Habitat Suitability Modeling and Mapping to Assess the Influence of Freshwater Withdrawals on Spatial Distributions and Population Numbers of Estuarine Species in the Lower Peace River and Charlotte Harbor, Florida (2024)

Search for other works by this author on:

Article history

Abstract

The assessment and management of freshwater inflow to estuaries have received increased emphasis in recent decades to account for the important ways in which freshwater inflow affects physical, chemical, and biological processes and the resources of estuaries, including relationships with the productivity of sport and commercial fisheries (Drinkwater and Frank 1994; Estevez 2002; Powell et al. 2002; Olsen et al. 2006; Gillson 2011; Adams 2014). Alber (2002) proposed a conceptual model to support management of freshwater inflows by establishing inflow standards that help to protect resources and functions of estuaries. The management approach can be inflow based (flow is kept within some prescribed bounds under the assumption that taking too much away is injurious to biological resources), condition based (inflow standards are set in order to maintain specified conditions in the estuary), or resource based (inflow standards are set based on the requirements of specific resources). Each approach is carried out by regulating reductions or other alterations of freshwater inflow.

In 1972, the Florida Legislature directed the five Florida water management districts to establish minimum flows and levels (MFLs) for rivers and streams within their boundaries (State of Florida 1972). As currently defined by statute, “the minimum flow for a given watercourse shall be the limit at which further withdrawals would be significantly harmful to the water resources or the ecology of the area.” The water management districts have taken different approaches to comply with the legislation (Alber 2002).

The Southwest Florida Water Management District (SWFWMD), due to its responsibility to permit the consumptive use of water and the legislative mandate to protect water resources from “significant harm,” has established minimum flows for the free‐flowing lower Peace River, which drains into Charlotte Harbor on the southwest coast of Florida (SWFWMD 2010). The SWFWMD has been conducting a re‐evaluation of the minimum flows and developing new ones for the lower portion of Shell Creek, a tributary to the lower Peace River that enters the river 12 km upstream of the river mouth. The Peace River and Shell Creek are both used for municipal water supplies. To evaluate minimum flows, historic freshwater withdrawals were added back into the flow records for each source and other hydrologic adjustments were made to create a baseline flow record that reflects natural flow conditions. A variety of analytical approaches was then used to evaluate the effects of different daily flow reductions on salinity, water quality, and various biological parameters in the tidal reaches of the lower Peace River and Shell Creek to determine the total amount of water available for withdrawal without causing significant harm to environmental resources (SWFWMD 2020).

Fundamental to the approach for developing MFLs is the understanding that freshwater flow regimes possessing largely natural patterns of variation are important for maintaining the ecological characteristics and productivity of riverine and estuarine systems (Browder 1991; Livingston 1997; Poff et al 1997; Richter et al. 1997; Flannery et al. 2002; Mattson 2002; Gillson 2011). By simulating maximum allowable percentage daily withdrawals from the baseline flow record of an unimpounded river, the evaluation of minimum flows for the Peace River could be considered a determination of “environmental flows,” the term that is now used to describe the timing, quality, and quantity of flows needed to sustain freshwater and estuarine ecosystems (Arthington 2012; Poff et al. 2017). However, the term “minimum flows” is used herein because it is established in Florida statutes and is the term given to the corresponding regulatory rules.

The locations of early life stages of fish and invertebrate species vary along the salinity gradient in the lower Peace River (Greenwood et al. 2004; Idelberger and Greenwood 2005; Greenwood 2007; Peebles et al. 2007; Stevens et al. 2013; Call et al. 2013). Greenwood et al. (2004) found that early life stages of 14 fish and invertebrate species had differing responses to freshwater inflow in the lower Peace River. None of these studies used GIS to map seasonal changes in spatial distributions of species life stages in the lower Peace River. Although numerous agencies have fisheries‐independent monitoring (FIM) programs, most do not use the data collected to estimate population numbers for the species that are found in the estuaries they monitor.

A new approach for estimating population numbers in estuaries was developed for juvenile pink shrimp Farfantepenaeus duorarum in Tampa Bay (Rubec et al. 2016a). Habitat suitability modeling (HSM) used delta‐type generalized additive models (GAMs) associated with GAMLSS software in R to create seasonal maps of species distributions and population abundance. Environmental data points were interpolated to create habitat grids. Gear‐corrected (GC) CPUEs from fitted splines and graphs derived from the habitat suitability models were assigned to corresponding cells in the habitat grids to create seasonal grids containing predicted GC‐CPUEs for 87 species × life stage combinations (hereafter, “species life stages”). The seasonal grid cells were then averaged to create continuous GC‐CPUE grids, which were partitioned into zones to create seasonal HSM maps for each species life stage. This provided a means to visualize the spatial distribution of mean GC‐CPUEs in each zone. Population number estimates were derived from the mean GC‐CPUEs associated with the HSM zones. The approach was applied to the Charlotte Harbor system (Rubec et al. 2019).

The main goal of the present study was to assess the influence of changes in freshwater inflows on spatial distributions and population abundance of fish and invertebrate species in the lower Peace River and Charlotte Harbor. We used HSM with GIS to assess the effects of freshwater withdrawals on fish and invertebrate species life stages in the study area. The first step was to map seasonal temperature and salinity conditions associated with baseline conditions (no water withdrawals) and minimum flows (with water withdrawals). A second step was to estimate relative population numbers from seasonal HSM maps to determine the impacts of water withdrawals on early life stages of fish and invertebrate species (biological resources). The third step was to use the HSM maps to spatially elucidate species life history patterns and compare those findings with the published literature. Since the abundance of species life stages changes between seasons and across years, modeling and mapping methods were required to separate climatic effects from the effects associated with water withdrawals.

METHODS

Fisheries‐independent monitoring

The main source of data for early life stages of fish species and blue crab Callinectes sapidus has been FIM data collected in the Charlotte Harbor study area (Figure 1). The FIM sampling has been conducted north of Pine Island within Charlotte Harbor, the lower Myakka River, and the Lower P segment of the lower Peace River and lower Shell Creek. Seasonal data extracted from the FIM database for 1996–2013 included catch numbers, effort, and physicochemical data (temperature, salinity, dissolved oxygen, and depth) along with the date, latitude/longitude, and associated gear types. The Upper P segment of the lower Peace River, north of its confluence with Shell Creek, was not part of long‐term FIM. For this segment, we used FIM data that were collected during two special studies conducted from April 1997 to March 1998 and from July 2007 to June 2010 (Stevens et al. 2013).

Sampling gears used in the study area from 1996 to 2013 included a 21.3‐m circular bag seine, a 21.3‐m boat bag seine, a 183‐m haul seine, and a 6.1‐m otter trawl. A 61‐m haul seine for sampling in the lower Peace River was added in 2007. As is typical of FIM data, catches have a high percentage of zeroes (no fish collected in samples). Gear corrections were used to standardize the CPUEs for different gear types (Robson 1966). This approach has been used with previous HSM studies in Florida estuaries (Rubec et al. 2001, 2006, 2016a, 2016b, 2019). To create GC‐CPUEs, gears with lower mean CPUEs were standardized within the R program GAMLSS by the ratio of each gear’s mean CPUE to the gear with the highest mean CPUE. Gear corrections were applied to seasonal data sets for all species life stages. Gear codes were used to illustrate which gear types were used in final habitat suitability models (Table 1). The ordering of gear codes from high to low indicates the gear types (first one for each species life stage) used for gear type standardization. The high percentage of zeroes necessitated the use of delta‐gamma GAMs to deal with zero inflation. Sample numbers indicate that sample sizes were adequate for all gear type combinations.

A seasonal approach was taken because the life stages of the species of interest show peak abundances in the Charlotte Harbor system during different times of year (Greenwood et al. 2004; Idelberger and Greenwood 2005; Peebles et al. 2006). In addition, salinity and water temperature exhibit typical seasonal variations in response to regional climatic patterns, including seasonal variations of rainfall and freshwater inflow. By applying the percent‐of‐flow approach (Flannery et al. 2002), the withdrawal of proportionately more freshwater inflow was simulated during seasons of the year when high flows were most common, suggesting that the wet seasons should be separately examined. On the other hand, the area and volume of salinity‐based habitats are generally more susceptible to impacts from a given percent flow reduction during the dry times of the year (SWFWMD 2010, 2020). Major episodic flow events, such as floods and droughts, can greatly influence the distribution and abundance of fish and invertebrate species and community composition in estuaries ( Livingston 1997; Whitfield 2005; Gillson 2011; Stevens et al. 2013). Given that this project allowed for the generation of a limited number of HSM maps, it was determined that average salinity and temperature conditions for each season would best be applied to represent typical changes in habitats between baseline and minimum flows, with the evaluation of specific high‐ or low‐flow events possible for future assessment.

Considering these factors, four 3‐month seasons were chosen to reflect seasonal changes in water temperature, salinity, and rainfall conditions in the region (PBS&J International 1999; Flannery et al. 2002). Water temperatures and freshwater inflows are greatest in the summer (July–September), which typically has the highest rainfall. The fall (October–December) has declining water temperatures and declining freshwater inflow, often followed by a minor increase in inflow during the winter (January–March) due to rains associated with cold fronts. However, from 2007 to 2014 the winters were unusually dry and had the lowest average inflow of the four seasons. The spring (April–June) has rising water temperatures and typically includes the lowest inflows, although inflows usually increase in mid‐June as the rainy season begins.

Habitat mapping

Bottom sediment types at FIM sampling locations were extracted from a National Oceanic and Atmospheric Administration (NOAA) fishing chart created in 1989. The NOAA map is based on mud/sand distributions determined using a plumb line dropped onto the bottom to assess the firmness of bottom sediments at stations across the estuary. When we initiated the present study in 2015, the NOAA sediment map was the best information available.

A bottom type grid/map was created with polygons coded for mud, sand, and submerged aquatic vegetation (SAV). Bottom types for mud and sand were digitized as polygons from the NOAA fishing chart for Charlotte Harbor (Rubec et al. 2019). Seagrass coverages in Charlotte Harbor were derived from images obtained using aerial photography conducted every 2 years since 2002 (Photo Science and Kaufman 2013). Most of the imagery was collected during winter. The imagery for SAV showed little or no change in the spatial extent of SAV from 2002 to 2013. Based on good water clarity, we chose the 2012 coverage as being most representative of the spatial extent of SAV. A bottom type grid was created using ArcGIS (ESRI 2014).

Bathymetry data derived from a sonar survey conducted in 2012 by Wang (2013) were obtained from the SWFWMD for the Charlotte Harbor study area. Additional data for Gasparilla Sound were obtained from NOAA for areas that were not described in the SWFWMD data set (Rubec et al. 2018). This included data obtained from hydrographic surveys conducted by NOAA in 1955 and 1956. The bathymetry data were merged into a single point feature class. Large backwaters and canals with no bathymetry data were removed. Some smaller backwaters were included. The bathymetry data points were interpolated in ArcGIS using empirical Bayesian kriging (Krivoruchka 2012). The output raster grid for bathymetry was clipped to the water extent within the study area.

We extracted averaged surface and bottom dissolved oxygen data from the FIM database using the Statistical Analysis System (SAS Institute, Cary, North Carolina). Point data collected from 1996 to 2013 at FIM sampling stations were interpolated in ArcGIS using empirical Bayesian kriging associated with the Geostatistical Analyst 10.3 extension to create seasonal dissolved oxygen grids (ESRI 2014). The seasonal dissolved oxygen grids were clipped to the same spatial extent as the bathymetry grid and bottom type grid, with each grid containing about 1.9 million total 15‐ × 15‐m cells. Since the methods are fairly complicated, a diagram outlining the methods is presented in Figure 2.

Hydrodynamic modeling using two flow scenarios

The effects of reductions in freshwater inflow to the estuary were evaluated for the years 2007–2014 using the percent‐of‐flow approach in which various percentage flow reductions were applied to daily flows in the baseline flow record (SWFWMD 2020). Based on other analyses that assessed changes in the volume, area, and shoreline length of various salinity zones, a minimum flows scenario was created by SWFWMD in which daily flow reduction percentages were applied within three flow ranges for both the lower Peace River and lower Shell Creek, ranging from 13% of low flows to 40% of high flows. On the lower Peace River, a low‐flow cutoff was also employed that prohibits any withdrawals, which was in effect on 26% of the days during the study period, while the 40% limit for high flows was applied to 22% of the days. Using this withdrawal schedule, the greatest differences in daily flows between the baseline and minimum flows scenarios were during periodic high‐flow events that were most common in the summer wet season (Figure 3). Much smaller flow reductions, both in terms of differences in daily flows and percent daily flow reductions, occurred during most of the year. The SWFWMD later added a total maximum withdrawal limit of 11.33 m³/s (400 ft³/s) between baseline and minimum flows in the lower Peace River, which was neither included nor simulated in our study.

To simulate the effects of these flow reductions on salinity distributions in tributary rivers, a dynamically coupled three‐dimensional (3D)–two dimensional vertical (2DV) model was developed by SWFWMD (Chen 2020). The Unstructured‐grid Lake and Estuarine Simulation System (UnLESS) hydrodynamic model, which dynamically couples the Laterally Averaged Model For Estuaries (LAMFE) and Lake and Estuarine Simulation System in Three Dimensions (UnLESS3D), was applied to greater Charlotte Harbor (Figure 4) using 4,790 grids in the horizontal plane and 17 layers in the vertical direction to discretize the 3D simulation domain, and 311 grids and 17 layers to discretize the 2DV simulation domain. The simulation domain for greater Charlotte Harbor included tidally influenced sections of the lower Peace River, lower Myakka River, and lower Shell Creek and the estuary extending southward past the mouth of the Caloosahatchee River and offshore about 20–30 km into the Gulf of Mexico. The tidally influenced sections included a 34.2‐km section of the lower Peace River, a 38.6‐km section of the lower Myakka River, and a 10‐km section of lower Shell Creek.

While the greatest changes in salinity and temperature distributions were mostly within the tidal rivers, the area modeled using UnLESS included greater Charlotte Harbor to eliminate the effect of barriers from predictions for salinity and temperature patterns in the tidal portion of the rivers (Chen 2020). Salinity and temperature fields were simulated in greater Charlotte Harbor and its major tributaries from 2007 to 2014 for both the baseline and minimum flow scenarios. The UnLESS model was used to predict temperature and salinity for each hour in a day (24 h) for 90 d within each season.

For the present study, seasonal files with 2,160 predictions for temperature and salinity within the Charlotte Harbor study area (Figure 1) were obtained for each year from 2007 to 2014 for both baseline and minimum flows (Rubec et al. 2018). Seasonal values for predicted salinity and temperature were averaged across years (2007–2014) for both scenarios. The point data were then interpolated using empirical Bayesian kriging to create seasonal temperature grids and seasonal salinity grids with the same cell size and spatial extent as the bathymetry grid, bottom type grid, and seasonal dissolved oxygen grids. For the spatial analyses, we held the bathymetry, bottom type, and seasonal dissolved oxygen grids constant between baseline and minimum flows. Only the seasonal temperature grids and seasonal salinity grids changed between the two scenarios.

Estuarine species life stages

Eight species life stages were selected based on the criterion that they (1) exhibit preferences for low or moderate salinity and (2) have been found to be abundant in the Charlotte Harbor study area. Six species life stages exhibited affinities for low salinity in a previous HSM study in Tampa Bay and Charlotte Harbor (Rubec et al. 2016b). These species were seasonally analyzed as early juvenile (EJ), juvenile (J), and adult (A) life stages (Table 2). The Hogchoker and blue crab were added because they have been found to exhibit affinities for low salinity (Flaherty and Guenther 2011; Stevens et al. 2013; Doering and Wan 2018). For these two species, juvenile and adult life stages were combined (JA) and analyzed together.

For the present study, the term “estuarine residents” refers to species life stages for Hogchoker, Sand Seatrout, blue crab, Southern Kingfish, and Bay Anchovy, which are abundant in the Charlotte Harbor study area during most seasons of the year (Idelberger and Greenwood 2005; Stevens et al. 2013). Estuarine transients, such as Red Drum and Spot, occupy nearshore waters of the Gulf of Mexico as adults but are abundant in the lower portions of rivers as juveniles during discrete time periods corresponding with the species’ spawning cycles (Stevens et al. 2013).

Habitat suitability modeling

The physicochemical data at FIM sampling locations were used with HSM to relate CPUEs to environmental conditions. Seasonal delta‐gamma GAMs were developed that relate GC‐CPUEs to environmental data collected in Charlotte Harbor. We used the online R‐based program GAMLSS (Rigby and Stasinopoulos 2005; Stasinopoulos and Rigby 2007) that was designed for data sets with a surplus of zero catch values (i.e., zero‐inflated data) and that was previously applied to FIM data from Tampa Bay (Rubec et al. 2016a). Log‐transformed cubic smoothing splines were fitted to nonzero (positive CPUE) data (mu) and logit‐transformed splines were fitted to probability of zero occurrence (P = 0) data (nu) across environmental gradients for both full and reduced models. The spline data were then back‐transformed and the two components were multiplied (mu × nu) to derive seasonal GC‐CPUE splines across gradients for water temperature, salinity, dissolved oxygen, and depth. Histograms were created for categorical variables with mean GC‐CPUEs by bottom type, gear type, and year. Predictions based on the combined mu and nu models account for uncertainty in predicted GC‐CPUEs. The HSM analyses and mapping methods using FIM data were described in more detail by Rubec et al. (2016a, 2016b, 2019).

First, a full model (i.e., all predictor variables included) was fitted using the penalized B‐spline. Next, 31 reduced models were developed, consisting of various combinations of one to five environmental factors, each with a different value of Akaike’s information criterion (AIC). The model with the lowest AIC was chosen as the final reduced model. Depending on selectivity for the size of the species life stage being analyzed, not all gear types were used for each HSM analysis (i.e., if a gear type did not catch any individuals of the species life stage being analyzed, that gear type was not included in the final analysis).

The delta‐gamma GAMs developed using reduced models were not spatial. A second part of the R program GAMLSS used GC‐CPUEs derived from fitted cubic smoothing splines to assign GC‐CPUEs to a data set representing the centers of the cells associated with the habitat grids, according to latitude and longitude. Unique combinations of values for temperature, salinity, oxygen, depth, and the three categorical variables were available for each grid cell. The GC‐CPUEs associated with the cell coordinates were then averaged to create a data set with mean GC‐CPUEs. Seasonal mean GC‐CPUE data sets for each species life stage were then imported into ArcGIS, and continuous GC‐CPUE grids were created across the estuary (the R code is available from gislibrarian@myfwc.com).

Zonal grids used to create habitat suitability modeling maps

Using the Slice tool in ArcGIS Spatial Analyst, the continuous GC‐CPUE grids were assigned to four habitat suitability zones by the Jenks (1967) natural breaks classification method. The natural breaks method associated with the Slice tool specifies “that the classes will be based on natural groupings inherent in the data.” Break points are identified by choosing the class breaks that best group similar values and that maximize the differences between classes. This provided an objective means of partitioning continuous GC‐CPUE grids into four zones for each species life stage for the baseline scenario. The HSM zones for minimum flows were created using the same natural breaks that were calculated for baseline. Seasonal HSM maps were created from the zonal grids for each species life stage associated with baseline and with minimum flows (Rubec et al. 2018). Each HSM map had four habitat suitability zones: low, moderate, high, and optimum, representing predicted mean GC‐CPUEs increasing across the HSM zones.

Validation graphs

Gear‐standardized FIM data (observed GC‐CPUEs) for species life stages within each season were spatially joined to the zonal grid data using ArcGIS to create validation data sets. Each FIM data point was joined to the closest habitat point within 50 m. We validated each model by overlaying mean observed data onto predicted HSM zones to create validation graphs for each season (Rubec et al. 2018). Increasing trends in mean observed GC‐CPUE across HSM zones indicated spatial agreement between mean observed GC‐CPUEs and mean predicted GC‐CPUEs within HSM zones. Validation graphs with increasing mean observed GC‐CPUEs across all four suitability zones (low to optimum) were scored as 1.0. When the mean observed GC‐CPUEs exhibited increasing trends across three HSM zones (low to high suitability) instead of across all four zones, they were scored as 0.5.

Computation of zonal areas and population numbers

For each species life stage under baseline and minimum flows scenarios, tables were created that presented mean GC‐CPUEs (individuals/m²) and zonal areas (m²) for each seasonal HSM zone. The study area is comprised of 1,906,683 total 15‐ × 15‐m cells, with a total area of 429,003,675 m². Changes in percent zonal area (A) were calculated as the relative difference between percent areas for baseline and minimum flows:

Zonal population number estimates by season for each species life stage were derived by multiplying mean GC‐CPUEs (individuals/m²) by the areas (m²) associated with the HSM zones. Total population numbers in the study area were then estimated for each season by summing the zonal population estimates.

Confidence intervals

The R program GAMLSS computed CIs around Owens plots, around fitted splines, and associated with validation data sets, allowing assessment of uncertainty in the GC‐CPUE data. The population number estimates were not computed by the delta‐gamma GAMs used with the habitat suitability model. They were derived from the GC‐CPUEs from fitted splines applied to data sets containing the central coordinates of cells in the habitat grids. The averaged GC‐CPUEs associated with the cell coordinates were imported into ArcGIS to produce continuous GC‐CPUE grids. The continuous GC‐CPUE grids for each species life stage were then partitioned to produce seasonal HSM maps. Due to the averaging, the estimated population numbers derived from continuous GC‐CPUE grids do not have CIs.

RESULTS

Habitat Maps

Seasonal maps illustrate spatial patterns in salinity and temperature in the Charlotte Harbor estuary (Figures 5 and 6). The maps were similar for both the baseline condition and minimum flows. The percent salinity changes within zones, obtained by subtracting percent salinity for minimum flows from percent salinity for baseline, were mostly <1% for salinity ranges up to 20 psu and <4% for higher salinity ranges up to 35 psu (Table 3). The percent temperature changes within zones, obtained by subtracting percent temperature for minimum flows from percent temperature for baseline, were <0.5% for temperature ranges up to 30°C and <3% for higher temperature ranges up to 34°C (Rubec et al. 2018). Although the salinity maps and temperature maps changed between seasons, they appeared almost identical between the two scenarios within each season.

Statistical Table of Reduced Habitat Suitability Models

Final reduced habitat suitability models had the lowest AIC values and contained the best combinations of environmental variables (Table 4). Salinity was significant, based on high abundance (GC‐CPUEs) at low‐salinity ranges, for species life stages during most seasons on both the mu and nu sides of the models. Temperature was significant for some seasonal species life stages. However, there was no seasonal preference for temperature with most species life stages. Depth, dissolved oxygen, mud, and SAV were less significant for most species life stages, but these environmental variables were included in many of the final seasonal habitat suitability models.

Fitted Splines and Histograms from Habitat Suitability Models

Seasonal abundance from fitted splines (for continuous variables) and histograms (for categorical variables) illustrated marked preferences for salinity based on outputs from full delta‐gamma GAMs for all species life stages. In the reduced models, salinity was significant for most resident species life stages (Table 4). For example, JA‐Hogchoker during spring showed marked preferences for low salinities (0.5–5.0 psu) and for temperatures >30°C (Figure 7). The broader fitted splines for dissolved oxygen and depth suggested that these environmental factors were not as significant for JA‐Hogchoker. This was confirmed with the statistical output during spring for the reduced model. Plots of seasonal abundances showed that each of the species life history stages of interest exhibited strong affinities for salinity—the variable of greatest interest in comparing minimum flows to baseline conditions (Figure 8).

Juvenile and adult Hogchoker exhibited highest abundance at oligohaline salinity (0.5–5.0 psu) in the Upper P segment for all four seasons. Juvenile Sand Seatrout and JA‐blue crab were most abundant in the Upper P and Lower P segments. The abundance for J‐Sand Seatrout peaked near 7 psu during all four seasons. The abundance for JA‐blue crab peaked at 10 psu in fall and winter, at 8 psu in spring, and at <5 psu in summer. The abundance for EJ‐Southern Kingfish peaked at 15 psu in the fall and near 18 psu during winter, spring, and summer. The abundance‐by‐salinity relationships for A‐Bay Anchovy were similar during all four seasons, with peaks for the fitted GC‐CPUEs occurring near 18 psu.

The optimum zones derived from HSM analyses indicated that each species was most abundant at different salinity ranges proceeding downstream. The species order from lower to higher salinities was as follows: JA‐Hogchoker, J‐Sand Seatrout, JA‐blue crab, EJ‐Southern Kingfish, A‐Bay Anchovy, EJ‐Red Drum, and EJ‐Spot. The optimum salinity ranges for each species life stage when compared by season were similar for estuarine resident species, but as expected they differed for estuarine transient species as they progressed through their life history.

The estuarine transient species exhibited affinities for a broad range of moderate salinity as they first settled in the estuary; they then exhibited affinities for a narrow range of low salinity as early juveniles, followed later by affinities for moderate salinity. For example, EJ‐Red Drum abundance peaked near 18 psu in the fall, when they were first settling, and then peaked at <10 psu during winter and spring (Figure 8). There also were increasing relationships associated with the GC‐CPUE‐by‐salinity splines during winter and spring at salinities exceeding 30 psu. The fitted GC‐CPUE spline for EJ‐Red Drum in summer indicated that they were most abundant over a broad range of salinities <20 psu.

For EJ‐Spot during winter, the GC‐CPUE‐by‐salinity spline indicated that they were most abundant at salinities ranging from 5 to 25 psu (Figure 8). In spring, the abundance‐by‐salinity spline decreased at salinities ranging from 0.5 to 20.0 psu and then increased at salinities >30 psu. In summer, the abundance‐by‐salinity spline for EJ‐Spot declined at salinities ranging from 0.1 to 10.0 psu and then increased at salinities exceeding 20 psu. During fall, the abundance spline increased at salinities >30 psu.

Habitat Suitability Modeling Maps

Seasonal HSM maps illustrated spatial use by species life stages. In most cases for resident species, the predicted optimum zones occurred in low‐ or moderate‐salinity segments of the lower Peace River, lower Shell Creek, and lower Myakka River rather than in Charlotte Harbor. Seasonal baseline and minimum flows HSM maps were created, but only the baseline maps are presented in this paper because the minimum flows HSM maps are so similar. The report (Rubec et al. 2018) submitted to SWFWMD containing all of the seasonal HSM maps is available separately online as a Supplement to this article.

Most of the seasonal HSM maps for species life stages show an expansion of their spatial distributions in the lower Peace River during summer that is associated with higher freshwater inflows. In most cases, optimum zones of abundance expanded to match the expansion of low‐ and moderate‐salinity zones during the high‐inflow summer months (Figure 5).

Juvenile and adult Hogchoker

Seasonal HSM maps for JA‐Hogchoker associated with baseline showed optimum zones of abundance in the Upper P segment during all seasons (Figure 9). During fall, winter, and spring dry seasons, the optimum zones contracted within the upper part of the Upper P segment. For both baseline and minimum flows, the spatial extent of optimum zones expanded within the Upper P segment during the summer (Rubec et al. 2018). Close examination revealed that the optimum zone for minimum flows contracted slightly in association with water withdrawals during the summer.

Juvenile Sand Seatrout

Seasonal HSM maps for J‐Sand Seatrout were so similar between baseline and minimum flows that it was difficult to visually discern whether there was any effect of water withdrawals (see the Supplement). The HSM maps showed J‐Sand Seatrout occurring in both the Upper P and Lower P segments during fall. Small blue polygons representing the optimum zone were present in the Upper P segment during the winter. In spring, the optimum zones expanded, indicating that J‐Sand Seatrout became very abundant throughout the Upper P and Lower P segments. During summer, the optimum zone contracted, whereas the northern part of Charlotte Harbor had high abundance (green polygon). However, there was also a small optimum zone near the mouth of Charlotte Harbor during the summer.

Juvenile and adult blue crab

The optimum zone in the baseline HSM map for JA‐blue crab during winter was situated in the Upper P and Lower P segments of the lower Peace River and in the lower Myakka River (Figure 10). In spring, JA‐blue crab were abundant in the rivers, but the optimum zones diminished in area. In summer, the optimum zone expanded in the two rivers and into northern Charlotte Harbor. The optimum zone contracted in the fall and was present in the Lower P segment.

Early juvenile Southern Kingfish

The optimum zones in baseline and minimum flows HSM maps for EJ‐Southern Kingfish indicated that they were most abundant in the Lower P segment of the lower Peace River during all seasons (see the Supplement). The optimum zones expanded in the river during winter, with high abundance in the northern part of Charlotte Harbor. There was a contraction of the spatial extents of high and optimum zones in the maps during spring for both scenarios. During summer, the optimum and high zones expanded, possibly in relation to higher freshwater inflows. In summer, the optimum zone in the river for minimum flows was visibly smaller than the optimum zone associated with baseline.

Adult Bay Anchovy

The seasonal baseline HSM maps for A‐Bay Anchovy (Figure 11) were very similar to the seasonal HSM maps associated with minimum flows (Rubec et al. 2018). The optimum zones for both scenarios expanded during summer from the rivers into Charlotte Harbor, specifically in shallow‐water areas (<2 m) associated with increases in freshwater inflow.

Early juvenile Red Drum

Based on optimum zones in baseline and minimum flows HSM maps, EJ‐Red Drum during the fall were most abundant over SAV in segments A and B (Figure 1) of Charlotte Harbor and in the Lower P segment of the lower Peace River (Figure 12). In winter, the optimum zones shifted upriver into the Upper P segment. However, EJ‐Red Drum also were abundant in segment B near the mouth of the estuary. During spring, the optimum zones shifted downstream into the Lower P segment and into segments A and B of Charlotte Harbor in shallow water over SAV. In summer, the optimum zones for baseline and minimum flows indicated that EJ‐Red Drum were most abundant in the Lower P segment. However, zones of high abundance (shown in green) in the HSM maps indicated that EJ‐Red Drum also were prevalent in deeper water within segment A of northern Charlotte Harbor. During summer, there was no longer an optimum zone for EJ‐Red Drum in segment B near the mouth of the estuary.

Early juvenile Spot

The optimum zone in the baseline HSM map for EJ‐Spot in winter (Figure 13) indicated that they were abundant in the lower Peace River and in shallow areas within the northern part of Charlotte Harbor. During spring, the optimum zone situated in the Upper P segment may have been due to an affinity by EJ‐Spot for low salinity. Optimum zones for summer indicated that some EJ‐Spot still were present in the Upper P segment at low salinities, with the rest of the population present in segment B near the mouth of the estuary, where high salinities were found. In the fall, small optimum zones were located in segment B near the estuary mouth.

Validation

We validated the delineation of HSM zones for each species life stage by confirming that seasonal mean observed GC‐CPUEs increased across the four zones (Table 5). An example validation graph is presented for A‐Bay Anchovy in summer for minimum flows (Figure 14). Most species life stages received a validation score of 1.0. For JA‐Hogchoker, three of four seasons (spring, summer, and fall) had validation scores of 0.5 for both baseline and minimum flows. Lower mean observed GC‐CPUEs for these seasons were associated with the optimum zone in comparison with the high zone in the JA‐Hogchoker validation graphs.

Zonal Areas and Population Numbers

Zonal areas and population number estimates were derived from seasonal HSM maps for A‐Bay Anchovy during spring with both baseline and minimum flows (Table 6). Total spring population numbers estimated for A‐Bay Anchovy were 2,098,463,644 for baseline and 1,995,985,434 for minimum flows. Similar computations were performed seasonally to derive zonal areas and population number estimates for other species life stages. Seasonal population number estimates for species life stages are presented for each scenario (Table 7). Percentage changes between baseline and minimum flows indicated that total population numbers declined between 0.3% and 21.0%.

Although declines in total population numbers were <15% for most species, higher percent reductions were observed for JA‐Hogchoker in spring (19.0%), J‐Sand Seatrout in spring (21.0%), and EJ‐Southern Kingfish in summer (17.5%; Table 7). Total population numbers for A‐Bay Anchovy increased by 1.1% in summer and decreased during the other seasons. Total population numbers for EJ‐Spot decreased between baseline and minimum flows during winter (1.5%) and spring (1.7%) and increased during summer (4.8%) and fall (7.8%). Summer and fall were the seasons when EJ‐Spot left the estuary and estimated population numbers were low.

In most cases, there were higher seasonal percent population numbers of fish or crabs in the moderate and high HSM zones than in the optimum zones (Table 8). This was because the moderate and high zones had larger zonal areas than the optimum zones. Since population numbers were derived by multiplying mean GC‐CPUEs (individuals/m²) by zonal areas (m²), it was possible to obtain higher estimates for percent population numbers for the moderate and high zones, despite these zones having lower mean GC‐CPUEs than were estimated for the optimum zones.

DISCUSSION

This study, based on HSM analyses of FIM data, showed that the abundance‐by‐salinity relationships for resident species were similar across seasons for JA‐Hogchoker, J‐Sand Seatrout, JA‐blue crab, EJ‐Southern Kingfish, and A‐Bay Anchovy, indicating that each species life stage has a preferred salinity range that does not change much between seasons (Figure 8). This was previously verified by overlaying optimum zones from HSM grids onto salinity grids and extracting the salinity ranges by season (Rubec et al. 2019).

Short‐term studies of benthic macroinvertebrates, fish, and nektonic invertebrate communities in relation to physical habitat, salinity, and freshwater inflow in the lower Peace River were summarized by SWFWMD (2010). The SWFWMD has also sponsored surveys to relate distributions and relative abundances of phytoplankton, zooplankton, larval fish, and larval macroinvertebrate communities to salinity and water temperature changes associated with freshwater inflow and other habitat variables in the lower Peace River and lower Shell Creek (Peebles 2002a; Vargo et al. 2004; Atkins North America 2014; Janicki Environmental 2015). Regression analyses indicated that the life stages of various fish and invertebrate species displayed different responses to freshwater inflow in the lower Peace River (Greenwood et al. 2004; Peebles and Burghart 2013). Estuarine studies of responses to changes in freshwater inflows have been conducted in other regions of Florida (Browder and Moore 1981; Livingston 1997; Sklar and Browder 1998; Doering and Wan 2018); in other states, such as Texas (Longley 1994; Powell et al. 2002), California (Kimmerer et al. 2009; Weber‐Stover et al. 2016), and Georgia (Alber and Flory 2002); and in other countries (Drinkwater and Frank 1994; Estevez 2002; Gillson 2011; Adams 2014). These studies used a variety of methods for different purposes, thus preventing comparisons between them. What distinguishes the present study is the population number estimates derived from seasonal temperature and salinity grids predicted from hydrodynamic modeling of freshwater inflows by SWFWMD (Chen 2020).

The Peace River is not impounded and is free‐flowing along its length, whereas Shell Creek is impounded by a low‐head structure located 10 km upstream of its confluence with the Peace River. However, water is infrequently taken from storage in this small impoundment, so freshwater flows to lower Shell Creek largely follow the normal seasonal pattern for southwest Florida. Corrections to the flow records for both systems were made for existing withdrawals and other anthropogenic impacts to flow, so the baseline flow record examined in this study and the corresponding salinity distributions predicted from the hydrodynamic model reflect natural seasonal variations in the region. This allowed for the HSM‐GIS assessment of changes in fish habitat and species population abundance due to freshwater withdrawals that are representative of four seasons in southwest Florida, with the caveat that freshwater inflows during the winter were unusually low due to dry climatic conditions during the study period.

Peebles and Greenwood (2009) used spatial abundance quantiles to demonstrate impingement of EJ‐Hogchoker smaller than 31 mm SL on Shell Creek’s estuarine dam in association with a reduction of 3‐d mean freshwater inflow and the loss of low‐salinity habitat. The JA‐Hogchoker caught in seines also showed evidence of crowding below the dam. The loss of low‐salinity habitat (0.5–5.0 psu) was detrimental for EJ‐Hogchoker and JA‐Hogchoker.

Stevens et al. (2013) compared the abundance of Hogchoker (and other species not discussed here) within an oligohaline stretch of the lower Peace River during periods of varying freshwater inflow (wet and dry periods). The study used FIM data collected with a 21.3‐m circular seine during a wet period (April 1997–May 1999) and a dry period (July 2007–April 2010). Mean salinity in the oligohaline zone (<5 psu) during the wet period (1.7 psu) was less than that observed during the dry period (4.6 psu). Hogchoker had a significantly higher mean CPUE in the oligohaline zone during the wet period (37 fish/100 m²) than during the dry period (10 fish/100 m²). However, the spatial extent of the oligohaline zone was not determined from measurements of salinity. Its location was inferred based on shoreline vegetation usually associated with low salinity. Stevens et al. (2013) speculated that the decline in mean CPUE of Hogchoker during the dry period could have occurred because (1) Hogchoker moved upstream to remain within the oligohaline zone as the isohaline shifted upstream (outside of the FIM zone) or (2) production of Hogchoker decreased as a result of higher salinities in what was believed to be the oligohaline zone.

In the present study, reduced freshwater inflows due to water withdrawals from the lower Peace River and Shell Creek appeared to be detrimental for JA‐Hogchoker due to reduced areas of low‐salinity habitats (0.5–5.0 psu). Although the seasonal validation scores for most species life stages were 1.0, the scores for JA‐Hogchoker during spring, summer, and fall were 0.5 for both baseline and minimum flows (Table 5). Insufficient FIM samples in the Upper P segment, associated with two special studies, may account for the mean observed GC‐CPUEs within optimum zones being lower than those in the high HSM zones. Another potential explanation is that the predicted optimum zones in HSM maps for baseline and minimum flows in the Charlotte Harbor study area have smaller areas than the optimum zones predicted from the analyses of FIM data (Rubec et al. 2019). In that study, the validation scores for JA‐Hogchoker were 1.0 for all four seasons.

Juvenile Sand Seatrout were abundant in the lower Peace and lower Myakka rivers during spring and less abundant in Charlotte Harbor during summer, fall, and winter (Rubec et al. 2018). The summer HSM maps had a small optimum zone near the mouth of the estuary, which might indicate that J‐Sand Seatrout move out of Charlotte Harbor in summer and later spawn in the Gulf of Mexico (Cowan and Shaw 1988). Another possibility is that J‐Sand Seatrout grow to adult size in Charlotte Harbor during the summer and fall and then spawn in the lower part of the estuary during the following spring (Knapp and Purtlebaugh 2008). Sand Seatrout sought a reduced salinity range when they reached a length of 30–70 mm SL and then moved to higher salinities as they approached 100 mm SL (Purtlebaugh and Rogers 2007). Peebles (2002a) found postflexion larval Sand Seatrout and J‐Sand Seatrout in the lower Peace River and Shell Creek during spring. Some spawning evidently took place there, but most of the postlarvae and juveniles were believed to originate from higher‐salinity areas within Charlotte Harbor.

Optimum zones for JA‐blue crab during all seasons were in the tidal portions of the Peace and Myakka rivers (Figure 10). Literature reviewed by Gandy et al. (2011) indicated that immature females seek low‐salinity areas in estuaries (<15 psu), where they subsequently mate with mature males. When mating occurs in the spring and summer, the interval between mating and egg extrusion is about 2 months. When mating occurs in the fall and winter, spawning occurs during the following spring. Freshwater inflows were highest in the summer and lowest in the winter; thus, it is interesting that JA‐blue crab in our study were most abundant in expanded optimum zones during those seasons. The expansion of the optimum and high HSM zones in the lower Peace River during winter and summer appears to be partly related to peaks in recruitment of juvenile blue crab during winter and summer.

In their analysis of interannual trends in blue crab abundance, Flaherty and Guenther (2011) found that immature and adult blue crab indices of abundance and commercial landings were high in Tampa Bay during 1998 in association with increased rainfall during the 1997–1998 El Niño. This was followed by a reduction in the abundance of immature blue crabs in 2002 corresponding to lower‐than‐average river inflows from late 1998 to the beginning of 2002. Flaherty and Guenther (2011) noted that reduced rainfall and freshwater diversions had the potential to adversely affect recruitment and survival of young blue crab in the estuary. From 2003 to 2004, there was a dramatic increase in blue crab recruitment and a steady rise in adult abundance and commercial landings in Tampa Bay that appear to be linked to increases in freshwater inflows above historic means.

The distribution of EJ‐Southern Kingfish in the lower Peace and Myakka rivers during all seasons (Rubec et al. 2018) and in the northern part of Charlotte Harbor during winter and summer suggests expanded geographic ranges in optimum and high HSM zones during winter and summer, which may be related to spawning during fall and spring in the estuary. In comparison with baseline, a marked contraction of the optimum zone was apparent for minimum flows associated with water withdrawals during the summer.

Optimum zones for A‐Bay Anchovy demonstrated their prevalence in the lower Peace River during all seasons (Figure 11). Their optimum zones expanded into northern Charlotte Harbor in summer for both baseline and minimum flows and were apparently related to the expansion of low‐ to moderate‐salinity zones associated with increased freshwater inflows (Figure 5). Adult Bay Anchovy were the most populous species life stage, with their population numbers estimated to be five to six orders of magnitude greater than those of the other species life stages represented in the study (Table 7). Similar seasonal HSM maps and population number estimates for J‐Bay Anchovy were produced for both flow scenarios (Rubec et al. 2018; see also the Supplement) but are not presented in this paper.

These estimates appear reasonable for a species that is known to aggregate in great numbers. Bay Anchovy filter plankton from the water through their gills. Hence, one can assume that their population numbers in the lower Peace River are associated more with the flow‐related abundance of plankton in the river, which is tied to the influx of nutrients, than directly with salinity. Studies by Peebles et al. (1996, 2007) and Peebles (2002b) have demonstrated relationships between spatial distributions and abundance of A‐Bay Anchovy and larval prey, such as copepod larval abundance, which are tied to salinities behind frontal zones of river plumes in Tampa Bay.

In a community‐based study of the Charlotte Harbor system (Whaley et al. 2007), the frequency of occurrence of age‐0 Red Drum (15–100 mm SL) in the lower Peace River and lower Myakka River was highest during the fall, when this size range recruits to the estuary. In Tampa Bay, two size‐classes of age‐0 Red Drum (15–50 and 51–100 mm SL) were most abundant in the fall and winter, respectively, within the lower portions of three rivers adjacent to the bay, which collectively comprised only 2% of the study area but contained between 40% and 96% of the annual populations (Whaley et al. 2016). Freshwater inflows were positively related to spatial distributions and population abundance in the rivers, suggesting that reductions in inflow can reduce both habitat areas and population numbers. Due to their spatial distributions in the rivers, age‐0 Red Drum appear to be particularly vulnerable to modification of the riverine environment.

Peters and McMichael (1987) tracked the monthly growth of EJ‐Red Drum present in backwater areas and rivers adjoining Tampa Bay. Length frequencies indicated that Red Drum were <100 mm SL in September 1982, and they grew to lengths ranging from 200 to 300 mm SL by September 1983. With the present study in Charlotte Harbor, fitted splines indicated that EJ‐Red Drum (10–100 mm SL) were most abundant at salinities ranging from 10 to 20 psu in the fall and at salinities <10 psu during winter and spring (Figure 8). Based on optimum zones in HSM maps for baseline and minimum flows, EJ‐Red Drum that recruited during the fall were most abundant over SAV in northern Charlotte Harbor and in the Lower P segment of the lower Peace River (Figure 12). In winter, they moved upstream to the Upper P segment. During spring, they moved downstream into the Lower P segment and into segment A of Charlotte Harbor. In summer, the optimum zones showed that EJ‐Red Drum were most abundant in the Lower P segment, but the high HSM zones for summer indicated that they also were prevalent in deeper water within segment A. The optimum zones indicated that larger EJ‐Red Drum (101–299 mm SL), which probably recruited 1 year earlier, were present near the mouth of the estuary during fall, winter, and spring. Their absence near the mouth of Charlotte Harbor during the summer may be attributable to their growth beyond this size range (i.e., they reached the juvenile stage).

Seasonal HSM maps of Charlotte Harbor showed that EJ‐Spot were abundant in the lower Peace River during winter and spring (Figure 13). The fitted abundance splines by salinity and optimum zones in the HSM maps indicated that most EJ‐Spot left Charlotte Harbor in summer and fall starting at an age of about 6 months. The movement out of Charlotte Harbor starting in summer was unexpected, as peaks in larval abundance in the Gulf of Mexico suggest that Spot leave Louisiana estuaries to spawn in the Gulf of Mexico during fall and winter (Cowan and Shaw 1988).

The optimum zones in seasonal HSM maps for the resident species life stages indicated that they had high abundances within salinity ranges found in the lower Peace River. This could be due to habitat affinities for low or moderate salinities. However, it does not explain why the estuarine transients, EJ‐Red Drum and EJ‐Spot, were most abundant in the Upper P segment of the river during winter and spring, respectively. It seems likely that freshwater inflows to the river introduce nutrients that enhance production of phytoplankton, zooplankton, and larval fish that are exploited by early life stages of estuarine fish and invertebrates (Flannery et al. 2002; Peebles 2002a, 2002b, 2005). The relationships between species abundance and low or moderate salinities may be indicative of food availability created by nutrients and organic materials entering the lower Peace River in association with freshwater inflows from upstream (Rubec et al. 2019).

The population number estimates for species life stages averaged across years (2007–2014) in the present study were lower than those estimated across years (1996–2013) during the first phase of this study (Rubec et al. 2019). Since the same FIM data were used to create seasonal habitat suitability models for species life stages in both studies, the differences in population number estimates appear to be related to the habitat grids. The depth, bottom type, and dissolved oxygen grids used were the same in both studies. Hence, the lower population estimates in the present study must be related to the seasonal temperature grids and seasonal salinity grids derived from data produced by the hydrodynamic model and used for the baseline and minimum flow analyses. The most likely explanation is that the higher population estimates are related to generally higher rainfalls prior to 2007, which influenced the seasonal salinity grids created for the previous study (Rubec et al. 2019). Data derived from the SWFWMD database indicate that yearly rainfall totals in the Peace River watershed were higher than average during 1997, 1998, 2002, 2004, and 2005, whereas there were no years with above‐average rainfall between 2007 and 2014. The combined average flow for the Peace River and Shell Creek during 1996–2013 was 44.63 m³/s (1,576 ft³/s), while the average flow for 2007–2014 was considerably less at 31.32 m³/s (1,106 ft³/s).

Water managers have focused on volume, area, and shoreline length of low‐salinity zones (<2, <5, <10, and <15 psu) in the lower Peace River as their MFL management target because the low‐salinity zones there are most sensitive to change (SWFWMD 2020). The present study found that seasonal salinity maps for baseline in the Charlotte Harbor study area were very similar to those for minimum flows (Rubec et al. 2018). Seasonal temperature maps were also very similar for the two scenarios. Despite seasonal water withdrawals associated with minimum flows, the predicted seasonal salinity grids and seasonal temperature grids did not differ substantially between the flow scenarios. Since other environmental variables (depth, bottom type, and dissolved oxygen concentration) were kept constant, the similarities of salinity and temperature grids within each season largely explain why seasonal HSM maps associated with baseline and minimum flows are so similar within the Charlotte Harbor study area.

In the present study, there was an effect of seasonal water withdrawals on total population numbers associated with minimum flows in comparison to baseline (Table 7). The differences in percent population numbers for the two flow scenarios ranged between an increase of 7.8% for EJ‐Spot in fall to a decline of 21.0% for J‐Sand Seatrout in summer. Declines in percent population numbers for JA‐Hogchoker varied between 11.5% in fall and 19.0% in spring. The J‐Sand Seatrout and EJ‐Southern Kingfish losses were similar in magnitude to JA‐Hogchoker losses—over 10% in three of four seasons. By percentage, A‐Bay Anchovy population number losses associated with minimum flows appeared low, although by sheer numbers, their losses were huge.

Habitat suitability models linked to GIS provide meaningful quantitative comparisons to assist water managers with their decisions and help to explain the impacts of water withdrawals to policymakers and the interested public. We have illustrated a spatial HSM approach for predicting and comparing habitats and population numbers of selected species life stages of fish and crabs in the lower Peace River–Charlotte Harbor system under two water management scenarios, baseline and minimum flows. The approach was successful in quantifying small differences in seasonal salinity and temperature patterns (habitat areas) used to predict seasonal population numbers (biological resources) for species life stages; these small differences would not have been detectable otherwise. There were small reductions in habitat areas and population numbers associated with minimum flows in comparison to baseline across the study area, indicating that the percent‐of‐flow approach for regulating daily freshwater withdrawals was effective at preventing unacceptable impacts on habitat areas for the life stages of key species that would result from spatial shifts of salinity distributions in the estuary.

The HSM‐GIS approach, while providing estimates of habitat areas and population numbers associated with changes in water temperature and salinity distributions, does not account for other important mechanisms by which freshwater inflow can influence the abundance and distribution of fish and invertebrate populations in estuaries. Freshwater inflow delivers terrestrially derived organic matter and nutrients, and variations in the timing and volume of inflow can have pronounced effects on water quality, light penetration, primary production, trophic organization, and food webs in estuarine and coastal systems (Livingston et al. 1997; Darnaude 2005; Wissel and Fry 2005; Gillson 2011; Kim et al. 2014). Thus, using the HSM‐GIS approach in isolation without other analyses or modeling could result in erroneous conclusions and/or unintended consequences for particular species. Optimally, the HSM‐GIS approach should be accompanied by other assessments or modeling of the effects of freshwater inflow reductions on the physicochemical characteristics, trophic dynamics, and species interactions in an estuarine system.

Within a broader research and management strategy, the HSM‐GIS approach allows for the evaluation of how changes in salinity and temperature can affect the distribution and abundance of life stages of key species in an estuary. The availability of FIM data for Charlotte Harbor and its primary tributaries allowed for the determination of HSM zones ranging from low to optimum for the life stages of species known to be responsive to changes in freshwater inflow. When combined with salinity and temperature predictions from a hydrodynamic model, the HSM‐GIS approach allowed for the seasonal assessment of spatial changes in species life stages and population numbers for the Charlotte Harbor estuarine system across years (2007–2014). Using predictions from hydrodynamic modeling, the effects of reductions in freshwater inflow for specific peak flow events or prolonged low flows can be evaluated for any period within a year or across years for a selected duration of interest. Using these tools, potential impacts to fish habitats near and far from the location of the freshwater flow reductions and associated changes in population numbers of species life stages can be evaluated, allowing resource managers to focus their attention on zones of the estuary that are most vulnerable to change and providing sensitive indicators so that adverse impacts to the resources of the estuary can be avoided.

Acknowledgments

We thank Michael Flannery (formerly with SWFWMD), Joan Browder (National Marine Fisheries Service), and Philip Stevens and Richard Flamm (Fish and Wildlife Research Institute) for reviewing various drafts of the paper. We also appreciate Jennifer Shafer (Fish and Wildlife Research Institute) for her editorial review. Helpful improvements were provided by four anonymous reviewers. We are grateful for the financial assistance provided by SWFWMD (Purchase Order 15PW000010). There is no conflict of interest declared in this article.

REFERENCES