups. topography and bathymetry dramatically changed the shape of
Because of the relatively large amount of hydrologic loading the estuary that was simulated and should not be made in any of
received, salinity along the Florida Big Bend Coast is generally hydrodynamic modeling studies. Model results from an estuary
low. At the mouth of the Homosassa River, recorded salinity that has a dramatically different shape from the real estuary are
generally varies between < 15 psu to < 25 psu most of the time of meaningless.
the year. Only during the end of the dry season, which generally Because the Homosassa River is a narrow and meandering es-
starts in late April and ends in late June for the region, will salinity tuary, its circulation pattern and salinity and temperature distri-
at the river mouth reach 30 psu or more at high tides. Like some butions vary primarily in the vertical and longitudinal directions.
other estuaries along the Gulf coast of Florida, the Homosassa Vertically two-dimensional variations are typical for narrow es-
River estuary is under the action of microtidal forces, which tuaries (Prandle, 1985; Jay and Smith, 1990; Chen, 2003), for
are primary diurnal and semidiurnal. As the estuary is relatively which the most suitable simulation tool is a laterally averaged
shallow, the Homosassa River is generally well or partially mixed. model. This study used the Laterally Averaged Model for Estu-
There are only limited previous studies on effect of SGDs aries (LAMFE), developed by the author, for the simulation of
on salinity habitats in the Homosassa River estuary. Yobbi and circulations, salinity transport processes, and thermal dynamics
2
X. Chen Regional Studies in Marine Science 43 (2021) 101674

in the Homosassa River. The LAMFE model has been applied to 2.2. Water level, salinity, and temperature data
several estuaries over the last two decades, including the lower
Hillsborough River (Chen and Flannery, 1997; Chen et al., 2000; There were several fixed stations in the Homosassa River
Chen, 2004a), the lower Alafia River (Chen, 2004b; Flannery et al., system, at which real-time data were collected for water eleva-
2007), the lower Peace and Myakka Rivers (Chen, 2007, 2008, tion, salinity, and temperature with a time interval of 15 min.
2010), and the lower Manatee/Braden Rivers (Chen, 2012a). Most of these stations were maintained by the United States
Geological Survey (USGS). Yellow triangles in Fig. 1 show loca-
As mentioned in Chen (2014), most previous coastal and es-
tions of the USGS real-time stations, including (1) Homosassa
tuarine hydrodynamic modeling studies did not consider effects
River at Shell Island near Homosassa (#02310712), (2) Homosassa
of SGDs (Johnson et al., 1991; Blumberg and Kim, 2000), partly
River at Homosassa (#02310700), (3) Halls River near Homosassa
because SGDs are much lower than river flows and precipitations
(#02310690), (4) Halls River at Homosassa Springs (#02310689),
for these estuaries and partly because SGDs are very difficult to (5) Homosassa Springs at Homosassa Springs (#02310678), (6)
quantify in many cases. There are only very limited estuarine SE Fork Homosassa Spring at Homosassa Springs (#02310688),
and coastal hydrodynamic modeling studies which considered and (7) Hidden River near Homosassa (#02310675). More de-
SGDs. Ganju et al. (2012) applied the 3D model ROMS (Warner tails about these USGS real-time stations can be found in Chen
et al., 2008) to West Falmouth Harbor, Massachusetts to verify (2019a). In addition to the USGS stations, two additional fixed
their tidal and groundwater flux estimates to the estuary based stations by the Southwest Florida Water Management District
on velocity and salinity measurements. Chen (2014) estimated (SWFWMD) were available for the Salt River and Mason Creek
and considered SGDs in the hydrodynamic simulations for Crystal (see red triangles marks in Fig. 1 for their locations.)
River/Kings Bay in using the 3D model UnLESS3D to analyze Measured real-time data of water level, salinity, and tempera-
thermal habitats for manatees in the spring-fed estuary. It was ture indicate that the Homosassa River estuary is a low energy
found that SGDs in Crystal River/Kings is not only negatively system. Although tidal signals are recorded everywhere in the
river, they generally have a tidal range of less than 30 cm. Even
proportional to the tides but also positively proportional to the
at the station near the mouth of the estuary, tides generally
first derivative of tides with respect to time.
vary within a range of 50 cm or less. Tides in the estuary are
In the following sections, available data in the spring-fed es-
primarily diurnal and semidiurnal; however, non-tidal and low
tuary are described, with major physical characteristics being frequency signals also exist in measured real-time water levels at
discussed. The LAMFE model application to the Homosassa River all stations. These low frequency variations are mainly influenced
is then presented with a brief description of model setup and by meteorological and hydrological characteristics of the region.
model calibration and verification. Effects of SGD on salinity habi- Because of the physical setting of the estuarine system, tides
tats in the Homosassa River estuary are analyzed after a series of are attenuated quickly as they propagate from the mouth to the
model runs were conducted. Conclusions of the modeling study upstream segment of the Homosassa River. From the Homosassa
are provided at the end of the paper. River at Shell Island station near the mouth to the Homosassa
River at Homosassa station, about two thirds into the estuary,
tides lose about one half of their variabilities. In other word,
2. Measured and estimated data about 75% of tidal energy is lost along two thirds of the river. In
the tidal reach upstream of the Homosassa River at Homosassa
Different types of reliable data are needed for a successful station, up to the headwaters of the river, the tidal energy atten-
simulation of estuarine and coastal hydrodynamics. These data uation becomes small, as tidal waves attenuate only slightly while
propagating further upstream.
include topographic and bathymetric data, flow data, meteoro-
One of the reasons for this phenomenon of different tidal
logical data, and real-time data of water elevation, salinity, and
energy losses is that the downstream reach is almost two times
temperature. While topographic and bathymetric data are to
of the length of the upstream reach, causing the tidal energy
define the shape of the estuary, meteorological data and some to lose more in the downstream reach than in the upstream
other data are to drive the hydrodynamic model. For model reach. Another reason is that the tidal currents in the downstream
calibration and verification, real-time water levels, salinities, and part are stronger than these in the upstream part of the estuary.
temperatures recorded at stations inside the simulation domain Nevertheless, a major reason for the large tidal energy loss in the
will be needed to ensure that the numerical model performs well downstream reach may be caused by a few narrow segments in
for the waterbody. the downstream portion of the river, especially the one at around
the half way between the Homosassa River at Shell Island and
Homosassa River at Homosassa stations, which is about 2 km long
2.1. Topographic and bathymetric data and greatly restricts the passage of tidal energy passing through
it.
For the Homosassa River estuary, bathymetric data were col- Salinity and temperature variabilities in the downstream area
lected by surveying cross sections along the main stem of the of the Homosassa River are dominated by processes in the GOM,
river and its tributaries. Topographic data collected using the while in the upstream area, they are dominated by SGDs. Al-
though measured salinities and temperature at upstream stations
Light Detection and Ranging (LiDAR) technique are available for
contain tidal signals, their variabilities are weak. For example, at
the region. These data show that the Homosassa River estuary is
the Homosassa Springs at Homosassa Springs station, measured
generally shallow, with the water depth being less than one me-
temperature was almost constant at about 23 ◦ C, with a very
ter most of the area. The upstream segment of the main stem and insignificant daily and seasonal variations. During winter, water
the tributaries are shallower than the middle and downstream temperature at the mouth can drop to 10 ◦ C or lower but the
segment of the Homosassa River. The elevation of the thalweg of upstream area near Homosassa Springs and the SE Fork were
the Homosassa River main stem varies from about −1.0 m NAVD warm with temperature higher than 20 ◦ C during these cold days.
88 in the upstream area to about −6.6 m, NAVD88 in the middle Further downstream in the Homosassa River and in Halls River,
segment of the river. the low water temperature was about 15 ◦ C, which was still
3
X. Chen Regional Studies in Marine Science 43 (2021) 101674

roughly 5 ◦ C higher that at the river mouth, because of the warm comprises most of the hydrologic loading to the estuary. The only
SGDs flowing to the estuarine system. SGD missing from this station is that entering the Hidden River
At the cross section of the SE Folk station, flow is mostly (Fig. 1), which is gaged at the USGS Hidden River near Homosassa
fresh (< 0.5 psu) and only occasionally a salinity peak of only station.
1–1.5 psu would occur. The SE Fork Homosassa Spring station is Although Homosassa Springs are the headwater for the estu-
about 130 meters to the confluence with the Homosassa River, ary, the USGS does not have an ADCP deployed downstream of
which is roughly the same distance to the Homosassa Springs these spring vents to record SGDs due to the lack of a proper cross
at Homosassa Springs. While salinity at the SE Fork station is section downstream of the vents. Based on limited discharge data,
barely higher than 1 psu, salinity measured at the Homosassa a regression equation was obtained by the USGS that relates
Spring at Homosassa Springs is barely less 1 psu. This suggests the discharge of Homosassa Springs with the water level at the
that SGDs from SE Fork springs, including Belcher Spring, Trotter site and the groundwater level in a well in Weeki Wachee. This
Springs, McClain Spring, and Pumphouse Spring are fresh and regression equation was updated several times over the years.
the occasional salinity peaks are caused by high tides. Under The most recent version, valid since October 1, 2013, takes the
the normal condition, the freshwater–salt water interface moves following form
back and forth around the SE Fork station, with its location being
Qhs = 76.863 − 20.7636h + 5.3139W (1)
downstream of the station most of the time. As salinity at the SE
Fork station comes from the estuarine part of the river, not from where Qhs (in cfs) is the flow rate for Homosassa Springs at
the SGD, it can be concluded that SGD gaged at the SE Fork is Homosassa Springs, h is water level (feet, NGVD29) and W is
fresh. It is not expected that a reduction of SGD would cause any Weeki Wachee well level (feet, NGVD29).
salinity increases in SGD (Chen, 2019b). For the SE Fork discharge before October 1, 2011, the USGS
Although the Halls River at Homosassa Springs station is lo- also has a regression equation as follows
cated upstream of the Halls River near Homosassa station, salinity
measured at the upstream station is about 2 psu higher than that Qse = 18.6293 − 10.3114h − 418.139∆h + 3.31029W (2)
measured at the downstream station. Obviously, SGD entering where Qse (in cfs) is the flow rate through the SE Fork Homosassa
the upstream of the Halls River is brackish. The relatively flat station and ∆h is the water level change over the data collection
salinity peaks suggest that the upstream SGD is a salinity source interval (15 min).
for the Halls River and salinity in the SGD is quite stable under Both Equations (1) and (2) show that the SGD is positively
various discharge conditions. The relatively lower salinity at the proportional to the groundwater level but negatively proportional
downstream station is caused by the fresher SGDs from SE Fork to the water elevation in the Homosassa River. However, Eq. (2)
and Homosassa headwaters, because the Halls River near Ho- has an extra term suggesting that SGDs from spring vents up-
mosassa station is close to the confluence of the Halls River with stream of the SE Folk station is also linearly related to the water
the Homosassa River. The tidal currents transport fresher water level change over the 15-minute interval of data collection. This
upstream to dilute the brackish SGD in the upstream portion of is consistent with a previous study on effects of tides on SGD in
the Halls River. The moderately flat salinity peaks generally varied Crystal River/Kings Bay (Chen, 2012b, 2014). Chen (2014) found
at around 5 psu and it is possible that salinity in the upstream that the time derivative term is necessary not only to properly
SGD of the Halls River is the same or slightly higher than that match the phase in the tidal signal of the measured SGD but also
measured at the Halls River at Homosassa Springs station. to model the high frequency variabilities observed in the SGD
One thing that can be noticed from the measured salinity data.
data is that salinity at the Homosassa River at Homosassa station Fig. 2 shows discharge data at all the stations in the Homosassa
does not correlate well with salinity at the river mouth. Several River system. For clarity, only 2016 was shown in the figure
salinity signals that are strong at the mouth are nonexistent or and discussed here. Discharge data collected/estimated for other
become weak at the Homosassa River at Homosassa station, and years can be found in Chen (2019a). As can be seen from Fig. 2,
vice versa. A regression between salinities at the two stations has most discharge data contain significant tidal signals, mainly due
a R2 of less than 0.5, even when various time leads/lags were to the existence of tidal prisms upstream of these stations. The
considered, suggesting that salinity in the upstream half of the Homosassa River at Homosassa has the biggest upstream tidal
estuaries are more dominated by SGDs and/or factors other than prism, resulting in the biggest tidal flux varying between about
those occurring in the GOM. -1,000 cfs to about 1,000 cfs. The Hidden River discharge has only
insignificant tidal signals but has a clear seasonal variability.
2.3. Flow data The big positive/negative peaks shown in Fig. 2 were during
days of Hurricane Hermine in 2016. Hermine developed in the
As mentioned above, most of the hydrologic loading to the Florida Straits on August 28, 2016 and made landfall in the Florida
Homosassa River estuary comes from the various spring vents and Panhandle on September 2, 2016. As presented in Chen (2019a),
the quantification of the SGD is very difficult because of the com- during these hurricane days, the SE Fork Homosassa Spring sta-
plexity of spring vents and the impact of tides on the discharge tion had a negative discharge peak of more than 400 cfs, while the
rate. Some of the limestone vents are clustered together and Homosassa River at Homosassa station had a negative discharge
connected. There could be interactions among these spring vents. peak of almost 10, 000 cfs.
It is generally infeasible to measure the discharge from each An analysis of measured SGDs was conducted for all the sta-
individual spring vent. As a result, the combined SGD from several tions and it was found that the long-term average of SGD gaged
spring vents is generally measured at a proper downstream cross at the Homosassa River at Homosassa station is roughly equal to
section of these vents. The USGS has been deploying acoustic the sum of those gaged at the SE Fork Homosassa Spring station,
Doppler current Profilers (ADCPs) to the estuary and using the the Homosassa Springs at Homosassa Springs station, and the
index velocity method to calculate SGDs entering the SE Folk Halls River at Homosassa Springs station. In other words, there
and the Halls River since October 1, 2011 and March 10, 2012, is no noticeable ungaged flow upstream of the Homosassa River
respectively. The USGS also has an ADCP at the Homosassa River at Homosassa station. For example, during September 30, 2016–
at Homosassa station to record the combined SGD upstream of March 13, 2018, average SGDs were 177.09 cfs, 56.77 cfs, 90.90
the cross section since October 1, 2006. SGD gaged at this station cfs, and 30.58 cfs at the Homosassa River at Homosassa, SE Fork
4
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 2. Measured and estimated discharge data in the Homosassa River system during 2016.

Homosassa Spring, Homosassa Springs at Homosassa Springs, and the grids varies between 29 to 277 m. Fig. 3 shows cross sections
Halls River at Homosassa Springs stations, respectively. The three that form the 406 LAMFE grids along the estuarine system. In the
upstream stations had a total of 178.25 cfs, which was very close vertical direction, 15 layers, varying from 0.3 to 1.76 m, were used
to the average discharge of 177.09 cfs at the Homosassa River at to discretize the water depth ranging between −6.56 m, NAVD88
Homosassa station. to 3.0 m, NAVD88.
Based on the availability of measured data which were used
2.4. Meteorological data to drive the model, a 34-month period between 11/4/2014 and
8/31/2017 was chosen for model calibration and verification.
Meteorological data that are available for study included wind During this 34-month period, SGDs and meteorological data were
speed, wind direction, air temperature, air humidity, and solar available. Water elevation, salinity, and temperature data at the
radiation at a SWFWMD weather station near Inglis, Florida and mouth of the Homosassa River were also available. However,
at a Florida Automated Weather Network (FAWN) station by the there were no water elevation, salinity, and temperature data
University of Florida’s Institute of Food and Agricultural Science for the open boundaries in the Salt River and the Mason Creek
(UF/IFAS). The former is about 16 miles north of the Homosassa for the entire 34 months, except for a 4-month period in 2017.
River, while the latter is about 6 miles northeast of the river. The To obtain boundary conditions of water elevation, salinity, and
region has distinct winter and summer patterns. In the winter, temperature in the Salt River and Mason Creek for the entire
frequent frontal incursions and extratropical cyclones can pro-
34 months, an effort was made to correlate water elevations,
duce large shifts in wind speed and wind direction in response to
salinities, and temperatures measured at the Salt River and Mason
rapidly changing atmospheric pressure and thermal gradients. In
Creek stations with those at other data collection stations with
the summer, the weather is generally characterized by light and
longer period of record. It was found that water level, salinity,
variable winds originating from the northeast trade wind circu-
and temperature data in the Salt River and Mason Creek are best
lation. Sea/land breezes are typical due to the strong differential
correlated with those measured at the mouth of the Homosassa
heating of the land and adjacent waters along the coast during
River. A trial and error approach with various time lags or leads
the summer months. Occasional tropical storms can move to the
area during summer, causing a temporal but sometimes intense was used to obtain the best correlations. Details about estimating
modification to the meteorological conditions of the region. boundary conditions of water level, salinity, and temperature in
the Salt River and Mason Creek are documented in Chen (2019a).
3. Model calibration and verification As discharges from spring vents are measured at proper tran-
sects downstream of these groundwater sources, measured flow
3.1. Model setup data contain tidal fluxes through these transects. To obtain net
SGDs entering the upstream reaches of the cross sections, tidal
For the LAMFE model application, the Homosassa River estu- fluxes were estimated through the following formula and taken
ary, was discretized with 406 grids along the river main stem and away from the reported discharge data.
its 21 branches to ensure that the main physical of the system ∂η
can be properly resolved by the model. The horizontal spacing of qt = −A (3)
∂t
5
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 3. Cross sections (yellow segments) that form the LAMFE grids for the Homosassa River and its branches. Numbers in green are grid numbers in the longitudinal
direction. Orange arrows are locations where SGDs enter the model domain. (For interpretation of the references to colour in this figure legend, the reader is referred
to the web version of this article.)

where qt is the tidal flow, with the positive flow pointing toward model parameters within certain allowable ranges to obtain the
downstream, η is measured water level, t is time, and A is the best match of simulated water levels, salinities, and temperatures
surface water area upstream of the cross section. with measured data collected in the field. The LAMFE model for
Except for the Hidden River, net SGDs were added to the the Homosassa River was mainly calibrated against measured
model domain at the most upstream grids of the main stem and real-time data at the five USGS station: SE Fork Homosassa
branches of the estuarine system, instead of at the stations where Spring, Homosassa Springs at Homosassa Springs, Halls River at
discharges were measured or estimated. Orange arrows in Fig. 3 Homosassa Spring, Halls River near Homosassa, and Homosassa
indicate locations where SGDs enter the model domain. River at Homosassa stations. Only limited LAMFE model param-
Because no direct measurements of salinity and temperature eters had to be tuned in the calibration process, including the
in the spring vents were available for this modeling study, salinity bottom roughness, ambient vertical eddy viscosity and diffusivity,
and temperature in SGD was an unknown and needed to be and light attenuation. LAMFE allows different river segments to
reasonably estimated. This study used a trial and error approach be assigned with different values for these model parameters,
to estimate salinities and temperatures in all the SGDs. Based on among which, ambient vertical eddy viscosity is the most sen-
measured salinities and temperatures at the SE Fork Homosassa sitive parameter for salinity simulation and light attenuation is
Spring station, the Homosassa Springs at Homosassa Springs sta- the most sensitive parameter for temperature simulation. The
tion, and the Halls River at Homosassa Springs station, many model was best calibrated with an ambient vertical eddy viscosity
salinity and temperature estimates were tested in model runs. of 0.10 cm2 /s, a light attenuation coefficient of 0.004 1/cm, and
After a careful analysis of simulated salinity and temperature a bottom roughness height (z0 ) and an ambient vertical eddy
results, it was found that it is suitable to use measured salinity at diffusivity in the ranges of 0.10–0.12 cm and 0.05–0.10 cm2 /s,
the Homosassa Springs at Homosassa Spring for the Homosassa respectively.
Main SGD and measured salinity at the Halls River at Homosassa Comparisons of model results with measured field data at
Springs station for the Halls River SGD. For the SE Fork, the best the five measurement stations in the Homosassa River estuarine
SGD salinity estimate takes the following form system are presented in Figs. 4–6. In these time series plots, red
lines are model results, while dashed green lines are measured
Sse = max(0.952Ssef , 0.3) (4)
data. For simplicity and clarity, only 60 days of model results
where ssef represents measured salinity at the SE Fork Homosassa and field data of water level, salinity, and temperature, between
Spring station and sse is the estimated salinity in SGD entering the 1/5/2015 and 3/5/2015, are shown in Figs. 4–6. The choice of
SE Fork. The best estimate for SGD temperature is a combination these 60-day period is arbitrary. Comparisons of model results
of measured temperature and a constant value of 23.5 ◦ C. More with real-time field data during other time periods are similar,
details about estimating salinity and temperature in SGDs are with some having a slightly better match and some a slightly
reported in Chen (2019a). worse match. Time series plots showing comparisons of model
results with field data of water level, salinity, and temperature
3.2. Comparisons of model results with field data during other time periods are reported in Chen (2019a).
As can be seen from Fig. 4, simulated water elevations at all
The 34-month simulation period was divided into a model five USGS stations agree very well with measured data. Simulated
calibration period and a model verification period. The former water levels have the same long-term and short-term variations
was 11/4/2014–5/31/2016, while the latter was 6/1/2016– as measured data. Modeled salinity and temperature results also
8/31/2017. Model calibration involves a series of adjustment of have good agreement with field data (Figs. 5 and 6), both in
6
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 4. Comparison of measured and simulated water levels at the SE Fork Homosassa Spring, Homosassa Springs at Homosassa Spring, Halls River at Homosassa
Spring, Halls River near Homosassa, and Homosassa River at Homosassa stations during 1/5/2015 through 3/5/2015.

terms of long-term and short-term variations. Because of the Table 1 summarizes MEs, MAEs, RMSEs, NRMSEs, R2 , and
uncertainties included in the input data that drive salinity and Skills for simulated water elevations, salinities, and tempera-
temperature simulations, it is expected that the match between tures. These statistics were calculated for the calibration period
model results and data for salinity or temperature is not as good (11/4/2014–5/31/2016) and the verification period (6/1/2016–
as that for water level simulation. Nevertheless, simulated salin- 8/31/2017) separately. They were also calculated for the en-
ity and temperature results are satisfactorily well matched with tire simulation (calibration & verification) period (11/4/2014–
measured data. The reversed salinity distribution in Halls River, 8/31/2017).
where upstream salinity is higher than downstream salinity, was As can be seen from Table 1, the ME between simulated and
correctly simulated. measured water elevations is about -1.53 cm during the calibra-
tion period, 0.79 cm during the verification period, and -0.64 cm
for the entire period. MAEs for the water elevation simulation
3.3. Model performance metrics
are 5.44, 5.68, and 5.54 cm for the calibration, verification, and
entire period, respectively. RMSEs are 7.17 cm, 7.29, and 7.21 cm,
The performance of the LAMFE model for the Homosassa River
for the calibration, verification, and the entire simulation periods,
is accessed quantitatively with several statistics, including the respectively, with the corresponding NRMSEs being 0.04, 0.04,
mean error (ME), mean absolute error (MAE), root-mean-square and 0.03. The R2 values between simulated and measured water
error (RMSE), normalized root-mean-square error (NRMSE), co- levels are all 0.86 for the three periods, during which the Willmott
efficient of determination (R2 ), and a skill assessment param- skills are all 0.96.
eter (Skill) introduced by Willmott (1981). The Willmott skill For salinity, MEs between model results and field real-time
assessment parameter takes the following form data are −0.07, 0.14, and 0.0 psu, respectively during the calibra-
tion, verification, and entire periods, respectively. The MAEs are
∑ M
(y − yD )2
Sk = 1 − ∑ ⏐ ⏐ (5) about 0.39, 0.61, and 0.46 psu, respectively during the three peri-
(⏐yM − yD ⏐ + |yD − yD |)2
⏐ ⏐
ods, while RMSEs between simulated and measured salinities are
0.69, 1.01, and 0.81 psu, respectively. The NRMSEs for the salinity
where Sk is the skill assessment parameter; yM and yD represent prediction is 0.03, 0.04, and 0.03, respectively for the calibration,
simulated and measured variables; and yD is the expectation of verification, and entire periods. The R2 values between simulated
yD . Sk in the above equation varies between 0 and 1, with one be- and measured salinity are 0.85, 0.81, 0.83, respectively for the
ing a perfect agreement and zero being a complete disagreement calibration, verification, and entire simulation periods, while the
between simulated results and measured data. Skills are 0.96, 0.95 and 0.95, respectively for the three periods.
7
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 5. Comparison of measured and simulated salinities at the SE Fork Homosassa Spring, Homosassa Springs at Homosassa Spring, Halls River at Homosassa Spring,
Halls River near Homosassa, and Homosassa River at Homosassa stations during 1/5/2015 through 3/5/2015.

Table 1
MEs, MAEs, RMSEs, NRMSEs, R2 values, and Willmott skill assessment parameters for simulated water levels, salinities, and
temperatures, in comparison with real-time field data measured at the five stations within the simulation domain during the
calibration period, the verification, and the entire simulation period (11/4/2014–8/31/2017.)
Simulation Period ME MAE RMSE NRMSE R2 Skill
Water Level (cm)
Calibration −1.53 5.44 7.17 0.04 0.86 0.96
Verification 0.79 5.68 7.29 0.04 0.86 0.96
Calib. + Verif. −0.64 5.54 7.21 0.03 0.86 0.96
Salinity (psu)
Calibration −0.07 0.39 0.69 0.03 0.85 0.96
Verification 0.14 0.61 1.01 0.04 0.81 0.95
Calib. + Verif. 0.00 0.46 0.81 0.03 0.83 0.95
Temperature ()
Calibration −0.18 0.70 1.02 0.04 0.93 0.98
Verification −0.07 0.59 0.89 0.03 0.93 0.98
Calib. + Verif. −0.14 0.66 0.98 0.03 0.93 0.98

MEs for the temperature prediction by the LAMFE model are In summary, different assessments of the model skill show
−0.18, −0.07, and -0.14 ◦ C, respectively during the calibration, that the LAMFE model performed very well for the simulation
verification, and entire periods. The corresponding MAEs during of hydrodynamics, salinity transport processes, and thermal dy-
the three periods are 0.70, 0.59, and 0.66 ◦ C, respectively. RMSEs namics in the Homosassa system. If measured data at the USGS
between simulated and measured temperatures are 1.02, 0.89,
stations at all real-time stations within the simulation domain of
and 0.98 ◦ C during the calibration, verification, and entire periods,
the Homosassa River estuary were all perfectly accurate, statis-
respectively, with their NRMSEs being consistent with those for
predicted water levels and salinities; namely about 0.03–0.04 tics listed in Table 1 would represent errors or uncertainties of
for the three simulation periods. The R2 values between simu- simulated water levels, salinities, and temperatures by the LAMFE
lated and measured temperature are 0.93 for all three simulation model. These errors or uncertainties are combined results of all
periods, during which the Skills are all 0.98. uncertainties in input data, model parameters, and numerical
8
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 6. Comparison of measured and simulated temperatures at the SE Fork Homosassa Spring, Homosassa Springs at Homosassa Spring, Halls River at Homosassa
Spring, Halls River near Homosassa, and Homosassa River at Homosassa stations during 1/5/2015 through 3/5/2015.

schemes used in the model and they are all within acceptable conditions and other data which drive the LAMFE model for the
ranges. 125-month simulations can be found in Chen (2019a).
As the ecological health of the estuary depends on salinity
4. Effects of SGD on salinity habitats distributions in the waterbody, it is important to ensure that
the estuarine system has a proper salinity gradient along the
With the LAMFE model being well calibrated and verified, water course. Salinity habitats have been used as surrogate in
effects of the SGD reduction on salinity habitats were investi- water resource regulations (Flannery et al., 2007; Ghile et al.,
gated by conducting a series of simulations with various flow 2020; Murphy and Weiland, 2014; Fish and Wildlife Service,
conditions, including the existing flow condition, a baseline flow 2008; Rose et al., 2013). Plenty of water volumes of different
condition, and 12 flow reduction scenarios, ranging from 2.5% salinity zones are critical for various fish species. Generally, three
reduction to 30% reduction with a constant 2.5% increment. The types of fish species exist in a healthy estuary, including freshwa-
baseline flow condition is an imaginary flow condition that would ter, estuarine-resident, and estuarine-dependent species (Peebles,
exist if no ground water were withdrawn in the springshed. It 2005). Freshwater fishes inhabit the tidal freshwater zones and
is estimated that the existing withdrawal causes about 1.85% the upstream non-tidal segments of the estuarine system. Many
reduction of SGDs entering the Homosassa River estuary. As such, freshwater fishes also migrate into and feed in low salinity wa-
the baseline flow is obtained by dividing the existing SGDs by ters, as many of them can tolerate some amount of salt for
0.9815. a short time (Peterson and Meador, 1994). Estuarine-resident
It is desirable to run the Homosassa River LAMFE model for species spend their entire life cycle in the estuary and often have
various flow reduction conditions for as many years as possible, broad salinity tolerances, but they do not migrate away from
if input data which are used to drive the model are available. In the tidal river for feeding or reproduction. Estuarine dependent
this study, the critical data needed for conducting simulations for species typically spend a portion of their early life cycle in the
multiple years are the downstream boundary conditions. With estuary, with a later return to higher salinity coastal waters as
an effort to fill gaps in measured data based on available water they mature. They include many fish and shellfish species of sport
level, salinity, and temperature data in the GOM not far from or commercial importance in coastal areas.
the mouth of the Homosassa River, the boundary conditions at While sufficient water volumes of different salinity zones
the downstream open boundary of the simulation domain be- are needed for various fish types, changes in the bottom areas
came available for a 125-month period from October 9, 2007 to of salinity zones are important to the distribution of benthic
March 12, 2018, allowing a continuous simulation for more than macroinvertebrate communities, as different communities have
10 years. Details on the gap filling in the downstream boundary different levels of salinity tolerance. As such, both water volumes
9
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 7. Time series of water volume, bottom area, and shoreline length of salinity ≤ 2 psu for 0%, 10%, 20%, and 30% reductions from the baseline flow conditions
in the Homosassa River estuary during September 3, 2011 through August 27, 2012.

and bottom areas of various salinity zones should be evaluated only salinity ≤ 2 and 10 psu and only four flow conditions are
in the ecological assessment of the estuary. In addition, shoreline presented here in the figures. Time series plots of salinity habitats
lengths of various salinity zones should also be evaluated, as their for other salinity zones and flow scenarios can be found in Chen
distributions are correlated with different vegetation types. (2019a), though they are presented in a different way.
In this study, water volumes, bottom areas, and shoreline As shown in Fig. 7, ≤2 psu water volumes varied within a
lengths of salinity ≤ 1, 2, 3, 5, 10, and 15 psu were calculated range roughly between 20,000 m3 and 1 million m3 , while ≤2 psu
using a post-processing program from simulated salinities in all bottom areas varied from about 20,000 m2 to about 1 million m2
grid cells that were written out hourly. As mentioned before, during the 360 days. The ≤2 psu shoreline length was roughly
except for days near the end of the dry season, salinity at the between 300 m and 15,000 m. When the baseline flow was
mouth of the Homosassa River estuary is generally lower than 25 decreased by 10%, 20%, and 30%, ≤2 psu water volume, bottom
psu. Water volume, bottom area, and shoreline length habitats for area, and shoreline length were decreased accordingly at each
salinity ≤ 20 psu represent most of the entire estuary. As a result, time point, with all the peaks being evidently reduced for all three
salinity zones with higher upper limits (e.g., ≤20 or 25 psu) are salinity habitats.
not as sensitive to the SGD as those of lower salinity zones and Simulated salinity habitats for salinity ≤10 psu ranged approx-
are not included in the following discussion. imately between 1.6 and 9.0 million m3 , 1.0 and 5.5 million m2 ,
Figs. 7 and 8 are time series of simulated water volumes, and 20 to 78 km for water volume, bottom area, shoreline length,
bottom areas, and shoreline lengths for salinity ≤ 2 and 10 respectively. The relative variabilities of ≤10 psu salinity habitats
psu, respectively during September 3, 2011 through August 27, are much smaller than those of ≤2 psu salinity habitats. Salinity
2012 under the baseline flow condition and three flow reduction habitats of ≤10 psu are not very sensitive to an SGD reduction
conditions: 10%, 20%, and 30%. The top panels in the figures are most of the year; however, during the dry season, starting in late
water volumes, while middle and bottom panels are respectively April and ending in late June, when the region generally receives
bottom areas and shoreline lengths of the corresponding salinity the least rainfall of the year, the effect of an SGD reduction on the
zones. For simplicity, only a randomly selected 360-day period availability of ≤10 psu salinity habitats becomes noteworthy.
are shown in the figures, though flow scenarios were run for a Both the ≤2 and 10 psu salinity habitats show seasonal vari-
simulation period of more than ten years. For the same reason, abilities. During the wet season, there exist much more ≤2 and
10
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 8. Time series of water volume, bottom area, and shoreline length of salinity ≤ 10 psu for 0%, 10%, 20%, and 30% reductions from the baseline flow conditions
in the Homosassa River estuary during September 3, 2011 through August 27, 2012.

10 psu salinity habitats than during the dry season. Because tidal Figs. 9–11 are CDFs of water volume, bottom area, and shore-
signals are included not only in the downstream boundary con- line length for various salinity zones and flow reduction scenarios.
ditions but also in the upstream SGDs, simulated water volume, For a better inspection of the differences among different CDFs
bottom area, and shoreline length for any salinity zones also con- for different flow reductions, only the baseline, existing, 2.5%, 5%,
tain tidal signals with both diurnal and semidiurnal variabilities. 10%, 15%, 20%, 25%, and 30% flow reduction scenarios are included
From Figs. 7 and 8, the diurnal and semidiurnal signals in the in Figs. 9–11. From these figures, it is apparent that habitats of
simulated salinity habitats are weaker than those of weekly vari- lower salinity zones are generally more sensitive to a flow change
abilities, about one half of the spring–neap cycle. This is different than those of higher salinity zone. For example, CDF curves for
from other estuaries, where freshwater inflows are mainly from ≤15 psu water volume, bottom area, and shoreline length are
surface water runoff and thus no tidal signals in the hydrologic only altered slightly for different flow scenarios, but CDF curves
loading. For the Homosassa River, because SGDs are affected for ≤2 psu are changed significantly under different flow con-
by tides, effects of tides on salinity habitats are enhanced and ditions. This property is consistent with other estuarine systems
generally much stronger than these in estuaries without SGD evaluated in southwest Florida, which receive hydrologic loadings
being the major source for hydrologic loading. mainly from surface water runoff. In other words, regardless of
Direct comparisons and analyses of hourly model results of the source of freshwater inflow or if the inflow contains tidal
water volume, bottom area, and shoreline length for different signals, lower salinity habitats are more sensitive than those of
flow reductions are very tedious and not necessary, because one other salinity zones. Nevertheless, there is an exception for the
would easily lose the forest for the trees. While the temporal Homosassa River, where salinity habitats of ≤1 psu is not very
variation of salinity zones in the estuary may have a time scale sensitive to SGD reductions. The reason for the relatively weak
of hours or days, the time scale of the response of various species sensitivity of ≤1 psu salinity habitats to flow is that these near
to salinity changes is normally much longer. As such, cumulative freshwater habitats are mainly found in the SE Fork of the river
distribution functions (CDFs) of salinity habitats are often gener- (Fig. 12), which contributes less than half of the total hydrologic
ated and analyzed to get an overall picture of the effects of flow loading to the estuary. On the other hand, the Halls River receives
reductions on salinity habitats (Flannery et al., 2007; Ghile et al., SGDs that are much saltier than SE Fork SGDs. As mentioned
2020). before, SGDs entering the Hall River are to some degree salinity
11
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 9. Cumulative distribution functions of simulated water volumes for various flow reductions for salinity ≤ 1, 2, 3, 5, 10, and 15 psu.

source for the Hall River and thereby for the entire estuary. Fig. 12 is a map showing the long-term mean of simulated
Because SGDs from all the spring vents are reduced by the same depth-averaged salinity during the 125-month simulation period
percentage from the baseline condition in the flow reduction for the baseline flow condition in the estuary. Although the actual
simulations, saltier water inflow to Halls River was reduced while salinity distribution varies with time, Fig. 12 provides some gen-
freshwater inflow to the SE Fork was reduced. The combined eral information about the area of different salinity zones most
effect is the relatively weak sensitivity of ≤1 psu salinity habitats of time in the estuary. The 2 psu isohaline is generally located
to the SGD. upstream of the confluences of Halls River with the Homosassa
From Figs. 9–11, one can see that ≤2 and 3 psu salinity River, while the 3 psu isohaline moves within a segment down-
habitats within the middle half of the range are much more stream of the Halls River confluence. One can see that there
sensitive to an SGD reduction than those above the 90th per-
is narrow transect downstream of the Halls River confluence,
centile, which occurs mainly during the wet season. While ≤2 psu
where the Homosassa River becomes not only deeper but also
salinity habitats are most sensitive to an SGD reduction near the
much wider as one travel from upstream to downstream. This
70th percentile, ≤3 psu salinity habitats are most sensitive near
the 25th percentile. For ≤5 psu salinity habitats, they are most bathymetric change affects the sensitivity of the ≤2 and 3 psu
sensitive at near the 85th percentile; however, the most sensitive water volumes and bottom areas to the SGD. In Fig. 9, one can
≤10 psu salinity habitats occur near the 15th percentile. Such a see that the 70th percentile of ≤2 psu water volume is about
response of salinity habitats to an SGD reduction is associated 1.1 million m3 , which is about the same as the 25th percentile
with the location of the individual salinity isohaline, which varies of ≤3 psu of water volume. Similarly, one can see in Fig. 10 that
depending on the tides and salinity at the downstream bound- the 70th percentile of ≤2 psu bottom area is about 0.65 million
aries as well as SGDs received. To this end, the physical setting of m2 , which is roughly the same as the 25th percentile of ≤3 psu
the Homosassa River also plays a big role influencing the salinity bottom area. In other words, the 70th percentile of ≤2 psu and
isohaline and the response of the salinity habitats. the 25th percentile of ≤3 psu represent approximately the same
12
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 10. Cumulative distribution functions of simulated bottom areas for various flow reductions for salinity ≤ 1, 2, 3, 5, 10, and 15 psu.

area of the estuarine system, which includes a short downstream To quantify the long-term changes of salinity habitats caused
segment of the Halls River and all the areas of the Homosassa by a flow reduction, average water volume, bottom area, and
River that are upstream of the narrow cross section located just shoreline length for different salinity zones can be calculated over
downstream of the Halls River confluence. the entire 125-month period. Numerically, these average values
The 5 psu contour isohaline is generally located near the are equal to areas between the y-axis and the corresponding
town Homosassa. During the dry days, it can migrate toward the CDF curves shown in Figs. 9–11. Averages of ≤1, 2, 3, 5, 10,
Halls River confluence; however, during the wet days, it can be and 15 psu salinity water volumes, bottom areas, and shoreline
pushed to cross sections downstream of the town Homosassa and lengths for various SGD reductions are plotted in Fig. 13. As
into several distributaries that branch off the Homosassa River, can be seen from the figure, no obvious refection points can be
including the Salt River and Mason Creek. On the other hand, found in these curves representing relationships between salinity
the 10 psu isohaline generally occurs downstream of these dis- habitats and the percentage SGD reduction and the relationships
tributaries. In relatively dry days, the 10 psu isohaline will retreat are practically linear. With a linear regression, the R2 values of
toward the town Homosassa and into the distributaries. It turns the linear model are all higher than 0.99, except for ≤1 salinity
out that these distributaries can cause ≤5 and 10 psu salinity water volume, bottom area, and shoreline length, which have R2
habitats to be very sensitive to a flow reduction. With about at 6.0 values of 0.97, 0.96, and 0.96, respectively. The slope of each re-
million m3 , 3.4 million m2 , and 50 km of water volume, bottom gression line represents response of the long-term average of the
area, shoreline length respectively, the 85th percentile of ≤5 psu simulated salinity habitat to an SGD change. Using the long-term
salinity habitats cover basically the same area of the estuary as averages of simulated water volume, bottom area, or shoreline
that of the 15th percentile of ≤10 psu salinity habitats. At these length for salinity ≤ 1, 2, 3, 5, 10, and 15 psu under the baseline
numbers, an SGD increase or decrease would cause the most flow condition to normalize the slopes, normalized response rates
relative increase/decrease for ≤5 and 10 psu salinity habitats. can be calculated and listed in Table 2. In the table, BLV denotes
13
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 11. Cumulative distribution functions of simulated shoreline length for various flow reductions for salinity ≤ 1, 2, 3, 5, 10, and 15 psu.

Table 2
Simulated salinity habitats under the baseline flow condition (BLV) and normalized response rate
for different salinity zones in the Homosassa River estuary.
Salinity zone Water volume Bottom area Shoreline length
BLV (mil m3 ) NRR BLV (mil m2 ) NRR BLV (km) NRR
≤1 psu 0.02 0.41 0.02 0.31 1.62 0.25
≤2 psu 0.47 1.38 0.32 1.38 6.73 0.96
≤3 psu 1.00 1.22 0.73 1.25 13.85 1.02
≤5 psu 1.95 1.00 1.52 1.01 28.42 0.92
≤10 psu 3.85 0.43 3.05 0.42 54.32 0.38
≤15 psu 5.30 0.15 4.11 0.14 68.62 0.11

the 125-month average of simulated water volume, bottom area, will cause 1.38% change of ≤2 psu water volume but only 0.43%
or shoreline length for any of the salinity zones under the baseline change of ≤10 psu water volume. Generally, oligohaline habitats
flow condition, while NRR is the normalized response rate, a in the Homosassa River estuary increase/decrease roughly 1% or
dimensionless parameter representing the sensitivity of salinity more with a 1% increase/decrease of SGD. However, as the salinity
habitats to the SGD change. zone is expanded to include mesohaline, the sensitivity becomes
From Table 2, it can be seen that except for the ≤1 salinity weaker and weaker. The most sensitive salinity zones are ≤2 psu
habitats, the sensitivity of the simulated salinity habitat decreases for the water volume and bottom area. For the shoreline length, it
as the upper limit of the salinity zone increases. For example, the is the ≤3 psu salinity zone the most sensitive to an SGD change.
NRR for ≤2 psu water volume is 1.38, while NRR for ≤10 psu The ≤2 psu water volume and bottom area are also the most
water volume is 0.43. In other words, every 1% of SGD change sensitive among all the salinity habitats studied here.
14
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Fig. 12. Depth-averaged, 125-month mean salinity distribution during October 9, 2007 to March 12, 2018 in the Homosassa River for the baseline flow condition.
White contour lines are spaced with a 1 psu interval.

Fig. 13. Relationships of salinity habitats with SGD reduction in the Homosassa River Estuary. Sal_1, Sal_2, Sal_3, Sal_5, Sal_10, and Sal_15 in the legends represent
salinity ≤ 1, 2, 3, 5, 10, and 15 psu.

From Table 2, one can see that the water volume and bottom length in these other estuaries. In the Homosassa River, however,
for each of the salinity zone have about the same NRR, which the estuary is shallow and except for the navigation channel and
is generally higher than that for the shoreline length in the
some deep holes, most riverbed varies between a small range
Homosassa River estuary. This is slightly different from other es-
tuaries in southwest Florida, where the water volume of a salinity around 0.5 m deep. Because the estuary is generally well-mixed,
zone is generally more sensitive to the freshwater inflow than the the water volume of a salinity zone varies in a similar way as
bottom area of the same salinity zone, which is more sensitive
the bottom area and both are just two-dimensional functions of
than the shoreline length. The reason for such a response is that
water volume, bottom area, and shoreline length of a salinity length. As such, the water volume and bottom area for the same
zone are respectively three-, two-, one-dimensional functions of salinity zone have roughly the same sensitivity to an SGD change.
15
X. Chen Regional Studies in Marine Science 43 (2021) 101674

5. Conclusions Chen, X., 2004b. Modeling hydrodynamics and salt transport in the Alafia
River estuary, Florida during 1999–2001. Estuar. Coast. Shelf Sci. 61 (2004),
477–490.
The laterally averaged hydrodynamic model LAMFE was ap-
Chen, X., 2007. Dynamic coupling of a three-dimensional hydrodynamic model
plied to the Homosassa River estuary to simulate circulations, with a laterally averaged, two-dimensional hydrodynamic model. J. Geophys.
salinity transport processes, and thermal dynamics in the spring- Res. 112, C07022. http://dx.doi.org/10.1029/2006JC003805.
fed estuary on the Gulf coast of Florida peninsula. SGDs received Chen, X., 2008. Simulating hydrodynamics in the upper charlotte harbor and
by the estuary were either recorded at proper transects down- its major tributaries in Florida. J. Coastal Res. 2008 (10052), 149–162. http:
//dx.doi.org/10.2112/1551-5036-52.sp1.149.
stream of spring vents using ADCPs or estimated using regression Chen, X., 2010. Hydrodynamic Simulations of the Lower Peace River – Lower
equations relating spring flows with tides and groundwater levels Myakka River – Upper Charlotte Harbor System in Support of Determining
in the region. The model was calibrated and verified against real- Minimum Flows for the LPR and LMR in Southwest Florida. Technical Report,
time data of water level, salinity, and temperature collected at Southwest Florida Water Management District, Tampa, Florida.
Chen, X., 2012a. Simulating Hydrodynamics and Salinity Transport Processes
five stations within the simulation domain. Various skill assess-
in the Manatee/Braden River Estuarine System using a Multi-Block Model.
ments were conducted and have shown that the LAMFE performs Technical Report, Southwest Florida Water Management District, Tampa,
very well for the spring-fed estuary. Florida.
As many estuarine species depend on the availability of differ- Chen, X., 2012b. Simulating hydrodynamics in a spring-fed estuary using a three-
ent salinity zones, salinity habitats, especially oligohaline salinity dimensional unstructured Cartesian grid model. Estuar. Coast. Shelf Sci. 115,
246–259. http://dx.doi.org/10.1016/j.ecss.2012.09.007.
habitats, which are often used as surrogate for determining the Chen, X., 2014. Estimate submarine groundwater discharge to crystal river/kings
health of the estuary, need to be protected. For this purpose, this bay in florida with the help of a hydrodynamic model. J. Mar. Sci. Eng. 2,
study investigated the impact of reduced SGDs on the salinity 66–80. http://dx.doi.org/10.3390/jmse2010066.
habitats in the Homosassa River using the calibrated/verified Chen, X., 2019a. Hydrodynamic Modeling of Effects of Flow Reduction on Salinity
and Thermal Habitats in the Homosassa River. Technical Report, Southwest
LAMFE model. Salinity habitats considered here included ≤1, 2,
Florida Water Management District, Tampa, Fla, p. 247.
3, 5, 10, and 15 psu water volumes, bottom areas, and shoreline Chen, X., 2019b. Will a Reduction of Submarine Groundwater Discharge
lengths. Cause Salinity in SGD to Increase in the Chassahowitzka and Homosassa
Model results show that different salinity habitats respond to Rivers? Technical Memorandum, Southwest Florida Water Management
the SGD differently in the Homosassa estuary. Over the short District, Tampa, Florida, p. 14.
Chen, X., Flannery, M.S., 1997. Use of a hydrodynamic model for establishing a
term, the response is greatly affected by some special physical minimum freshwater flow to the lower hillsborough river. In: Spaulding,
features of the waterbody and the SGD that the estuarine system M.L., Blumberg, A.F. (Eds.), Proceedings of ASCE Estuarine and Coastal
receives. The sensitivity of the instantaneous salinity habitat to Modeling 5, pp. 663–678.
the SGD varies with the location of the isohaline (and thereby the Chen, X., Flannery, M.S., Moore, D.L., 2000. Salinity response to the change of the
upstream freshwater flow in the Lower Hillsborough River, Florida. Estuaries
percentile of the habitat) in relationship with these physical fea-
23, 735–742.
tures. For example, the response of ≤2 and 3 psu salinity habitats Fish and Wildlife Service, 2008. Biological opinion on the proposed coordinated
is affected by a cross section change, which is a short distance operations of the central valley project and state water project. Avail-
downstream of the Halls River confluence with the Homosassa able from http://www.fws.gov/sfbaydelta/documents/SWP-CVP_OPs_BO_12-
River. For ≤5 and 10 psu salinity habitats, their sensitivities are 15_final_OCR.pdf.
Flannery, M.S., Chen, X., Hey, M., Munson, A., Dachsteiner, M., 2007. The Deter-
affected by the distributaries located downstream of the town mination of Minimum Flows for the Lower Alafia River Estuary. Southwest
Homosassa. Florida Water Management District Brooksville, Florida, p. 393.
Model results also show that long-term averages of salinity Ganju, N.K., Hayn, M., Chen, S.-N., Howarth, R.W., Dickhudt, P.J., Aretxa-
habitats in the Homosassa River are linearly decreased with an baleta, A.L., Marino, R., 2012. Tidal and groundwater fluxes to a shallow,
microtidal estuary: Constraining inputs through field observations and hy-
increasing SGD reduction percentage. Normalized response rates
drodynamic modeling. Estuar. Coasts 35 (5), 1285–1298. http://dx.doi.org/10.
to an SGD change were calculated for all the salinity zones con- 1007/s12237-012-9515-x.
sidered in order to quantify sensitivities of different salinity zone Ghile, Y.B., Chen, X., Leeper, D.A., Anastasiou, C., Deak, K., 2020. Recommended
to the SGD. It was found that water volumes and bottom areas of Minimum Flows for the Lower Peace River and Proposed Minimum Flows
the same salinity zone are similarly sensitive to an SGD change for Lower Shell Creek. Draft Report, Southwest Florida Water Management
District, Brooksville, Florida, p. 172.
and both are more sensitive than the shoreline length. Except for
Hamrick, J.M., 1992. A three-dimensional environmental fluid dynamic computer
those of ≤1 psu, oligohaline salinity habitats are more sensitive code: Theoretical and computational aspects. Special Report 317, The College
to SGD than higher salinity zones. Salinity habitats for ≤2, 3, and of William and Mary, Virginia Institute of Marine Science, VA.
5 psu linearly increase/decrease about 1% or more with every 1% HSW Engineering, Inc., 2011. A Modeling Study of the Relationship of Freshwater
Flow with the Salinity and Thermal Characteristic of the Homosassa River.
increase/decrease of SGD. The ≤2 psu water volume and bottom
Prepared for Southwest Florida Water Management District, Tampa, Fla., p.
area are the most sensitive habitats to a SGD change and they are 33637.
reduced about 1.38% for every 1% reduction of the SGD. Jay, D.A., Smith, J.D., 1990. Residual circulation in shallow estuaries, 2, Weakly
stratified and partially mixed, narrow estuaries. J. Geophys. Res. 95, 733–748.
Declaration of competing interest Johnson, B.H., Kim, K.W., Heath, R., Hsieh, B., Butler, L., 1991. Validation of a
three-dimensional hydrodynamic model of Chesapeake Bay. J. Hydraul. Eng.
117 (1), 2–20.
The authors declare that they have no known competing finan- Leeper, D.A., Flannery, M.S., Heyl, M.G., Basso, R., Kelly, M., 2012. Recommended
cial interests or personal relationships that could have appeared Minimum Flows for the Homosassa River System. Southwest Florida Water
to influence the work reported in this paper. Management District, Brooksville, Fla, p. 223.
Murphy, D.D., Weiland, P.S., 2014. The use of surrogates in implementation of
the federal Endangered Species Act—proposed fixes to a proposed rule. J.
References Environ. Stud. Sci. 4, 156–162. http://dx.doi.org/10.1007/s13412-014-0167-y.
Peebles, E.B., 2005. An Analysis of Freshwater Inflow Effects on the Early Stages
Blumberg, A.F., Kim, B.N., 2000. Flow balance in st. Andrew bay revealed through of Fish and their Invertebrate Prey in the Alafia River Estuary. Report
hydrodynamic simulations. Estuaries 23, 21–33. Prepared By the University of South Florida College of Marine Science for the
Chen, X., 2003. An efficient finite difference scheme for simulating hydrody- Southwest Florida Water Management District, Brooksville, Florida, p. 147.
namics in narrow rivers and estuaries. Internat. J. Numer. Methods Fluids Peterson, M.S., Meador, M.R., 1994. Effects of salinity on freshwater fishes in
42, 233–247. coastal plain drainages in the southeastern U.S. Rev. Fish. Sci. 2 (2), 95–121.
Chen, X., 2004a. Using a piecewise linear bottom to _t the bed variation in a http://dx.doi.org/10.1080/10641269409388554.
laterally averaged, z-co-ordinate hydrodynamic model. Internat. J. Numer. Prandle, D., 1985. On salinity regimes and the vertical structure of residual flows
Methods Fluids 44 (2004), 1185–1205. http://dx.doi.org/10.1002/_d.680. in narrow tidal estuaries. Estuar. Coast. Shelf Sci. 20, 615–635.

16
X. Chen Regional Studies in Marine Science 43 (2021) 101674

Rose, K.A., Kimmerer, W.J., Edwards, K.P., Bennett, W.A., 2013. Individual-based Willmott, C.J., 1981. On the validation of models. Phys. Geogr. 2, 184–194.
modeling of delta smelt population dynamics in the upper San Francisco Yobbi, D.K., Knochenmus, L.A., 1989. Salinity and Flow Relations and Effects of
estuary: I. Model description and baseline results. Trans. Am. Fish. Soc. 142, Reduced Flow in the Chassahowitzka River and Homosassa River Estuar-
1238–1259. ies. Southwest Florida, Water-Resources Investigations Report 88-4044, U.S.
Warner, J.C., Sherwood, C.R., Signell, R.P., Harris, C.K., Arango, H.G., 2008. Geological Survey, Tallahassee, Fla., p. 42.
Development of a three-dimensional, regional, coupled wave, current, and
sediment-transport model. Comput. Geosci. 34, 1284–1306.

17

You might also like