Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 19
Chapter 2 Hydrology and Hydrodynamics The goal of the Water Supply Impact Study (WSIS) was to assess potential impacts from withdrawing freshwater from the St. Johns River. It follows that the withdrawal’s influence on hydrology and hydrodynamics sets the stage for analyzing and understanding possible ecological effects. The District used three different types of models for analyzing water flow: surface hydrology, hydrodynamic, and groundwater. The surface water hydrology modeling used the well-known program HSPF and was reviewed in NRC (2010). This work served both to analyze flows through the landscape in the Upper St. Johns River and to provide the inflow boundary conditions for the hydrodynamic model of the Lower St. Johns River (LSJR) and Middle St. Johns River (MSJR). The hydrodynamic modeling used a well-established three-dimensional model, the Environmental Fluid Dynamics Code (EFDC). Hydrodynamic model development and calibration were discussed in NRC (2010); results from the modeling efforts were described in detail in Sucsy et al. (2010, 2011) and are further analyzed in this chapter. Steady-state groundwater flow models based on the U.S. Geological Survey’s MODFLOW were used to compute groundwater base flows along the river from the surficial aquifer system and the upper Floridan aquifer. The purposes of the groundwater modeling were to estimate the response of the underlying aquifer to potential water withdrawals in terms of discharge and aquifer head change, to provide boundary conditions for the mainstem hydrodynamic model, and to provide groundwater data for water budget calculations. Because most of the hydrologic and hydrodynamic modeling efforts were reviewed extensively in NRC (2010), the main conclusions and recommendations from that report are discussed only briefly here. Rather, this chapter deals primarily with work done subsequent to the publication of NRC (2010). WITHDRAWAL SCENARIOS Hydrodynamic modeling was used to examine the possible effects of a water withdrawal on water surface level, water age, and salinity along the LSJR and MSJR. These three features are the key physical1 changes that will occur as a result of withdrawals. The water withdrawals, if approved, would likely occur in stages over a period of years to decades, and thus it was necessary for SJRWMD to assess effects of other forcing conditions2, specifically (1) changes in 1 We use physical in the most restrictive sense, to exclude chemical changes. 2 In hydraulic/hydrologic modeling terminology, water withdrawals, tides, rainfall, and landscape characteristics are input forcing conditions that alter the model output. The model does not predict the forcing conditions; instead, it predicts the system behavior as a result of the forcing conditions. 19

OCR for page 19
20 Review of the St. Johns River Water Supply Impact Study: Final Report landscape runoff as predicted by the hydrologic model, (2) impacts from the proposed Upper St. Johns River (USJR) projects, and (3) predicted sea level rise (SLR). The hydrodynamic model was run through a comprehensive set of scenarios to evaluate model sensitivity to different forcing functions. The scenario nomenclature follows a keyword pattern of: Extent of withdrawal land-use year Upper basin project status sea-level rise where the values for the keywords are: withdrawal = {Base, Half, Full, FwOR} land-use = {1995, 2030} project = {N, P} sea-level rise = {N, S, H} The withdrawal conditions are defined as: Base: Zero surface water withdrawal (existing condition). Half: 50% of the proposed maximum withdrawal rate (77.5 MGD) from the SJR. Full: 100% of the proposed maximum withdrawal rate (155 MGD) from the SJR. FwOR: Full (155 MGD) SJR withdrawal plus 107 MGD from the Ocklawaha River for a combined 262 MGD withdrawal The land-use conditions are: 1995: Historic land-use patterns from 1995 data. 2030: Forecast land use patterns for year 2030. The projects are: N: No projects: hydrologic effects of USJR projects are neglected. P: Completed projects: all hydrologic effects of USJR projects are included. The sea-level rise conditions are: N: No sea-level rise (i.e., ignore historic trend and use 1995 data). S: Sea-level rise for 2030 based on historic trend at Mayport (14 cm). H: Higher estimate of sea-level rise at 2030 (28 cm). Thus, the scenario name Half1995PN indicates the “Half” withdrawal of 77.5 MGD with historic 1995 land-use, the completed USJR projects, and zero sea-level rise. Table 2-1 shows some of the most commonly used scenarios (and thus represents only a subset of the scenarios used and referred to throughout this report).

OCR for page 19
Hydrology and Hydrodynamics 21 TABLE 2-1 More frequently used hydrodynamic model scenarios for the WSIS. Base1995NN (green highlight) is the hindcast scenario of the historic baseline. Upper Withdrawal Sea Level LandUse Basin (mgd) Rise (cm) Projects Base1995NN 0 historic none 0 Half1995NN 77.5 historic none 0 Full1995NN 155 historic none 0 Base1995PN 0 historic completed 0 Half1995PN 77.5 historic completed 0 Full1995PN 155 historic completed 0 Base1995PS 0 historic completed 14 Full1995PS 155 historic completed 14 Base2030PN 0 forecast completed 0 Half2030PN 77.5 forecast completed 0 Full2030PN 155 forecast completed 0 FwOR2030PS 262 forecast completed 14 At first glance, some of the scenario choices in Table 2-1 may seem strange because they do not correspond to any reasonably expected condition. For example, Base1995PS combines 1995 land-use conditions with 2030 forecast sea-level rise and the upstream projects installed in the present decade, and this combination does not reflect any reasonable historic or forecast condition of the river. Indeed, the purpose of most of these scenarios was neither to forecast nor hindcast3 a possible river condition but to isolate effects and evaluate model sensitivity for different forcing conditions. Thus, by comparing Base1995PS to Base2030PS the modelers could evaluate how land-use changes (1995 to 2030) affect the system for the same sea-level and project conditions. Further comparison of these ‘S’ sets with the similar ‘N’ sets of Base1995PN and Base2030PN allowed the modelers to compare how land-use changes interact with sea-level rise. The hydrology and hydrodynamics (H&H) workgroup analyzed three scenarios that they considered to represent reasonable “near-term” possibilities: Base1995PN, Half1995PN, and Full1995PN, and four scenarios considered to be reasonable long-term possibilities: Base2030PS, Half2030PS, Full2030PS, and FwOR2030PS. The short- and long-term scenarios thus differed in terms of land use conditions and salinity effects from sea level rise (and also inclusion of withdrawals from the Ocklawaha River). The short-term scenarios were considered to be plausible conditions for approximately the next decade, whereas the long-term scenarios were considered to be plausible for later decades. The outputs from the above seven scenarios, coupled with the Base1995NN and Full1995NN scenarios, were the primary hydrologic and hydrodynamic information provided to the ecological workgroups for their analyses. Note that the environmental workgroups considered different scenarios to be the “extreme,” depending on river segment and the type of ecological components being considered; details on these choices can be found in the individual workgroup reports. Several additional scenarios also were run to gain insight into possible future events that might change the predicted outcomes. These scenarios were not intended for use by the ecological workgroups because of the high level of uncertainty associated with whether or not these conditions might occur. The additional nomenclature for these scenarios is: 3 Hindcasting is a term used by modelers to describe models of past system behavior. Such models are used to understand the system, to improve the model through calibration, and to estimate model uncertainty.

OCR for page 19
22 Review of the St. Johns River Water Supply Impact Study: Final Report CHND: Channel deepening by the U.S. Navy in the lower SRJ WWTP: Reuse (diversion) of outflow from wastewater treatment plants It should be noted that not all the scenarios that were run were analyzed in the H & H report. GROUNDWATER MODELS Modeling of groundwater flow was used in the WSIS to examine the impact that surface water withdrawals may have on groundwater discharge to the river due to a declining river stage. The models were used mainly in the middle SJR, where surface waters are influenced by groundwater, and to set up groundwater discharge and salinity boundary conditions for the hydrodynamic model. An important issue regarding groundwater discharge to the St. Johns River is its role in adding chloride to the river from saline portions of the regional aquifer, i.e., the Upper Floridan Aquifer (UFA). Because the groundwater flow models used by the District do not simulate transient flows or chloride transport, the Committee recommended in NRC (2009) that the District determine whether the assumptions inherent in applying the models for these purposes were valid. The District conducted various analyses and concluded that their use of the steady-state density-independent groundwater models was appropriate (see Sucsy et al., 2011, Volume 2, Chapter 5 for details of these analyses as well as more comprehensive results for the groundwater modeling). The sections below briefly revisit these issues. Analysis of Methods to Estimate Chloride Loads The District evaluated the validity of three simplifying assumptions needed to compute chloride loadings from groundwater discharge from the UFA to the St. Johns River using their models: (1) temporally constant chloride concentrations, (2) constant chloride–salinity relationships for diffuse groundwater flow, and (3) constant density of groundwater for vertical flow calculations. Temporally constant chloride concentrations. According to Sucsy et al. (2011), chloride concentrations in groundwater of the UFA vary widely across the study area, but the temporal variations in chloride concentrations at a given location are small. The stability of chloride concentrations at a given site means that estimates of chloride loads to the river can be simplified by matching observed chloride concentrations at a given site in the aquifer with estimates of groundwater discharge. The temporal stability of chloride within observation wells was demonstrated using four analyses: calculation of relative standard error, visualization of chloride time-series, visualization of ranked chloride observations, and comparison of chloride with rainfall indicators. Constant chloride–salinity relationships for diffuse groundwater flow discharge. Even if chloride concentrations at a given site do not change significantly over time, the proportion of chloride relative to other major ions could change. This question is relevant when using chloride concentrations to estimate salinity levels for use in the hydrodynamic model. The ratio of eight major ions found in water was used by the District to evaluate the stability of

OCR for page 19
Hydrology and Hydrodynamics 23 salt composition in UFA wells. Three analyses were used: normalization of ionic concentrations, USGS chemical classification, and Maucha diagrams. Results indicate that salt composition, like chloride concentration, is stable in groundwater discharging into the MSJR, which allows accurate conversion of chloride concentrations to salinity. Previous studies on groundwater in the region have shown that trapped Pleistocene seawater (often referred to as “relict seawater,” RSW) is the main source of chloride in the UFA. Although chloride varies widely in observation wells across the study area, a common source (RSW) helps to predict that the chloride–salinity relationship is spatially constant. Constant density of groundwater for vertical flow calculations. Vertical density gradients were found to have only a minor effect on calculated vertical groundwater discharge to the river, because the gradient between the UFA hydraulic head and river stage is relatively large. The District’s analyses of groundwater discharge indicated that density differences between brackish groundwater and fresh river water do not appreciably increase groundwater discharge to the river. Based on the above findings, the District concluded that chloride loads could be calculated as the simple product of simulated groundwater discharge and observed chloride concentrations, and that density differences resulting from spatial chloride variability are unimportant in the simulation of vertical groundwater discharge. Constant-density groundwater flow models thus were considered to be valid for the study area. Groundwater Modeling Results and Steady-State Assumptions As noted above, the groundwater modeling studies had two main goals: (1) provide boundary conditions for a hydrodynamic model of the MSJR, and (2) test whether increased groundwater discharge and chloride load would appreciably alter river conditions if river water levels declined due to surface water withdrawals. Because groundwater discharge is nearly impossible to measure by direct observation, modeling or indirect estimates based on a river basin mass balance are the only practical alternatives. Two steady-state groundwater flow models based on the USGS’s MODFLOW were used by the District to estimate groundwater discharge and chloride load to the St. Johns River: the North-Central Florida Model (NCF) and the East-Central Florida Model (ECF). The NCF covers the Lower Ocklawaha River and a small portion of the northern MSJR; the ECF covers the northern USJR and most of the MSJR. Actual groundwater discharge varies seasonally in response to wet and dry periods; the District’s use of steady-state values for the WSIS assumes that seasonal variability of discharge is of secondary importance compared to average discharge. The District concluded from the modeling studies that diffuse groundwater discharge is the dominant source of chloride to the MSJR (~75% of the chloride load to the reach upstream of US17) and that water withdrawals would have insignificant effects on overall discharge and chloride budgets of associated river segments. These conclusions were based on the finding that neither groundwater discharge or chloride load is appreciably altered by lowered river stage because the hydraulic head that drives groundwater discharge is much larger than the maximum expected reduction in river stage.

OCR for page 19
24 Review of the St. Johns River Water Supply Impact Study: Final Report The steady-state approximation for groundwater discharge to EFDC requires testing because groundwater discharge does respond to seasonal and inter-annual variations of rainfall and pumping. The groundwater models provided constant groundwater discharge to EFDC, but the latter is a dynamic model that simulates hydrodynamic variables at hourly time scales. EFDC simulations made using steady-state groundwater discharge were found to be nearly indistinguishable from simulations made using observed transient groundwater discharge data (Belaineh, 2010), however, and District scientists thus concluded that the use of steady-state groundwater discharge as a boundary condition to EFDC is justified. Critique The District’s scientists are commended for their efforts to compile much more complete documentation to defend the assumptions that allowed them to use a simplified approach to compute groundwater discharges and salinity fluxes to the river. Their report (Sucsy et al., 2011) covers the key factors involved in computing groundwater discharge and salinity boundary conditions for the hydrodynamic EFDC model and presents significantly improved documentation on the validity of the assumptions regarding constant density and steady state modeling, thus resolving the Committee’s concerns as expressed in NRC (2009). Three issues were not adequately addressed in the WSIS. First, the sensitivity of groundwater discharge to hydraulic conductivity was not investigated. It is well known that there are large uncertainties and ranges associated with defining hydraulic conductivity in regional groundwater modeling studies. Second, Sucsy et al. (2011) states that the models were calibrated to average groundwater conditions for 1995 but it does not discuss the calibration of diffuse groundwater discharge into the river. Finally, a major concern that the Committee expressed in NRC (2010) is that groundwater discharge from the surficial aquifer system (SAS) was not adequately modeled in the WSIS (due to the constraints of the HSPF model). In some reaches of the river this discharge may be an important contribution to flow and river stage, and it can influence the extent of inundation in riparian wetlands during periods of low river stage. HYDROLOGIC MODELING The District’s hydrologic modeling used standard approaches for meteorological forcing and watershed runoff. As discussed in NRC (2010), the Committee considers these approaches to be reasonable with some reservations. This is summarized below along with some additional comments based on more recent information. To clarify the Committee’s concerns, it is useful to recognize that the hydrologic modeling effort served two roles in the overall study: (1) Providing time- and space-varying flow rates into the middle and lower SJR as forcing conditions for the hydrodynamic model, which in turn provides water level and flow conditions for the ecological models for those segments, (2) Providing time- and space-varying flow conditions in the upper SJR for direct use in the ecological models for these predominantly wetland areas. In the first role, the hydrologic model can be considered reasonably successful. Because runoff is a one-way process that accumulates flows into the main-stem river, local catchment

OCR for page 19
Hydrology and Hydrodynamics 25 runoff errors in a calibrated hydrologic model will tend to be offsetting. That is, given the limited availability of calibration data and its bias towards downstream locations, the hydrologic model generally will calculate the calibrated accumulated fluxes with acceptable accuracy. Local fluxes that contribute to the accumulated fluxes may not have sufficient calibration data, however, to adequately distinguish the partitioning of flows from multiple upstream catchments, and thus they may have more significant error. In the second role, the hydrologic model has limitations in its ability to model the catchment-by-catchment wetland effects in the USJR. The Committee has four reservations associated with hydrological modeling as boundary conditions to hydrodynamic modeling of the MSJR and LSJR. First, future modeling should be conducted using the best available rainfall data (NEXRAD Doppler) rather than NWS gages. Second, the uncertainty associated with using a model calibrated to 1995 land use conditions for 2030 land use conditions needs to be analyzed. One approach to this effort would be to use 2010 land use data without further calibration and analyze the model errors introduced. Third, land- use decisions and best management practices may affect the assumptions used to generate 2030 land use conditions and stormwater yield. The District should evaluate the significance of possible changes relative to the predicted catchment fluxes. Fourth, the analysis across multiple scenarios should examine whether confounding processes can lead to offsetting errors that underestimate impacts. The Committee has two principal reservations with regard to the hydrologic model's application to wetlands. First and foremost, the HSPF model does not include critical interactions between wetlands and surficial groundwater through water table elevation, in situ storage, or unconfined aquifer and river exchanges. In addition, the spatial resolution of HSPF sub-basins is too coarse to adequately represent the scale of wetland response needed for ecological modeling. The Hydroperiod Tool provides some help in these areas, but how errors in the original HSPF water flux estimation affect the Hydroperiod results is an unanswered question. HYDRODYNAMIC MODEL CALIBRATION AND CONFIRMATION The Committee reviewed preliminary results of the LSRJ/MSJR hydrodynamic model calibration in a previous report (NRC, 2010). More extensive methods and results for the hydrodynamic model calibration are presented in Sucsy et al. (2011) Volume 2, Chapter 6. The Committee regards this chapter as one of the most comprehensive model calibration documents produced for any U.S. river study. The authors carefully responded to the Committee’s comments and discussions from NRC (2010). The calibration report covers all the key factors involved in designing a working hydrodynamic model, including development of a digital elevation model, grid selection and testing, initial and final calibration results, comparisons with observed data, and confirmation (verification) against a data set not included in the calibration data set. The results show that the calibrated model provides a good representation of the physics of transport through the LSJR and MSJR and thus is a reasonable tool for development of the WSIS. The Committee congratulates the District on providing a well-documented scientific foundation for their modeling work.

OCR for page 19
26 Review of the St. Johns River Water Supply Impact Study: Final Report Analysis Methods In Volume 2, Chapter 7 of Sucsy et al. (2011), three modeled river properties (water level, salinity, and water age) were analyzed across three different scenario sets (hindcast, forecast, future) using seven different methods: (1) time-series, (2) point-to-point, (3) statistics, (4) cumulative distribution functions, (5) longitudinal distributions along the river, (6) discharge- difference, and (7) intensity-frequency duration. Examples of each method are given below. The hindcast analyses were based on scenarios Base1995NN, Half1995NN, and Full1995NN. These analyses used the known conditions in 1995 to examine sensitivity of the historic system to different withdrawal rates (without any competing effects). The forecast analyses examined how expected changes in land use, sea level rise, and the USJR projects would affect the system both with and without withdrawals. The analyses of future scenarios provided a set of “what if” scenarios for insight into other large changes that might occur. All of the above analyses were model-model comparisons, i.e., the model results under different forcing conditions were compared.4 By comparing these differences, insight was gained into how different factors affect the model. The extensive work done to calibrate and confirm (verify) the model (Volume 2, Chapter 6) provides confidence that the model provides a reasonable representation of the river physics, and so the comparison of different model results provides insight into how the river is expected to respond. Time-series Analysis An example of a time-series analysis is shown below as Figure 2-1. Each line on the figure represents the difference between the salinity at Acosta Bridge in one of the water withdrawal scenarios (either Half or Full) and the zero-withdrawal scenario (Base). This comparison allows the reader to observe how salinity increases due to water withdrawal, without the complicating factors of the USJR projects, sea level rise, or changes in land use. Time series analyses require a separate graph for each location where the analysis is desired. Because the model produces a time series for each model grid cell, thousands of pages could be filled with these graphs, which would be of little practical utility. The District appropriately limited itself to presenting a few time-series graphs for key locations that illustrate the principal physical responses. Point-to-Point Analysis Because a time-series figure provides results at a single location, understanding the overall effects for different scenarios across multiple locations with these graphs is difficult. To provide better insight, model scenarios were compared on point-to-point graphs, as shown in Figure 2-2, which compares the salinity (ppt) in one scenario at selected points in space over all time (daily-averaged) to the same points in space and time for another modeled scenario. These comparisons were made for the entire ten-year data series, providing 3,650 data points for each 4 Model-observation comparisons are provided extensively in the calibration section, Volume 2, Chapter 6 of Sucsy et al. (2011).

OCR for page 19
Hydrology and Hydrodynamics 27 FIGURE 2-1 Salinity difference at Acosta bridge. At this location, the vertically averaged mean salinity over this ten-year period is 5.9 ppt under base conditions, 6.1 ppt under the half withdrawal condition, and 6.2 ppt under the full withdrawal condition. SOURCE: Sucsy et al. (2011). of the ten selected spatial locations. Figure 2-2 thus shows that the Full1995NN withdrawal scenario produces slightly higher daily-averaged salinities than the Base1995NN no-withdrawal scenario. This approach allows readers to easily find any major effects induced by changes to forcing functions. Unfortunately, because of the large number of points and the relatively small differences between scenarios the point-to-point plots have quite a bit of over-printing. This effect makes it difficult to discern differences between the spatial locations. For example, in Figure 2-2 the salinities at Buffalo Bluff and Lake George have been entirely overprinted by salinities at Shands Bridge, Orange Park, and Acosta. The overprinting effect is more dramatic in several other graphs, most notably in Figure 4-10 of Sucsy et al. (2011), Volume 2, Chapter 7 where the wide range of water levels at Lake Harney completely obscures all the other data. Statistical Analysis To further collapse the daily-averaged values of two models from point-to-point graphs, several statistical measures of the paired values were computed at different points along the river, which collapses the time variable at each point so that the overall effect over a year can be evaluated. The statistic of the greatest importance is the root mean square difference (RMSD) between model results, which is equivalent to the standard deviation caused by the added forcing. For example, Table 4-2 in SJRWMD (2011), Volume 2, Chapter 7 shows that the

OCR for page 19
28 Review of the St. Johns River Water Supply Impact Study: Final Report FIGURE 2-2 Point-to-Point comparison for model salinity (ppt) comparing full withdrawal with zero withdrawal. SOURCE: Sucsy et al. (2011). RMSD for daily averaged water surface levels at Lake Harney for Full1995NN and Base1995NN is 4.9 cm. This number characterizes the overall change in the range of water surface level variability at Lake Harney predicted by the model. Caution should be used with these statistics, however, because the collapse of time-varying effects into a single number may obscure unusual or rare events that might be important; for this reason the District also employed Difference-Discharge analyses (discussed below). Cumulative Distribution Function (CDF) The CDF was calculated from the output of each modeled scenario to show the percentage of the time that a variable would be at or below possible values at particular points in space. These graphs allow the reader to evaluate whether different models have observably different distributions over the ten years of simulation. For example, Figure 2-3, which shows the CDF of salinity at Shands Bridge, indicates that all the model scenarios produce a salinity of less than 0.5 ppt 90 percent of the time, and higher salinities occur at this site relatively rarely. The difference between the CDF lines shows that increasing water withdrawal coincides with some increases in salinities during the 5 percent of the time when the salinity is above 0.5 ppt.

OCR for page 19
Hydrology and Hydrodynamics 29 FIGURE 2-3 CDF of Shands Bridge salinity using models for different withdrawal conditions. SOURCE: Sucsy et al. (2011). Longitudinal Distributions Plots of the longitudinal distribution of mean variables and differences between mean variables in models provide a way to collapse the time variability and emphasize the space variability in model output, e.g., see Figure 2-4. The key difficulty with these figures is that they average the variable at a location over the entire ten-year simulation, which may hide effects from short-term events. Providing these graphs with standard error bars would provide further insight. FIGURE 2-4 Difference in mean salinity between withdrawal and no-withdrawal scenarios over a ten-year simulation. SOURCE: Sucsy et al. (2011).

OCR for page 19
30 Review of the St. Johns River Water Supply Impact Study: Final Report Discharge-Difference The difference in a variable between two model outputs can be analyzed based on when the difference occurs relative to river discharge. This approach provides a method for understanding the relationship between the river discharge and the changes associated with different forcing conditions. For example, the District grouped the model results for river discharge using 25 m3 s-1 intervals and plotted key statistical characteristics (median, minimum maximum and quartiles; see Figure 2-5) to illustrate how changes in the water surface level vary with discharge at DeLand. These graphs put the statistical analysis described earlier into context. For example, the statistical analysis for Lake Harney showed a mean change of 4.9 cm in the lake level, but the Discharge-Difference analysis showed that much larger differences can be expected on a routine basis, especially at higher flow rates. As Figure 2-5 shows, during high flows at DeLand the change in the water surface level could be substantially greater than 4.9 cm; indeed, the range of the 25th percentile value for the four box plots for flows between ~200 and ~300 cfs is 8 to 14 cm (mean of 12 cm), meaning that the water level during withdrawals on average would be ~12 cm lower when flows were ~200-300 cfs. Such decreases may affect littoral inundation. FIGURE 2-5 Water level difference between full withdrawal (Full1995NN) and no-withdrawal (Base1995NN) scenarios for Lake Harney over a ten-year simulation. SOURCE: Sucsy et al. (2011).

OCR for page 19
Hydrology and Hydrodynamics 31 Intensity-Frequency-Duration The intensity of an effect (e.g., a high water surface level), its frequency (denoted by the return period, or how often the effect is experienced), and its duration (how long it lasts) can be analyzed together in an Intensity-Frequency-Duration plot, as shown in Figure 2-6. One can see that the full withdrawal case can be understood to slightly reduce the intensity (water level) of an event with a given return period and a given duration (i.e., by comparing the black and red lines of similar return periods at a selected number of consecutive days). Alternatively, one might also say that for a particular intensity, the full withdrawal decreases the duration of that event for a given return period. Because of the log scale and the curvature of the graphs, this effect may be more significant than the reduction of intensity. For example, Figure 2-6 shows that for a two-year return period, a water level of 1.7 m is expected to occur with a 100-day duration for the Base1995NN case, whereas the same water level will occur with only a 70-day duration for the Full1995NN withdrawal case. Understanding this behavior should be a critical part of analyzing duration changes for littoral inundation and drying. Similarly, to understand the impact of salinity changes, Intensity-Frequency-Duration curves should be analyzed with respect to changes in duration and intensity that might adversely impact ecological conditions. For example, Figure 2-7 shows that a salinity of 4 ppt with a return period of ten years has a duration of about three days in the baseline model, but a duration of ten days under the full withdrawal5. FIGURE 2-6 High water levels in Lake Harney over 10 year simulation. SOURCE: Sucsy et al. (2011). 5 Note that Figure 2-7 is difficult to interpret because there seems to be a mismatch between the numbers and the tick marks on the vertical axis.

OCR for page 19
32 Review of the St. Johns River Water Supply Impact Study: Final Report FIGURE 2-7 Salinity at Shands Bridge over a 10-year simulation. SOURCE: Sucsy et al. (2011). Linearity Test The range of scenarios used by the District (Table 2-1) allowed it to develop a linearity test to better understand the relative contributions of different forcing conditions to physical effects and whether or not interactions between different forcing conditions lead to nonlinear behavior. Linear behavior allows different forcing conditions and model response to be analyzed individually, whereas nonlinear behavior requires different forcing conditions to be analyzed together. For example, with four different forcing conditions that each might be modeled at three different levels, linear behavior requires only 12 model simulations (i.e., 4×3 simulations) to fully understand effects from different forcing conditions, but nonlinear behavior requires 64 simulations (i.e.,43 simulations), making both the computations and analysis a much more substantial task. Figure 2-8 provides an example of a linearity test that illustrates the overall impact of the different forcing conditions along the river length. The size of the color-coded components in each bar indicates the magnitude of the effect attributable to a given forcing condition, and in each case the bar components add up to close to 100 percent indicating the effects are substantially linear. In contrast, Figure 2-9 shows how the possible Future Conditions (Ocklawaha withdrawal, dredging, high sea-level rise, and water reuse) have a nonlinear effect on salinity.

OCR for page 19
Hydrology and Hydrodynamics 33 FIGURE 2-8 Linearity test for water level effects in Forecast scenarios. SOURCE: Sucsy et al. (2011). FIGURE 2-9 Linearity test for salinity effects in Future Conditions scenarios. SOURCE: Sucsy et al. (2011).

OCR for page 19
34 Review of the St. Johns River Water Supply Impact Study: Final Report Uncertainty Analysis The District has made impressive efforts to develop a formal uncertainty analysis for the hydrodynamic model of the St. Johns River (Sucsy et al., 2011, Section 9, Volume 2, Chapter 7). They used a method called “First Order Error Analysis” (FOEA), through which they looked at the sensitivity of the model results to different model inputs. The principal idea behind FOEA is to estimate how uncertainty in the model results might be impacted by uncertainty in the forcing conditions. The sensitivity of the model results to the uncertainty in forcing is evaluated with Dimensionless Sensitivity Coefficients (DSC), which represent the relative change in model output for the change in model input. Using the uncertainty in the forcing conditions and the DSCs, the District was able to quantify confidence intervals for model time series to help bound their expected model accuracy. In addition to estimating the model uncertainty associated with input data, the District analyzed uncertainty in the model-predicted changes between the base and withdrawal models. The principal limitation of the District’s uncertainty analysis is that it focuses on evaluating model performance uncertainty relative to the hindcast conditions and does not address the underlying uncertainty in forecast conditions. The District’s focus on hindcast uncertainty is commensurate with the state-of-the-art in hydrodynamic modeling; however, future land-use conditions, future sea-level rise, and permitted water withdrawal restrictions are also subject to considerable uncertainty. Unfortunately, such forecast conditions are uncertain in both time and space, which presents an ongoing challenge to uncertainty quantification for hydrodynamic modeling. HYDRODYNAMIC MODEL RESULTS Sections 4, 5 and 8 in Vol. 2, Chapter 7 of Sucsy et al. (2011) provide item-by-item analyses of model results using the tools described above. A more general discussion of these results is found in the Section 10 Summary of the same document, which provides individual summaries for prior sections, sequentially dealing with each effect (water level, salinity, water age). These summaries focus on general observations extracted from the detailed graphical and statistical analyses. Overall, the District has presented a reasonable effort to encapsulate a variety of complex analyses conducted using the hydrodynamic model. Critique The work done on building, testing, and analyzing the hydrodynamic model is state-of- the-art science. The Committee congratulates the District’s scientists for their meticulous efforts to make this complicated model work and to make sure that they can quantify the propagation of data uncertainty into hydrodynamic model uncertainty. The Committee has confidence that the District is building their WSIS analyses on a hydrodynamic foundation that is well-tested, robust, and well-understood. The report lacks a comprehensive synthesis of the model results, however. To some extent, the discussion of the 155 MGD withdrawal effects in Section 10.1 comes close to a synthesis, but it remains mired in an item-by-item analysis that in some ways misses the forest

OCR for page 19
Hydrology and Hydrodynamics 35 for the trees. After reading Chapter 7, it is clear what the District has done and how each piece behaves, but not necessarily what it means. Of particular concern is that the uncertainty analysis has not been synthesized with the water level, salinity, and age analyses to provide a deeper understanding of the model’s ability to explain the system. Associated with the lack of synthesis is the lack of discussion and graphs that could put into context the relationships between key mechanisms of the river system and their responses to forecast conditions. It appears likely that uncertainties in forecast conditions may be the critical unknown. From results shown during the committee meetings, the Committee concludes that there are two major competing effects—sea-level rise and increased runoff due to future land use changes such as development—that both affect water surface levels and salinity. Sea-level rise will push salinity further upstream, which effectively adds to withdrawal impacts, but increased runoff pushes salinity further downstream, subtracting from withdrawal. Both will lead to increased inundation of littoral zones, which counters the hydrologic effects of withdrawals. The uncertainties associated with these effects need to be discussed and analyzed. The District’s presentations during committee meetings indicated the magnitudes of these forecast effects are significantly larger than their net effect on withdrawal, and so the uncertainty in the forecasts could easily dominate the system. The District now has the tools and the data to put together a more complete picture of how the physics of this system will behave and they should be able to quantify the uncertainties associated with future conditions. CONCLUSIONS AND RECOMMENDATIONS The Committee recommends that the District develop a separate hydrodynamic/ hydrologic synthesis study that uses results from all the models to present a clear picture of the state of the science with regard to the river response. This should be thought of as a “big-picture view” of the river mechanics based on the understanding developed through the three models. This report should be directed at non-modelers and non-hydrologists so that they can better understand the implications of the extensive modeling study. The District should focus on the type of questions that will concern ecological scientists and the general public. Over the long term, the District should develop a groundwater model that simulates the full interaction of the river with the SAS and UFA under both steady state and transient conditions. In this way the District will have a model that represents the physical system that allows a better analysis of the river response. The District also should develop an uncertainty analysis for groundwater discharge to the river based on hydraulic conductivity, which is well known to have uncertainties that can be an order of magnitude or more for basins the size of the middle St. Johns. REFERENCES Belaineh, G. 2010. Investigation of the interaction between St Johns River and the underlying aquifers in the Middle Basin. Presentation to the NRC committee. March 29, 2010. NRC (National Research Council). 2009a. Review of the St. Johns River Water Supply Impact Study: Report 1. Washington, DC: The National Academies Press.

OCR for page 19
36 Review of the St. Johns River Water Supply Impact Study: Final Report NRC. 2010. Review of the St. Johns River Water Supply Impact Study: Report 3. Washington, DC: The National Academies Press. Sucsy, P., G. Belaineh, E. Carter, D. Christian, M. Cullum, J. Stewart, and Y. Zhang. 2010. Hydrodynamic Modeling Results. Palatka, FL: SJRWMD. Sucsy, P., G. Belaineh, E. Carter, D. Christian, M. Cullum, J. Stewart, and Y. Zhang. 2011. Hydrodynamic Modeling Results. Palatka, FL: SJRWMD. Note: This is the same report as above, but updated in May 2011 as Chapter 7 of draft final H&H report. It is this report that contains all the chapters that are referred to.