Review of the St. Johns River Water Supply Impact Study: Report 1 (2009)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

3 Hydrodynamic and Hydrologic Modeling Hydrodynamic and hydrologic modeling of the St. Johns River basin provides a critical foundation for the District’s assessment of the impacts of the proposed 262 million gallons per day (MGD) of water withdrawals. Extensive surface water and groundwater modeling was conducted as part of Phase I of the Water Supply Impact Study (WSIS) so that the ecological and biogeochemistry working groups could use the model output to assess potential impacts on biota in the region. This chapter reviews the hydrodynamic and hydrologic modeling efforts and provides recommendations for improvements for Phase II of the study. SURFACE WATER HYDRODYNAMICS AND HYDROLOGY In the WSIS Phase I assessment, the surface water hydrodynamic and hydrologic studies focused on understanding the changes in surface water depth, discharge, age, salinity, turbidity and wetland dewatering caused by surface water withdrawals from the St. Johns River. Hydrodynamic and hydrologic changes are the underlying drivers of ecological change resulting from surface water withdrawals; therefore, the hydrologic effects must be predicted accurately to understand the probable ecological effects. An overall project goal is to develop “quantitative response functions” that link river water withdrawal to hydrologic responses and ecological effects. The Phase II Methods report (SJRWMD, 2009c) provides an overview of the hydrodynamic and hydrologic modeling milestones and tasks that will be undertaken as part of the continuing study. Assessment of Phase I Surface Water Studies The Phase I surface water hydrodynamic and hydrologic studies (SJRWMD, 2008, Volume 1, Chapters 1 to 4) were screening studies based on readily available science and hydrologic models. Their objectives were twofold: (1) to understand which physical processes are affected by surface water withdrawals and (2) to develop the modeling infrastructure for Phase II studies by building on prior modeling efforts of the St. Johns River Water Management District (SJRWMD or “the District”). The surface water sections of the Phase I report provided an overview of the tools and capabilities that the District had in hand. Historical data were used in Chapter 1 to analyze water surface elevations in the middle St. Johns River (near DeLand). Chapter 2 provided an overview and analysis of hydrodynamic modeling of the lower St. Johns River. Chapter 3 provided information on several salinity scenarios conducted for the lower St. Johns River. Chapter 4 examined a model analysis of possible changes in middle St. Johns River sediment loading. The 27 

28 Review of the St. Johns River Water Supply Impact Study: Report 1 hydrodynamic models used were well-established three-dimensional models: the Environmental Fluid Dynamics Model (EFDC) in the lower St. Johns River and the Curvilinear Hydrodynamics 3-Dimensional (CH3D) model in the middle St. Johns River. Three hydrodynamic questions were addressed in Phase I: 1. Under what conditions is the water level at and downstream of DeLand, Florida, dominated by tidal flows rather than river discharge? (Chapter 1) 2. Can the pre-existing lower St. Johns River hydrodynamic model used for total maximum daily load (TMDL) studies be applied to upstream oceanic salinity modeling, and if so, what does it predict for upstream movement of brackish water under different dredging and surface-water withdrawal conditions? (Chapters 2 and 3) 3. Do water level changes in the middle St. Johns River impact sediment resuspension and hence turbidity? (Chapter 4) The Phase I report adequately addressed these questions for the purposes of a screening study. Specific suggestions on the analyses in support of these three questions follow. Chapter 1 examined historical hydrological data and behaviors, concluding that water surface elevation (stage) will be relatively unchanged in the lower and middle St. Johns River despite surface water withdrawals. The analysis showed that river stage is not a good indicator of either local hydrologic or ecological effects. The key observation is that the lower and middle reaches of the St. Johns River are tidally influenced under the critical low-flow regimes, such that reverse flows (i.e., upstream flows) will tend to maintain a relatively constant water level regardless of the downstream discharge. A clear stage-height vs. discharge relationship only occurs at higher flow rates when the proposed withdrawals would not be a significant fraction of the overall flow. This fact significantly simplifies hydrological analyses: the critical questions for withdrawals under low-flow conditions in the lower and middle St. Johns River are related to upstream transport of oceanic salinity (in the lower reaches) and changes in baseflow and water age (in the middle reaches). Hydrodynamic modeling of water levels in the middle and lower St. Johns River provided further support for these conclusions (as discussed below). It should be noted that the historical data and analysis methods in Chapter 1 are inadequate to demonstrate conclusively the potential effects of sea level rise on the middle St. Johns River, so future modeling will need to examine how sea level rise may affect upstream propagation of reverse- flow tidal effects. The possibility of climate-change induced sea level rise providing a nonlinear interaction with surface water withdrawals cannot be neglected, and the hydrodynamic models the District has should be adequate for this analysis. The Phase II methods report is not specific in its modeling scenarios, and so it is not clear whether or not sea level rise will be considered. Chapter 2 demonstrated that the EFDC hydrodynamic model could represent the tidally influenced surface water elevations in the lower St. Johns River. The report compared observed and simulated values, which are provided for the principal tidal component’s amplitude and phase (i.e., the M2 tide), hourly water levels, tidal discharge, and salinity. The committee agrees with the report that “the results presented [in Chapter 2] were adequate for the intended use of the model as a screening-level tool for the Phase I study.” The District used the tools available to rapidly get robust answers that would help narrow the scope of future studies. The committee also agrees that during Phase II the District should “further refine the model and incorporate additional verification, skill tests, sensitivity tests, and uncertainty analysis.” Tasks outlining calibration, verification and uncertainty analyses have been included in the Phase II Methods report; details of the methods used for these have not been provided for review. 

Hydrodynamic and Hydrologic Modeling 29 Chapter 3 described a limited study of some possible salinity effects associated with changes in sea level, wastewater diversion, surface water withdrawals, and channel dredging in the lower St. Johns River. The District’s analysis indicated minor increases (< 1 ppt) of average salt concentration due to withdrawals and higher increases due to channel deepening. This chapter provided large amounts of raw data and graphs but only limited text and analysis. The analysis focused on increases in average salinity and did not discuss extremes. Average salinity changes may not provide a firm foundation for quantifying possible ecological impacts. The need for extreme event data is discussed in the Phase II Methods report for the littoral zone working group (SJRWMD, 2009c, pg. 68, SAV Risk Model paragraph and 69, last paragraph); such data may also be important for understanding other environmental effects. The Phase II Hydrodynamic Methods report included a task for providing simulation results to the other six workgroups, but the details regarding the kind of data and how the groups will interact have not been provided. It will be important for the Phase II modeling and analyses to be tied carefully to the time-space scales of salinity that are important to ecology, and for the hydrodynamic modelers to have a documented process for determining scenarios and data needs for the other six working groups. Chapter 4 provided information and results from application of the CH3D model to the middle St. Johns River to study sediment transport and resuspension and water level changes (including linkage to the lower St. Johns River for the latter issue). The hydrological analyses from Chapter 1 implied that water level changes would be small due to the low gradient of the middle river and the lack of any coherent stage-discharge relationship at low flow rates. Results from the hydrodynamic modeling in Chapter 4 provided further support for this idea, indicating that water level changes of 1 to 4 cm could be expected for the proposed 155 MGD withdrawal. The maximum modeled change was 4.3 cm at the very most upstream section of the middle St. Johns River. Three scenarios for different distributions of withdrawals at five possible locations were used to examine water level changes using inflow data from 2005. Although this year was chosen for its data availability rather than as representative of an extreme year, the data set includes flow rates below 1800 MGD at all stations for two contiguous months; during this period the proposed withdrawal was 8 percent or more of the instream flow. Similar to Chapter 3, this chapter also included extensive graphs and data but limited analysis. Results presented in Chapter 4 demonstrated that the CH3D model can be calibrated to the middle St. Johns River, and that changes in suspended solids due to withdrawals in the middle St. Johns River should be negligible. Indeed, the findings in Chapter 1 that the water level in the middle St. Johns River is principally determined by sea level rather than discharge supports the conclusions of Chapter 4 that changes in sediment resuspension in the middle St. Johns River should not be significant. Recommendations for Phase II Surface Water Studies Based on the committee’s analysis of the Phase I report and a preliminary review of the Phase II work plan, the committee has several recommendations for improving the District’s Phase II surface water studies. Specifically, the District should (1) work to connect the separate modeling/analysis efforts of Phase I, (2) examine areas that were not studied in enough detail during Phase I, (3) document model calibration and sensitivity, and (4) develop methods to quantify model uncertainty on the time/space scales at which ecological effects may occur. These issues are described in more detail below. 

30 Review of the St. Johns River Water Supply Impact Study: Report 1 Connecting Hydrodynamic Models and Analyses The District needs to improve the integration of research questions and physical models across the principal regions of the St. Johns River. In Phase I, there was no linkage between the different physical models, and it was not clear how the results from the physical models would be qualitatively or quantitatively transferred to the ecological models. In particular, the Phase I surface water analyses did not examine how hydrologic changes in specific river sections relate to changes or effects in the rest of the river. Each change within the middle St. Johns River may have cascading influences into the lower St. Johns River and/or upper St. Johns River in hydrodynamics, hydrology, and ecology. For example, a reduction in the net downstream water flux may allow brackish water intrusion further up the lower St. Johns River. Increased upstream extent of reverse flows may change the balance between groundwater intrusion and wetland exchange with the surface water in the upper St. Johns River. An increase in water age may change water quality characteristics (e.g., nutrient concentrations) in the lower St. Johns River. Although the District no doubt understands these issues exist, the linkages need to be identified and studied more explicitly in the Phase II study. The Phase II work plan did not provide any significant detail on connections between hydrology and hydrodynamics. A critical limitation of the Phase I studies is that the hydrodynamics analyses were not informed by the time–space scales that are important to the ecological processes. Indeed, page 97 of the Phase II work plan indicates that information and data flow is only from the physical modeling to the ecology and does not have a defined feedback path. This one-way information flow may lead to the creation of problems after the hydrodynamic and hydrologic modeling work is essentially complete. As an example, hydrodynamic models can be used to produce detailed data and time-history statistics of salinity inundation; however, modelers often reduce data to simple daily averages and may not store all of the data necessary to later compute more detailed statistics. Thus, to study salinity stresses on submersed aquatic vegetation, the ecological scientists need to define the time-space characteristics of the salinity data that they would like to receive. Physical model development and analyses should be focused on the time-space scales important to ecology rather than the convenient time-space scales for hydrodynamics. This comment should not be construed as a recommendation to implement a monolithic hydrodynamic and ecological model from the estuary mouth to the uppermost reaches of the St. Johns River. However, the choice of model configurations and boundaries should be focused on representing the critical physical and ecological effects of surface water withdrawal rather than simply by the availability of pre-existing models and methods. To some extent, the Phase II work plan shows better integration between the hydrodynamics and hydrology workgroup and the ecology workgroups than the Phase I studies, although there is still limited communication across disciplinary boundaries. The conceptual framework discussed in Chapter 2 will help improve connections between the models and analyses. By developing a clear hierarchy from the conceptual framework through hypotheses and research questions that link the physics and ecology, modelers can develop a better understanding of the level of modeling necessary to answer the key ecological questions and develop the models accordingly. In particular, better linkages are needed between model extents, grid size, time steps, choice of dimensionality, analysis methods, and the time–space scales over which the physics drive ecological processes. 

Hydrodynamic and Hydrologic Modeling 31 Issues Requiring Further Study during Phase II Several important surface water hydrology issues were not studied or explained sufficiently in the Phase I report. The following section identifies specific issues that merit further exploration in the Phase II study. Modeling Hydrologic Extremes. The District should clarify how well the modeled flow conditions represent expected extremes, particularly drought. The District will need to examine thresholds beneath which water cannot be withdrawn, either due to negative consequences on baseflow and water levels in the upper St. Johns River or upstream propagation of oceanic salt water in the lower reaches. The fact that removal of 155 MGD may not significantly affect water levels in the middle St. Johns River cannot be used to assert that there will be no consequences. The District understands these ideas, but has not always presented them clearly. The St. Johns River system is different than most rivers in that an extended drought with reduced baseflow will not lead to significantly different water levels through the lower and middle reaches, but it will result in upstream propagation of oceanic salt water and/or changes in baseflow and water levels in the upper St Johns River. Thus, consequences of a long-term drought and practicality of water withdrawal in the lower and middle reaches should not be focused solely on water levels in those sections. Upper St. Johns River Hydrology. The Phase I studies did not include hydrodynamic or hydrologic models of the upper St. Johns River. Thus, in conjunction with improving the existing models, the District should develop a model to represent how surface water withdrawals affect the upstream wetlands and groundwater supply. The District needs to explore whether a pure hydrologic model (e.g., Hydrologic Simulation Program-Fortran [HSPF]) or a combination of a hydrologic model and a one-dimensional river model (e.g., Hydrologic Engineering Center- River Analysis System) or a two-dimensional floodplain inundation model is required to answer the key ecological questions for the upper St. Johns River. This would be determined best by close collaboration with the ecologists regarding the level of detail and certainty needed to understand possible impacts. Recognizing the above-mentioned deficiencies in the Phase I report, the District proposed in its Phase II work plan to build upon HSPF models previously developed for different purposes for “most of the watersheds.” The tasks outlined for this hydrologic modeling effort focus on rationalizing the disparate data sets used for prior models to develop land use, rainfall, and evaporation/transpiration data sets that are consistent across the models. These tasks appear to be reasonable, although the work plan gives no indication of how the concerns of the environmental science workgroups will be considered in developing scenarios and output data. Lower St. Johns River Stratification and Salinity Excursions. Vertical gradients of salinity (stratification) and horizontal upstream salinity excursions in the lower St. Johns River need to be more extensively studied during Phase II. There can be little doubt that sufficiently large withdrawals from the middle St. Johns River will allow greater upstream movement of higher salinity water in the lower St. Johns River, which may affect salinity-sensitive habitats. However, the model used in Phase I appears to have too few grid cells in the vertical direction to accurately represent the tendency of heavier saltier water to flow under lighter fresher water. This behavior produces a “salt wedge” that may propagate upstream against the river current based solely on density differences (i.e., a “baroclinic force”). Vertical stratification in the salt wedge reduces mixing of dissolved oxygen downward through the water column and may cause 

32 Review of the St. Johns River Water Supply Impact Study: Report 1 reduced oxygen levels (hypoxia) near the bottom. If a salt wedge develops, the horizontal extent of upstream salinity excursions along the bottom may be significantly greater than indicated by the existing EFDC model or by vertically averaged results. A technical difficulty in representing salt wedge behavior is that stable numerical models always have some positive numerical diffusion, which is cumulative and tends to smear the vertical salinity gradient over time. The reduced gradient leads to a smaller modeled baroclinic force, which reduces the modeled upstream excursion of salinity. To further complicate the physics, strong wind events may lead to rapid vertical mixing such that the interplay between the stratification, wind-driven turbulence, and current-induced shear requires very careful modeling. One of the difficulties of hydrodynamic modeling in a salt-wedge environment is validating a model. It is possible to get reasonable values for the average daily salinity while dramatically under-predicting the excursions, maximum salinities, and stratification that may be critical to ecological responses. To quantify how well its model predicts the dynamics of upstream salinity excursions and the development of salt wedge behavior, the District needs data on the evolution of the salinity field over sub-daily time scales. It is suggested that the District undertake several short-term field studies to monitor the spatial and temporal propagation of the salinity gradients in the lower St. Johns River over several tidal cycles. Such studies could be conducted with one or two small boats operating over 48 continuous hours using inexpensive conductivity-temperature-depth, dissolved oxygen profilers. Using these data, the District could examine their model performance with simplified domains to determine the necessary vertical grid resolution required for modeling salinity in the lower St. Johns River. Effects of Bridges. The lower St. Johns River model did not include the effects of bridges. In particular, the bridge for I-295 has a system of piers that blocks a substantial cross- section of the river. A visual estimate by a member of the committee was that between 10 and 20 percent of the river cross-section may be affected. Because of the shallow river surface gradient through this section, it seems unlikely that the bridge significantly alters water surface levels (at least not within the uncertainty of the model). However, the I-295 bridge (and others) may have significant effects on local velocities, turbulence, and mixing. It might be hypothesized that when the strong salinity gradients are undergoing tidal excursions in the vicinity of the bridge, there is much stronger mixing than when the salinity gradients have passed the bridge or are below it. Thus, there is the potential for significant amplification of the upstream salt wedge excursion if surface water withdrawals cause the salinity front to be pushed further upstream from the bridge (i.e., reducing mixing). Conversely, if the withdrawals lead to the salinity front spending more time around the bridge, then mixing may be increased, ameliorating some of the upstream salinity excursion. Modeling the Effects of Sea-Level Rise. Sea-level rise will be increasingly important as the century progresses, a fact that the District recognized in the Phase I report. Salinity scenarios corresponding to sea level increases of approximately 6, 10 and 20 cm at Mayport (near the mouth of the estuary) were used as input to the hydrodynamic model. The maximum increase in salinity predicted by the hydrodynamic model was considerably less than 0.5 ppt by 2033 (see Figure 27, Chapter 4,SJRWMD, 2008 draft). The committee believes that this is an excellent approach, but the District should extend this work to incorporate the recent higher-end estimates of sea level rise (e.g., Rahmstorf [2007] predicted rises of 0.4 meters by 2050 and 1.4 m by 2100). This would better quantify the long range impacts on salinity distributions in the lower St. Johns River with regard to future sea-level rise and provide managers with a more complete picture than using only the projections of Titus and Narayanan (1995) and the shorter time 

Hydrodynamic and Hydrologic Modeling 33 horizon of 2033. The worry is that if the 262 MGD are allocated by 2033, managers will have much less freeboard to keep the salt wedge downstream for the rest of the century. A more complete model would help quantify the uncertainty of what upstream influences sea-level rise could pose to future generations. Documenting Model Calibration, Validation and Sensitivity The usefulness of any model depends on model calibration and validation. “Calibration” requires adjusting model parameters to obtain a good fit between model results and observed data. “Validation” is the comparison of the model to observed data that were not used in the calibration to see how well the calibrated model predicts an independent data set1. For models that are entirely empirical (i.e., based only on observation), the calibration parameter adjustment process is irrelevant as long as the final state is acceptable. However, to evaluate deterministic models that are based on physical principles (such as the hydrodynamic model of the lower St. Johns River), documentation is needed on which parameters are adjusted and by how much. The need for these data can be illustrated by a simple example: some models have a parameter for vertical mixing efficiency that might be adjusted to 1.1 to provide an “improved” model output. Such a setting implies the mixing is “110% efficient” and therefore does more work than the energy available. Models that get the right answer for the wrong reasons are generally less useful (and less believable) than models that are wrong by some quantifiable amount but have physically reasonable parameterizations. Furthermore, models are applied within a framework of incomplete knowledge; for example, forcing conditions are not perfectly known, the model grid resolution is rarely ideal, and the model approximations for turbulence and hydrostatic pressure may affect the model output. The calibration process attempts to account for all of these uncertainties with the recognition that errors in the processes and model representations are interrelated. To provide transparency and confidence in modeling, the District should carefully document the methods and results of calibration. In particular, the following (at a minimum) should be identified. 1. Data used for calibration, 2. Data used for validation (these data should not be used in the calibration data set), 3. Types of data modified in calibration and the philosophy of their selection, 4. Range of modification for each variable changed, 5. Error measures of some calibrated variables in un-calibrated and calibrated simulations, 1 There is some confusion in the Phase II work plans regarding the use of the word “verification” for the hydrodynamic models. The hydrologic community stands in opposition to the majority of the modeling communities in that the word “verification” is often used in place of “validation.” In other modeling disciplines, the word “verification” is applied to the process of testing whether the model is bug-free (i.e., consistent with its desired computational equations) so that a verified model does exactly what it is supposed to do (even if it is wrong!). In contrast, a “validated” model is one that matches the observed data, even if it contains bugs and has not been “verified.” The District could make its case clearer by pointing out that the EFDC and HSPF models have been verified previously by developers, and their application to the St. Johns River is validated by comparison to data. 

34 Review of the St. Johns River Water Supply Impact Study: Report 1 6. Error measures of some non-calibrated variables in un-calibrated and calibrated simulations. The Phase II work plan includes tasks for model calibration, with some detail on the parameter estimation approach for calibrating the hydrological model. Calibration tasks for the hydrodynamic model are outlined, although there is insufficient information provided to determine whether or not the proposed calibration methods are sufficient. The governing equations that drive the physics in the model (and in the real world) may have varying sensitivity and robustness to different forms of error. Model sensitivity analyses need to be conducted with a particular focus on the physics that are critical to each study rather than the readily measured values. For example, sensitivity analyses for the vertical grid resolution cannot be based solely on water surface level because it is possible for a calibrated model to create an excellent and robust model of the water surface elevations that is insensitive to the choice of the vertical grid resolution, whereas modeling the upstream motion of salinity may be extremely sensitive to this parameter. The District should demonstrate the sensitivity of each model to various choices in the model setup. The key question is whether or not the model resolution is sufficient for the questions to be answered in light of the variability of the system. The committee recognizes that it may not be possible for a completely converged model to be applied (i.e., a model that is entirely insensitive to further refinements in time step or grid scale). Such unavoidable sensitivity to model grid scale and time step should be considered within the quantification of uncertainty (discussed below). Sensitivity tests are mentioned in the Phase II work plan as a part of the Task 4 Initial calibration (page 97), but the reference provided is circular (i.e., refers to itself for detail). Quantification of Uncertainty Uncertainty in surface water modeling will propagate through associated ecological analyses, adding to the range of possible outcomes and risks. It follows that Phase II should have a strong focus on understanding the underlying sensitivity and uncertainty in the hydrodynamic and hydrologic modeling studies. Two key issues contribute to model uncertainty: the sensitivity of the model setup and unknown future forcing. Both issues need to be estimated quantifiably so that the ecological analyses can be conducted in light of a range of possible future conditions. In quantifying the uncertainty associated with model sensitivity, the District will need to examine the most recent literature. This area is still relatively new for 3D modeling, and so there may not be clearly applicable guidelines. One aspect the District might consider is the difference between calibrated and uncalibrated model results, which may indicate the maximum range of uncertainty in the model behavior. Another possible avenue is to examine the variance associated with a collection of simulations using different (but reasonable) calibrations. When quantifying the uncertainty associated with unknown future forcing, it should be recognized that the deterministic hydrodynamic and hydrologic models provide a single answer for a given set of forcing conditions. Since the actual future forcing conditions are entirely unknown, they must be synthesized in some coherent approach. It is possible to use historical data to develop a stochastic understanding of the past for generating possible future “likely” scenarios, but such efforts should be tempered by the realization that we do not have a stationary climate system. 

Hydrodynamic and Hydrologic Modeling 35 To provide a sound basis for analyzing potential ecological effects, the hydrodynamic and hydrologic models should be used over a variety of scenarios as a means of quantifying uncertainty. In the past, the computational expense of 3D models made such efforts impractical. However, the latest generation of 8-processor workstations allows eight models to be run simultaneously with a relatively modest (<$5K) investment. Thus a bank of five computers can run 40 simulations at a time and rapidly develop a data set large enough to quantify the uncertainty associated with the different possible future conditions. To make this analysis tractable, the District should develop a rigorous approach that generates a variety of possible future scenarios and evaluates their likelihood, along with methods to analyze results from multiple model scenarios in an uncertainty framework. The Phase II work plan includes an Uncertainty Analysis Plan, apparently to be conducted by INTERA Geosciences & Engineering, but no details were provided. Summary of Surface Water Hydrodynamics and Hydrology The District is progressing along the correct track, but critical details discussed above either have not been considered or have not been sufficiently documented. The committee has three specific recommendations. First, the District should connect the separate modeling and analysis efforts of Phase I. For example, the Phase I analyses did not examine how hydrologic changes in specific river sections relate to changes or effects in the rest of the river. Second, several areas were not studied or explained in enough detail during Phase I, including modeling of extreme conditions, the hydrology of the upper St. John River, vertical gradients of salinity and horizontal upstream salinity excursions in the lower river, and the effects of bridges and sea level rise. Third, the District should document model calibration and sensitivity and develop methods to quantify model uncertainty on the time/space scales at which ecological effects occur. GROUNDWATER HYDROLOGY A primary goal of the Phase I groundwater modeling was to predict whether discharges of groundwater into the St. Johns River would change if water withdrawals caused the river stage drop. Two groundwater flow models were used during Phase I to compute groundwater base flows along the river from the surficial aquifer system (SAS) and the upper Floridan aquifer (UFA): the North Central Florida (NCF; Motz, and Dogan, 2004) and the East Central Florida (ECF; McGurk and Presley, 2002) MODFLOW (Harbaugh et al 2000) models (see Figure 3-1; Table 3-1). These steady-state models are not density-dependent groundwater flow models, such that solute transport is not addressed. The groundwater analyses were confined to the middle and upper St. Johns River basins (see Figure 1-1); for the purposes of this analysis, the District assumed that there is no groundwater discharge contribution to the lower St. Johns River basin. 

36 Review of the St. Johns River Water Supply Impact Study: Report 1 St John's River Basin North Central Florida (NCF) Inactive East Central Florida (ECF) East Central Florida Transient (ECFT) Inactive Lake Okeechobee FIGURE 3-1: SJRWMD Groundwater Models used in the WSIS. SOURCE: SJRWMD. TABLE 3-1: Groundwater Models in the WSIS Area Uniform Row, Square Flow Model Name Model Area (Status) Column Grid Size Type Aquifers Authors East Central Mainly the upper St. Johns 174, 2,500 ft on Steady SAS, McGurk Florida (ECF) River Basin 194 a side. State UFA, & & Presley (Used in Phase I CIA) LFA (2002) North Central Mainly the middle St. Johns 150, 2,500 ft on Steady SAS, Motz et Ground Water River Basin 168 a side. State UFA, & al. (2004) Model (NCF) (Used in Phase I CIA) LFA East Central Whole upper St. Johns River 472, 1,250 ft on Transient SAS, Butler et Florida Transient and Kissimmee Basins 388 a side UFA, & al. (2009) Model (ECFT) (Not Used in Phase I CIA) LFA Note: SAS = surficial aquifer system, UFA = upper Floridan aquifer, and LFA = lower Floridan aquifer. 

Hydrodynamic and Hydrologic Modeling 37 Assessment of the Phase I Efforts Based on the modeling results, the change in average discharge of groundwater following water withdrawal was not predicted to be significant in the middle and upper basins. More specifically, a 4-cm decrease in water level would increase the groundwater base flux by 1.1 percent (or a similarly small volume); it would increase the chloride flux by 13 million tons/day or 1.2 percent (SJRWMD, 2008). To further demonstrate that the groundwater discharge variability is not significant along the river, the District conducted a modeling analysis of two extreme cases. In the first case, the model results showed that lowering the water stage drastically (by 30.5 cm) would nearly double the groundwater discharge (which implies that a lowering of 4 cm stage would only change groundwater discharge by a small percentage, assuming a linear relationship). In the second case, the model predicted that if the current groundwater pumping in the St. Johns River basin (about 700 MGD) was stopped, the groundwater base flow to the river would increase to about 30 MGD, which is 5 percent of what is pumped (suggesting that discharge is not particularly sensitive to groundwater withdrawals). (The model results in the second case did not consider the direct surface water runoff that would increase as water moves from wetlands and other areas toward the river.) From these analyses, the District appears to have concluded that groundwater discharges to the river are relatively insensitive to the current proposed changes in river stage. This conclusion may be correct, but it is not yet technically defensible, for the reasons discussed below. Limitations of the Current Groundwater Modeling Several limitations in the current groundwater models may affect the accuracy of the output. First, the models are steady state models, they do not model density-dependent groundwater flow, and wetland hydraulics is not represented in any of the groundwater models. These facts are not surprising, given that it would be extremely difficult to develop a rigorous regional model within the timetable of the WSIS. To elaborate, the Floridan aquifer system of northeast Florida, mainly the middle St. Johns River basin (see Figure 3-1), was simulated using a steady state regional numerical model (the NCF). This revised groundwater flow model (Motz and Dogan, 2004) includes a change with respect to previous versions of the model in its lateral boundary conditions and recharge and evapotranspiration calculations. The newest design of the NCF includes all groundwater withdrawals for the 1995 conditions in the surficial aquifer system and upper Floridan aquifer. Vertically, the model consists of three aquifer units, separated by confining units. For both the NCF and the ECF, model calibration was based mainly on piezometric heads and spring flows. However, a rough calibration of the base flow using the average mass-balance also was conducted on a larger scale for the steady-state models. The models are not capable of computing changes in groundwater base flow resulting from the variability in rainfall conditions, nor are they capable of predicting saltwater intrusion into the river because variable-density flow cannot be accounted for. The ECF model (McGurk and Presley, 2002) was developed mainly for the upper St. Johns River basin to simulate the freshwater portions of the surficial and Floridan aquifer systems and to assess the potential impacts of future withdrawals to the Floridan aquifer (see Figure 3-1). The District calibrated the ECF model to average 1995 conditions by comparing the 

38 Review of the St. Johns River Water Supply Impact Study: Report 1 simulated groundwater levels to the surficial aquifer and the Floridan aquifer systems’ historical groundwater levels, and by comparing simulated spring flow conditions to the actual ones from 1995. McGurk and Presley (2002) recommended the following to further improve model results: calibrate the model to transient conditions, gather and use more information for recharge and evapotranspiration into the surficial aquifer system, and evaluate saline water intrusion. The District has not acted on these recommendations yet, which would help create a more defensible model and bolster their conclusions about the potential effect of water withdrawal on groundwater levels. Finally, the District’s 20-year groundwater projections do not include climate change factors, which would be a difficult and challenging difficult task. Sea level change is addressed in the Phase I report only in terms of the projected increase in salinity of the lower river. Adequacy of Groundwater Inputs into Surface Water Models The above limitations have direct consequences for the District’s hydrodynamic modeling because the groundwater model results (i.e., groundwater base flow from the aquifer) are used as input into the hydrodynamic models. Because the groundwater models are steady state, the salinity sources into the river are considered to be constant over time. However, rising seawater levels, intense droughts, and/or increasing withdrawals from the river could eventually change the contribution of groundwater to the river’s overall salinity. Data show that there are significant amounts of chloride in the groundwater that may make their way into the river (see Figure 3-2). The salinity in the river, specifically in Segment-7 (Figure 3-3), is dependent on geologically old saline water (around 5000 ppm) stored in the upper Floridan aquifer that has not been completely flushed out by meteoric water. (Indeed, the last opportunity for the aquifer system to have drained the saline water was approximately 18,000 years ago when the sea level was dramatically lower [by 450 ft.] than today.) The degree of uncertainty about salinity coming from this source was not evaluated in Phase I of the WSIS. Recommendations for Phase II Groundwater Studies Based on the committee’s analysis of the Phase I report and a preliminary review of the Phase II work plan, the committee has several recommendations for improving the District’s Phase II groundwater study. Improvements to the groundwater models during Phase II of the WSIS should include incorporating transient groundwater flow, two-dimensional density dependent flow, and the interactions between the wetlands and the aquifer system. These issues are described in more detail below. Transient Groundwater Flow A regional-scale, transient groundwater model covering east-central Florida, called the East-Central Florida Transient (ECFT) model, is in the final stages of development (Butler et al., 2009). ECFT expands upon the ECF model to include the Kissimmee Basin (see Figure 3-4). 

Hydrodynamic and Hydrologic Modeling 39 30 15 0 30 Miles St John's River Basin UFA Chloride Concentration 0-50 50-250 250-1000 1000-15000 ´ Lake Okeechobee FIGURE 3-2: Generalized Salinity Map, Layer 2, ECFT Model. SOURCE: SJRWMD Staff. 

40 Review of the St. Johns River Water Supply Impact Study: Report 1 FIGURE 3-3: St. Johns River Segments. SOURCE: SJRWD. The St. Johns River, Southwest Florida, and South Florida Water Management Districts are all in the final phase of calibrating the model under transient conditions. Following calibration, verification, and incorporation of peer reviewed comments, the District plans to use the model to evaluate future impacts on the hydrologic system, including wetlands, lakes, and springs. Use of this model will be an important step in the right direction for the Phase II WSIS. ECFT includes a Wetlands package, developed by the South Florida Water Management District and the Center for Hydrology and Water Resources at Florida Atlantic University (Restrepo et al., 1998). The current version of the Wetlands package, which has been applied to subregional models in south Florida, is incorporated into MODFLOW and enables the top layer of the grid system to contain overland or groundwater flow. Such an approach allows the flow equations to remain valid when the water surface falls below the soil surface. Furthermore, the package can account for vegetation characteristics, sheet flow, sloughs, levees, barriers to 

Hydrodynamic and Hydrologic Modeling 41 overland and groundwater flow, and evapotranspiration. Thus, it is suitable for modeling the wetting and drying of wetlands. The ECFT grid spacing (see Table 3-1) balances the need for resolution of surface water features and local impacts with data availability. The temporal discretization of the model (the process of transferring continuous models and equations into discrete counterparts over a specific time step) was chosen to accurately reflect the hydrologic system changes (e.g., rainfall and canal stages) over a recent wet and dry cycle. This model should be better able to predict wetlands behavior and interactions with the river, the aquifer, and the surface water storage. It will be critical to simulate the wetlands behavior in the specific locations where the water withdrawals may be occurring. ECFT should be a valuable screening tool because of its transient qualities, including the representation of various seasonal changes, wetland simulation capabilities, and increased horizontal resolution. 10 50 100 150 200 250 300 350 Marion 10 Volusia 50 Citrus 100 Lake Seminole Sumter 150 Hernando Orange 200 Pasco AT OC LA EAN 250 NT IC Brevard 300 Osceola Polk Hillsborough 350 400 C-3 8 Indian River 450 Manatee Hardee Highlands Okeechobee St. Lucie Every 10th Model Row Every 10th Model Column Kissimmee Region Inactive Model Area (Layer 1) 0 5 10 Study Area Miles Prepared by : MAPU Date: 9/14/06 Map Doc.: ModelMesh10x10.mxd FIGURE 3-4 Model Study Area and Grid for ECFT. SOURCE: SJRWMD. 

42 Review of the St. Johns River Water Supply Impact Study: Report 1 Density-Dependent Groundwater Flow The ECFT is not a density-dependent model, which limits its ability to predict changes in salinity that might result from water withdrawals. A density-dependent groundwater flow and transport model remains an important need for the St. Johns River basin. It would be extremely challenging to develop a three-dimensional model within the timeframe provided for the WSIS, but as an alternative the District should consider developing a groundwater flow and transport model for a cross-section of the basin, preferably in river segment 7 (see Figure 3-3) because that is the most critical location for potential changes in salt water intrusion. Seepage from the aquifer to the river in segment 7 is significant because the confining unit between the bottom of the river and the upper Floridan is thin in that zone, and the head in the upper Floridan is higher than the river stage. The gradient in the upper Floridan decreases significantly toward the ocean. Such a cross-sectional model could indicate the vulnerability of river segment 7 to changes in saline flow from groundwater following potential increases in water withdrawals from the river or other alternative water management scenarios. Such an effort is worthwhile and feasible because the problem of saline intrusion is density-dependent, the ideal tools to analyze the problem do not yet exist, and a two-dimensional cross-sectional model should not take long to develop. Indeed, there is no groundwater pumping from the upper Floridan aquifer in river segment 7 (because of high salinity), which simplifies the development of a cross- sectional model. Cross-sectional models are well suited for performing sensitivity analysis and can simulate groundwater flow patterns at a refined level of spatial resolution near the river locations of interest. Modeling Wetland-Aquifer-River Interactions As mentioned in Chapter 2, when water is taken out of the river, several hydrologic results can occur, including reduced flow downstream, increased base flow to the river due to a steeper gradient between the aquifer and the river, and increased drainage from upstream wetland areas. To analyze these occurrences, it is necessary to use an integrated surface water– groundwater flow model that is able to represent wetland and aquifer interactions. The current models (NCF and ECF) are not able to quantify the flow within wetlands and their change in storage. The Wetland package of ECFT is able to simulate these interactions sufficiently, although the committee has not yet seen the details regarding calibration of the ECFT model. Summary of Groundwater Hydrology In order to improve the groundwater modeling in Phase II of the WSIS, the District should consider using the ECFT transient model, which includes wetlands processes, and a cross-sectional density-dependent model. These models will be critical to understanding how groundwater flow and salinity flux into the river will vary with water withdrawals. The more sophisticated analyses that would derive from these models are needed to support the District’s assertion that water withdrawals will not significantly alter the groundwater discharge contribution to the St. Johns River basin.