ML23333A014

Attachment F - Groundwater Tek Inc - Peer Review Study Final-1

ML23333A014
Person / Time
Site:	Turkey Point
Issue date:	11/27/2023
From:	Office of Nuclear Material Safety and Safeguards
To:
Shared Package
ML23333A008	List: ML23333A009 ML23333A010 ML23333A011 ML23333A012 ML23333A013 ML23333A014 ML23333A015 ML23333A016 ML23333A017 ML23333A018 ML23333A019 ML23333A020 ML23333A021 NRC-2022-0172, Miami Waterkeeper Comment to NRC on Turkey Pt. 2023 Dseis (November 26, 2023)
References
Download: ML23333A014 (1)

Text

Attachment F Review of FPLs Groundwater Flow and Salt Transport Models and Assessment of the First Year Operation of the RWS Prepared for Division of Environment and Resources Management Miami-Dade County, Florida Groundwater Tek Inc.

Naples, Florida July 2020

Attachment F Table Contents Page Executive Summary 1

1. Introduction 6

2. Review the 2016 groundwater model and supporting documentation 8 2.1 General Model Description 8 2.2 Model Calibration 10 2.3 Prediction Model 13 2.4 Evaluation of 2016 Groundwater Flow and Salt Transport Models 13

3. Review of 2019 Groundwater Flow and Solute Transport Model (V3) 15 3.1 Revised Flow and Solute Transport Model (V3) 15 3.2 V3 Model Calibration 15

4. Model Verification Review of Model Verification Performed by FPL 18 4.1 Evaluation Based on Water Quality and Water Level Data 18 4.2 Comparison of Salt Extraction Rates 19 4.3 Modeled Extent of the Hypersaline Plume 20 4.4 CSEM Survey 20

5. Review of Remediation Prediction 24 5.1 Salt Extraction 24 5.2 Retraction of the Hypersaline Plume 24 5.3 Capture Zone Analysis 26 5.4 Water Levels and Drawdown in the Surficial Aquifer 27 5.5 Comments on FPLs Recommendations and Improvements 27

6. Sensitivity Analysis 29 6.1 FPL Sensitivity Simulations 29 6.2 Additional Sensitivity Runs 30 6.3 Summary of Sensitivity Analysis Simulations 32

7. Evaluation of FPL Conclusions and Recommendations 33

8. Summary, Conclusions and Recommendations 37 1.1 Summary 37 1.2 Conclusions 38 1.3 Recommendations 39

9. References 42 Tables Figures 1

Attachment F Executive Summary Groundwater Tek Inc. (GTI) was retained by the Miami-Dade County Division of Environment and Resources Management (Miami-Dade DERM) to evaluate the site-specific variable density groundwater flow and solute transport model developed by FPL and to assess the first year performance of the remediation system as whether the remedial objectives could be met as specified in the Miami-Dade County Consent Agreement (MDC, 2015).

The model was developed by FPL using the USGS computer code SEAWAT. Model simulations were used to develop a schedule for the mitigation and subsequently to evaluate the effectiveness of a recovery well system (RWS) to mitigation of salt water contamination of the Biscayne Aquifer resulting from the saline water present in the FPLs Cooling Canal System (CCS) at its Turkey Point Nuclear Power Plant in Southern Miami-Dade County.

Scope of Work Specifically, GTI was tasked by Miami-Dade in their scope of work (SOW) to

1. Review available historical reports to develop an understanding of the aquifer characteristic in the area of the project.

2. Review the 2016 groundwater model and supporting documentation.

3. Review and run the 2019 groundwater model to

i. Verify the appropriate incorporation of the monitoring data produced during the first year of operation as well as the CSEMs data.

ii. Verify model predictions with respect to the capacity of the RWS to retract the hypersaline plume in the overall time frame predicted as well as verify agreement with the model forecast at each milestone interval (e.g., year three, year five, etc.).

4. Review and evaluate any sensitivity analysis performed.

5. Evaluate FPL conclusions and recommendations.

Work Performed The following work elements were completed by GTI in accordance with the Miami-Dades SOW:

(1) Reviewed the historic and recent documents that pertain to the project; (2) Reviewed conceptual model, model input data, model structure, and model calibration procedure and statistics; 1

Attachment F (3) Reran and verified all the groundwater flow and salt transport models provided by FPL including:

a. The 4-stage calibration models;

b. The 13 month verification model;

c. The 9 year prediction model;

d. Two sensitivity analysis models.

(4) Performed additional simulations to investigate the sensitivity of model input parameters pertaining to salt transport; (5) Performed backward particle tracking and capture zone analyses; (6) Performed layer-by-layer plume volumetric change analysis; (7) Reviewed the water levels, water quality and salt extraction data collected during the first year operation of the RWS; (8) Reviewed the CSEM survey data presented in FPL reports; (9) Assessed the hydrogeological impacts from the RWS operation; (10) Assessed the performance of the first year operation and the effectiveness of the RWS to meet the remediation objectives; (11) Reviewed FPLs conclusions and recommendations.

Findings and Conclusions The following findings and conclusions, based on the data available at the time of this review and the results of the evaluations described above, are summarized below:

(1) FPL developed a site-specific variable density groundwater flow and solute transport model using the USGS computer code SEAWAT. The three-dimensional model was well designed, calibrated with data from more than 70 years. The model was evaluated with the data collected during the first year operation of the RWS (May 2018 to May 2019). The model predicted salt extraction rates, water levels and salinity appeared to be in acceptable agreement with the data measured in the field. However, comparison of the CSEM data regarding the extents of the plume of hypersaline plume (chloride concentration exceeding 19,000 mg/l) has not been consistent with the predictions from groundwater model simulations. As noticed by FPL, there is a disagreement for the plume extent in the deep portion of the aquifer between the model and the CSEM data.

(2) Model predictions indicated the RWS would be capable of retracting the hypersaline plume up to model layer 8 (about 60 feet deep) to FPLs property at west of the CCS and up to model layer 5 (about 35 ft deep) at north of the CCS.

2

Attachment F (3) The groundwater model predictions indicate that the existing RWS has little impact to the plume in the deep portion of the Biscayne aquifer (model layers 9 to 11) west of the CCS. To the north of the CCS, the groundwater modeling indicates the current RWS will be retract the flume up to model layer 5 and has little impact to the middle and lower portions of the aquifer (from model layer 7 to layer 11).

(4) CSEM surveys were conducted to delineate the hypersaline plume in the aquifer in April 2018 and May 2019. FPL reported a 22% reduction of the plume volume based on the CSEM survey data. However, the reduction of the plume varied greatly with depths, from 40% plume of volume reduction at the middle to only 5% to 6% of plume volume reduction at the bottom of the aquifer.

(5) The CSEM Voxel grid analysis indicated the plume volume within the CSEM survey area was reduced in the shallow and middle portions of the aquifer within the first year of the RWS operation. The plume volume analysis based on the numerical model showed a similar pattern but indicated the plume in the lower portion of the aquifer may not shrink or even get slightly larger instead.

(6) Both modeling results and the CSEM survey data indicate the RWS is capable of preventing the hypersaline plume from its further expansive migration at the west and north of the CCS in the Biscayne aquifer.

(7) Both groundwater modeling and the CSEM survey data indicate that the RWS has little remedial effect to the hypersaline plume currently in the deep portion of the Biscayne aquifer west and north of the CCS.

(8) Since the CSEM data cannot be used to predict how fast the hypersaline plume will be retracted, a well-calibrated and validated groundwater model is needed as a reliable prediction tool. It is strongly recommended to verify the chloride concentration interpreted from the CSEM survey using field measured data. The discrepancy between the CSEM data and modeling prediction in the deep portion of the aquifer should be resolved and the numerical model should be revised and recalibrated before using the model as a predictive tool.

(9) Data from TPGW-18 is useful to validate both the CSEM survey data and the numerical model. Its unique location also provides an opportunity for an early detection whether the plume is retracting eastward.

(10) Extra monitoring well(s) within the hypersaline plume (for example, between TPGW-4 and TPGW-17; between TPGW-18 and RWS-5, if applicable) might be helpful to detect and verify the progress of planned remediation action.

(11) Based on the information reviewed and analyzed, the current RWS does not appear to be capable of meeting the remediation objectives of retracting the hypersaline plume to the FPLs property from either west or north of the CCS per 3

Attachment F Miami-Dade County Consent Agreement and Florida Department of Environmental Protection Consent Order.

Suggestions and Recommendations The following suggestions and recommendations, based on the data available at the time of this review and the results of the evaluations described in this report, are presented by GTI as below:

(1) FPL should explain why the modeled salinity results for these lower layers in Year 1 from a calibrated and verified model did not match the CSEM data. If the CSEM data is proven to be more reliable, the SEAWAT model should be revised, re-calibrated and re-verified as a predictive tool. And all model predictions presented in the first year remediation action annual report (FPL, 2019) should be updated based on the revised model.

(2) It is recommended to verify the chloride concentration interpreted from the CSEM survey using field measured data. The discrepancy between the CSEM data and model prediction in the deep portion of the aquifer should be resolved and the numerical model should be revised and recalibrated before using the model as a predictive tool.

(3) FPL should explain why the plume volume in Voxel grid layer 4 indicated an unintuitive 140% increase over the first year of RWS operation based on the CSEM data so a higher level of confidence towards the CSEM data can be achieved.

(4) When the SEAWAT model is to be updated with Year 2 salinity data, FPL should demonstrate that the data that they will be using is real and representative.

(5) It would help better understand the RWS performance if a layer-by-layer water source analysis can be performed for each of the RWS wells using a backward particle tracking method.

(6) It might be more meaningful to compute the salt mass reduction in the plume with time by factoring in porosity and salt concentration. MDC CA actually requires FPL demonstrate a statistically valid reduction in the salt mass and volumetric extent of hypersaline water.

(7) An optimization analysis is recommended for the extraction wells by varying their pumping rates and schedules so the overall efficiency of the RWS can be maximized and the cost of remediation may be reduced.

(8) It is recommended to continuously monitor the water quality changes at TPGW-18 at its middle and deep screens. TPGW-18 may be useful to validate both the CSEM survey data and the numerical model. Its unique location also provides an opportunity for early detection whether the plume is retracting eastward in the deep portion of the aquifer. The chloride concentration was 25,400 mg/l (about 4

Attachment F relative salinity 1.3) in March 2019. New water quality data should be available soon to allow a progress assessment of the remediation system after its two years of operation.

(9) Extra monitoring wells open to the deep portion of the aquifer will be helpful and thus recommended to verify the progress of remediation, the CSEM data and numerical models. Suggested locations may be between TPGW-17 and TPGW-4, and TPGW-18 and RWS-5 if applicable.

(10) The drawdown data collected during the shut-off evens might be useful as aquifer performance tests and model calibration, particularly for the deep portion of the aquifer.

(11) In the groundwater model, the initial salt concentration at locations where water quality data were collected should be adjusted based on the field data so the error from the previous simulation would be carried into the next simulation.

(12) The initial extents of the hypersaline plume: FPL suggested the plume extents in deep layers were overestimated in the model, compared to the CSEM data. If the CSEM data are validated, the initial salt concentration in the model should updated by the CSEM data wherever the CSEM data available.

(13) To remediate the deep portion of the plume found west and north of the CCS, alternative technologies, such as directional drilling or horizontal wells (van Heest, 2013; U.S. EPA, 2017) may be considered.

(14) Extra extraction well may be needed to remediate the hypersaline plume north of the CCS. Caution should be excised to avoid introducing salt water intrusion to the aquifer.

5

Attachment F

1. Introduction Florida Power and Light (FPL)s Cooling Canal System (CCS) was constructed in 1973 to provide water for cooling the reactors at its Turkey Point Nuclear Power Plant in Southern Miami-Dade County (Figure 1-1). The CCS, which consists of a 5,900-acre network of canals, serves as a heat exchange for the power plant units at Turkey Point. An 18 foot deep ditch (the interceptor ditch) located west of the cooling canal system was constructed and designed to provide a hydraulic barrier to prevent to the inland westward migration of water from the CCS.

The system is closed to surface water bodies such as the Biscayne Bay or Card Sound, becoming a closed-loop system.

A plume of hypersaline water (chloride concentration exceeding 19,000 mg/l) has formed in the Biscayne aquifer below the CCS due to long term evaporation of the hot water in the CCS (Chin, 2016). The South Florida Water Management District has determined that the plume from the CCS has expanded westward beyond L-31E Canal (SFWMD, 2013).

A Consent Agreement (CA) has been executed between Miami-Dade County (MDC) and FPL (MDC 2015). The MDC consent agreement (Oct. 6, 2015) requires FPL to demonstrate a statistically valid reduction in the salt mass and volumetric extent of hypersaline water (as represented by chloride concentrations above 19,000 mg/l) in groundwater west and north of FPLs property without creating adverse environmental impacts. A further objective of this Consent Agreement is to reduce the rate of, and as an ultimate goal, arrest migration of hypersaline groundwater. According to the MDC CA (17.b): FPL shall intercept, capture, contain and retract hypersaline groundwater (groundwater with a chloride concentration of greater than 19,000 mg/l) to Property boundary to achieve the objectives of this Consent Agreement.

Florida Department of Environmental Protection (FDEP) issued a Consent Order (CO) on June 20, 2016 (FDEP 2016). This FDEP CO requires FPL to implement a remediation project that shall include a recovery well system that will halt the westward migration of hypersaline water from the CCS within 3 years and reduce the westward extent of the hypersaline plume to the L-31E canal within 10 years without adverse environmental impacts.

According to the MDC CA, FPL implemented a remedial action plan to retract the plume found west and north of the CCS back to the FPLs property.

A groundwater flow and solute transport variable density model was developed by FPL to facilitate proposed remediations options. The USGS computer code SEAWAT (Guo and Langevin, 2003; Langevin et al., 2008) was selected for the modeling development.

The variable-density groundwater flow and solute transport model developed for the design the remediation system went through a number of stages of model calibration and 6

Attachment F verification. According to the remediation action plan (FPL, 2017), 10 extraction wells, tapping at the bottom of the aquifer, were installed along the western and northern sides of the CCS to extract the saline water as shown in Figure 1-2. Each of these wells pumps at 1.5 million gallons per day (mgd).

The remedial system began operating in May, 2018. Data from the first year operation, including water levels, water quality and salt extration amount, etc. were used to veirify the numerical model and assess the effectiveness of the remediation system.

Groundwater Tek Inc. (GTI) was retained to conduct a review of the numerical models and data collected during the first-year operation and to assess the first year performance of the remediation system as whether the remedial objectives could be met as specified in the MDC CA (MDC 2015).

7

Attachment F

2. Review the 2016 Groundwater Model and Supporting Documentations 2.1 General Model Description A three-dimensional groundwater flow and solute transport model was developed to specifically address the remediation of the hypersaline conditions at the CCS (Tetra Tech, 2016).

This model incorporated detailed bathymetry, operational features, water level and salinity data. The model was capable of modeling groundwater flow, interactions between groundwater and surface water, transport of salt under non-isothermal conditions. The USGS computer code, SEAWAT (Guo and Langevin, 2003; Langevin et al., 2008) was selected to perform the numerical simulations.

The initial objectives of this model were (1) to support the design of a Recovery Well System (RWS) to intercept, capture and contain the hypersaline plume north and west of the CCS, and support authorization through appropriate regulatory processes and demonstrate the proposed remediation system would not create adverse impacts to groundwater, wetland or other environmental resources; and (2) to continue to assess the status and effectiveness of the remediation system operation in meeting the objectives of the Consent Agreement.

The model covers an area of 276 square miles. Its area extent of the model is bounded on all sides by major hydrogeologic features:

- The C-103/Mowry Canal forms the northern boundary;

- C-111 Canal bounds the model on the west and south;

- The eastern portion of Biscayne Bay for the eastern Boundary of the model.

The model has 295 rows and 274 columns. Variable grid spacings, ranging from 200 ft to 500 ft, were used in the model, with smaller grid cell dimensions located near the CCS.

Vertically, the model has 11 layers, with variable thicknesses, to represent the Biscayne aquifer. The uppermost layer consists of unconsolidated surficial sediments. Model layers 2 to 4 represent the Miami Oolite limestone, and model layers 5 through 11 represent the Fort Thompson Formation. Model layers 4 and 8, representing two High Flow Zones, were assigned with high permeability values, and based on available well logs and Aquifer Performance Test (APT) data.

The cooling canals were modeled using the MODFLOW RIVER package (Harbaugh et al.,

2000). Some of the river cells, representing the cooling canals, extended as deep as model layer

4. The stages and salinity in these river cells were derived from field measurements.

Two solute species which may affect the fluid density, salt and temperature, were considered in the model. The fluid density is computed as:

= 1.773125 ( ) 0.013252 ( ) + 62.3159 (1) 8

Attachment F where is fluid density [lb/ft3], S is fluid relative salinity [dimensionless], So is the reference-relative salinity (zero), T is fluid temperature [°C], and To is the reference fluid temperature (20°C).

The relationship of the fluid density-solute concentration is assumed to be linear within the range of the salinity in the model. Considering the hypersaline plume is actually originated from seawater, this assumption should be valid (Guo and Langevin, 2003).

Relative salinity (S) is used in all the models to represent the salt concentration in the groundwater. For sea water, S=1.0 which is equivalent to 35 PSU or 35,000 mg/l. Since a typical seawater has chloride concentration about 19,000 mg/l, so S=1 can be referred to the chloride concentration of 19,000 mg/l. It can be shown that the density would be 64.09 lbs/ft3 for sea water at 20 degrees of Celsius.

Boundary conditions General Head Boundaries (GHB) (Harbaugh et al., 2000) were assigned along the model perimeters. The head and salinity values for the boundaries were based on measured data.

Biscayne Bay was modeled as a time-varying constant head and constant concentration boundary. The head values specified for Biscayne Bay are based on the water level measurements collected continuously at two stations since 2010. The long-term average head value was applied to the steady-state flow model (1940-1960). Seasonal averages of the measured heads were applied to the seasonal transient model (1960-2010), and measured monthly average data were applied to the monthly transient model (2010-2015).

A methodology based on the recharge estimation approach of Hughes and White (2014) was used in these models. NEXRAD rainfall and reference evapotranspiration data from 1996 were used to estimate the spatial and temporal distribution of net recharge for the entire model domain.

Groundwater pumping is simulated using MODFLOWs WELL package (Harbaugh el al.,

2000). Municipal, industrial and agricultural groundwater pumping are all included. Pumping rates from the agricultural wells were estimated based on the NEXRAD rainfall and evapotranspiration data in the associated agricultural area. The monthly NEXRAD-based rainfall and evapotranspiration estimates in these agricultural areas were used to determine when and where an evapotranspiration deficit occurred, and the deficit was multiplied by the crop areas to obtain the volumetric pumping rate.

Initial Conditions A steady-state flow model was first built to simulate the groundwater flow for the conditions before the development of groundwater in the area. The initial salt concentration was then estimated based on the Ghyben-Herzberg relation (Bear, 1972) and the heads from the steady-state flow model.

9

Attachment F The groundwater flow and solute transport model was developed as a sequence of calibration models. In the sequence, the results from the earlier model were used as the initial conditions for the model of the next step. Finally the results from the monthly calibration model, including the heads, relative salinity and temperature were used as the initial conditions for the verification and prediction models (Tetra Tech, 2016).

Hydrogeologic properties Hydrostratographic layering for the regional Biscayne aquifer model was defined based on 41 well boring logs, including Turkey Point Ground Water wells (TPGW-1 through TPGW-14; JLA Geosciences, 2010) and additional boring logs interpreted and reported by Fish and Stewart (1991) and Parker et al. (1955). The Biscayne aquifer was divided into 11 model layers, representing a total thickness from about 60 ft at the western model border to about 100 ft at the CCS area.

Two zones of high hydraulic conductivity material (High Flow Zones) were represented in the regional model and were assumed to be continuous throughout the entire model domain. The upper High Flow Zone occurs at the base of the Miami Oolite, and the lower High Flow Zone is located in the approximate middle of the Fort Thompson formation (FPL, 2019).

These two High Flow Zones are represented by model layers 4 and 8 respectively.

A localized groundwater model was developed and calibrated by SDI Environmental Services (Enercon, 2016a) to match local changes in water levels and salinities measured during an APT which was conducted near TPGW-1. This local-scale model used three comparatively low-conductivity isotropic layers interspersed within the Biscayne aquifer to represent observed geology (Tetra Tech, 2016a). The hydrogeological parameters (horizontal and vertical hydraulic conductivities) were further adjusted during model calibrated.

Surface water and groundwater interaction There are a number canals in the model area. Interactions between the surface water and groundwater were represented using MODFLOWs DRAIN and River packages (Harbaugh et al., 2000). The CCS and primary canals (e.g. the perimeter canals, C-103 and C-111) were represented using MODFLOWs RIVER Package.

The CCS system was modeled as river cells. The head assigned to each river cell in the CCS were based on a north-south interpolation between five water level measurement stations.

The interceptor ditch and all secondary canals and drainage ditches were modeled using MODFLOWs DRAIN package. The river cell base and drain cell conductance were computed based on the length, width of the drain or river in the cell and hydraulic conductivities of the cell.

10

Attachment F 2.2 Model Calibration There are four stages of model calibration. The model calibration was first performed in June, 2016 (Tetra Tech, 2016). Then the model was calibrated using an automatic parameter estimation tool PEST (FPL, 2017). After one year of operation (from May 2018 to May 2019), the model was further calibration based on the latest data (FPL, 2019), which will be discussed in the next section in this report.

2.2.1 2016 Calibration Models The model went through several stages of model clibrations. The calibration model is subdivided into four timeframes, each of which simulates a particular set of hydrogeological conditions. The models developed for these four calibration timeframes:

(1) Pre-Development Steady-State Flow Model (Prior to 1940)

A steady-state flow model was used to define pre-development groundwater conditions. This model did not simulate salt and heat transport and only considered hydrologic processes which influenced local hydrology prior to human development: groundwater flow across the model domains perimeter, groundwater recharge, and groundwater discharge to Biscayne Bay. The resulting groundwater heads were used in conjunction with the Ghyben-Herzberg approximation (Bear, 1972) to estimate the shape and position of the freshwater-saltwater interface prior to human development for inclusion as an initial condition for the steady state model described below. Specification of the initial conditions using this methodology makes the modeling process more efficient than simply starting from an uninformed condition (Council and Richards, 2008).

(2) Steady-state Flow and Solute Transport Calibration Model (1940-1968)

The steady-state flow model calibration has 7 steady-state stress periods, for the time period from 1940 to 1968. The length of each of these 7 stress periods is 4 years. The time period was chosen based on the approximate start of groundwater development in the model area. All the stress periods were set as steady-state so aquifer storage was not considered.

The initial conditions of the steady-state model were derived from the model for pre-development conditions. The sea level for the Biscayne Bay was -0.71 ft NGVD.

The results, including simulated heads, salinity and temperature, were used as the initial conditions for the seasonal calibration model.

(3) Seasonal Calibration Model (1968-2010)

This model has 84 stress periods and each stress period lasts about half year long to represent the wet season (May-October)/dry season (November-April) hydrogeological features in southern Florida. The total simulation time is 42 years covering a time period from 1968 to 2010. The CCS was added to the model in May 1973 (the 10th stress period). The 2010 end-11

Attachment F date of this simulation was selected to coincide with beginning of the high-frequency data collection as a component if the Extended Power Uprate (EPU) monitoring program. The head, salt concentration and temperature results from the seasonal calibration model were used as the initial conditions.

(4) Monthly Calibration Model (2010 to 2018)

Initially the monthly calibration model covers the time period from Oct. 1, 2010 to March 31, 2015. It had a total of 63 monthly stress periods for a total simulation time of 5 years and 3 months. Late in 2019, the monthly model simulation time was extended to March 31, 2018, right before the RWS started operation. Simulated head, temperature, and salinity from the seasonal transient model were used as the initial conditions in the monthly stress-period transient flow and transport model. The results of this monthly calibration were used as the initial conditions for the prediction model.

2.2.2 Calibration parameters Calibration of the groundwater flow models included variations to both aquifer parameters and boundary conditions:

- River and drain conductance values;

- Heads and salinities in ungagged canals;

- Recharge rates;

- Hydraulic conductivities of High Flow Zones;

- Fort Thompson formation porosity;

- Longitudinal dispersivity.

2.2.3 Model Calibration with PEST The initial model calibration was performed by manually adjusting some of the parameters. After initial manual model calibration as reported in June 2016 (Tetra Tech 2016a),

the model was further calibrated against field data collected by JLA (2016) and CSEM data collected by Encore (2016a; 2016b). In this effort of model recalibration and revision, the popular automatic calibration tool, PEST (Doherty, 2010; Tetra Tech, 2016b) was applied to assist the process. The detail of this round model calibration is documented in FPLs Heterogeneous hydraulic conductivity analyses (FPL, 2017).

During the model recalibration with PEST, the hydraulic conductivity values at the 16 discrete locations (TPGW wells) were iteratively adjusted and re-interpolated in an effort to reduce the model error associated with groundwater levels and salt concentrations. Three separate PEST-based calibration efforts were conducted. In each of these calibrations, the pilot points hydraulic conductivities (in the High Flow Zones and deep model layers), the uniform 12

Attachment F horizontal hydraulic conductivities of other model layers, vertical hydraulic conductivities, and the cooling canal conductance were iteratively adjusted. Additionally, layer-wide porosities and dispersivity were adjusted (Tetra Tech, 2016b).

After the model recalibration with PEST, the representation of hydraulic conductivity became a more realistic spatially distributed aquifer parameter. As the results of model recalibration with PEST, the overall model error decreased from the 2016 groundwater flow and transport model. The normalized absolute mean error decreased for both the water level and salt concentration targets between 1972 and 2015. Plots of observed versus simulated water levels and relative salt concentration showed an overall good match to those observed conditions. The match to the CSEM survey-based salt concentration was reportedly improved, especially for the deep aquifer.

2.3 Prediction model The calibrated SEAWAT model was used to assess the effectiveness of a number of proposed remediation scenarios. After model calibration, seven remediation scenarios were designed and evaluated using the model. There were seven general scenarios, some of which had a number of configurations. A total number of 15 scenarios were evaluated based on evaluation rank matrix (Rose and Andersen, 2016), which includes drawdown, surface water seepages, wetland impacts, etc. Each scenario was a 10-yr simulation.

According to the technical memorandum (Tetra Tech, 2016), the hydrologic stresses and boundary conditions of these scenario models were derived from the time frame from 2011 to 2015 simulated in the monthly SEAWAT model and repeated once for the prediction simulations. The 2011-2015 timeframe experienced a reasonably wide-ranges of environmental conditions (dry and wet conditions).

ALT3D, among the 15 scenarios, was chosen as the basis for the RWS system design, based on the evaluation of a number of factors. According the performance Ranking Matrix (Tetra Tech, 2016), ALT3D had the highest performance score among 15 simulated remediation scenarios.

ALT3D involved one year of extraction from the base of the Biscayne aquifer beneath the CCS and adjacent to the Underground Injection Control (UIC) well followed by 9 years of pumping, at a total 15 mgd rate from 10 extraction or groundwater recovery wells tapping to the base of the Biscayne aquifer. These wells were spaced approximately 2000 ft apart at the base of the aquifer. The final RWS configuration is similar to ALT3D but the well locations and pumping schedule are slightly different.

2.4 Evaluation of 2016 Groundwater Flow and Salt Transport Models Overall, the models with different timeframes developed for model calibration are reasonably well built. The selection of model boundary conditions is appropriate. The input 13

Attachment F data are largely based on the available hydrogeological data. The model calibration over almost 75 years is rigorous and the match between simulated and observed the heads and relative salt concentration is in general agreement.

The matches between modeled and observed heads are reasonably good. The comparison of relative salinity is acceptable at most stations. However, large discrepancy can be seen at some of the stations. The reasons for the discrepancy seem to come from the initial conditions for solute transport when the errors were not fixed and carried into the model at next stage. The break through curves indicate the change of salinity was relatively small and the discrepancy between observed and measured concentrations showed little changes over time.

Comparison between the modeled position of the freshwater/saltwater interface, defined as the 1,000 mg/l isochlor, and the locations estimated by Parker et al., 1955 and Klein and Hull (1978) indicate the model is capable of simulating the historical extent of fresh/salt water interface.

The modeled hypersaline interface in the lower High Flow Zone was also compared to estimated Biscayne aquifer salt concentrations produced by a Continuous Source Electromagnetic Survey (CSEM) data (Enercon, 2016b). The modeled hypersaline interface is generally consistent with the CSEM-based estimated interface location. However, this survey also revealed that the most westward extent of the saltwater wedge west of the CCS is located in the lower High Flow Zone, not in the deep part of the aquifer as shown by the groundwater model.

14

Attachment F 3 Review of 2019 Groundwater Flow and Solute Transport Model (V3)

The groundwater flow and solute transport model, also referred as version V1, developed in 2016-2017 (Tetra Tech, 2016; FPL, 2017) went through two additional rounds of model calibration processes, based on one-year operation of the RWS from May 2018 to May 2019, as well as the newly available field data.

3.1 Revised Flow and Solute Transport Model (V3)

After the flow and solute transport model (V1) was calibrated and used to assess a number of scenarios of remediation in 2016-2017, it was updated and recalibrated to create the Version 2 (V2) model for apportionment of contributing factors on migration of saltwater in the region. Primary changes that were incorporated into the V2 model were:

(1) Data from Miami Dade County wellfields; (2) Land use time-series, including rock mines and quarries; (3) Separate simulation of recharge and evapotranspiration processes rather than a single net recharge term; (4) Detailed precipitation data; (5) Incorporation of canal and surface water methodology from the Hughes and White (2014) model; and (6) Incorporation of sea level rise.

The model was recalibrated using PEST (Doherty, 2010). The resulting V2 error statistics were well within criteria used to determine a good calibration was achieved. The V2 model was then used to simulate alternate historical scenarios and assess their individual and cumulative impacts on the saltwater interface.

3.2 V3 Model Calibration The model calibration involved (1) saltwater interface (2) heads; (3) the CSEM survey data; (4) salt extraction rates.

The re-calibration of V3 model involved adjustment of horizontal hydraulic conductivity, vertical hydraulic conductivity, and the cooling canal river conductance. During the calibration of V3 model, horizontal hydraulic conductivities in the Miami Oolite and Biscayne Bay, all of the horizontal conductivity pilot points in model layer 4, and most horizontal conductivity pilot points in model layers 9 and 10 were adjusted as a result of recalibration (FPL, 2019).

15

Attachment F The change of porosity during the V3 model calibration is relatively minor. The porosities increased in model layers 4 through 11, but the greatest change in porosity (model layers 5 through 7) only consisted of an increase from 0.248 to 0.309.

Multipliers applied to recharge, evapotranspiration rates, and canal bed leakance (including the cooling canals) differed by less than 20% from the initial V2 values.

One parameter value altered substantially during calibration was the reduction factor applied to the bed leakance of cooling canals (and, subsequently, conductance) after October 2013the initial value was 18 and the final value was approximately 58. The more than three-fold increase in this parameterby which the 1973-2013 cooling canal bed leakance is divided to generate the clogged canal bed leakanceresults in more than a 70% reduction in canal bed leakance relative to the V2 model (after October 2013).

The calibration statistics for water levels, relative salinities, and the CSEM surveys showed that all four seasonal and monthly water level and salinity target groups met the goal of 10% normalized mean absolute error (MAE).

The shape and position of the 1968 freshwater-saltwater interface simulated by the steady-state model at the base of the Biscayne aquifer generally matches the prior estimates from Parker et al. (1955) and Klein and Hull (1978).

Comparison of modeled groundwater levels over time at the monitoring wells used as the calibration targets showed the model results are in close agreement in simulating the changes of water levels at most of those monitoring wells.

Inspection of the plots of simulated salinity over time indicates that the modeled breakthough curves at most observation wells are in good agreement with observed data.

However, modeled breakthrough curves at some of the monitoring wells especially for the screens that open at the deep portion of the aquifer (e.g. TPGW-6D, TPGW-17D and TPGW-19D) are in significant discrepancy from observed. The model also under-simulated the breakthrough in the deep (58-ft) interval of well G-28.

The CSEM survey revealed that the most westward extent of the saltwater wedge west of the CCS was located in the lower High Flow Zone (model layer 8), not in the deep part of the aquifer. Comparison of the modeled hypersaline interface in the lower High Flow Zone to estimated Biscayne aquifer salt concentrations produced by CSEM survey (FPL, 2019) shows that the modeled hypersaline interface is generally consistent with the CSEM-based estimated interface location.

FPL indicated that there is a disagreement for the hypersaline plume extent in the deep portion of the aquifer between the model and the CSEM data. FPL showed that the plume extent determined based on the CSEM data is smaller than that predicted in the model in the deep portion of the aquifer (FPL, 2017; FPL, 2019). However, the comparison of the plume extents delineated by the CSEM survey and predicted by the groundwater model at TPGW-18 seems the disparity in March 2019 is not significant. Due to the lack of the Voxel grid data, a 16

Attachment F direct comparison of chloride concentration between the numerical model and the CSEM survey could not be made in this review.

It should be mentioned that the model prediction indicated that the remediation system may not meet the objectives set up by the MDC CA or FDEP CO, since only a portion the plume may be retracted back to FPLs property or to the east of the L-31E canal. The modeling results indicated the plume in the deep portion of the Biscayne aquifer would likely remain outside the western and northern CCS boundaries under current remediation action plan (FPL, 2019).

17

Attachment F

4. Review of Model Verification Performed by FPL The first year of remediation operation started in May 2018 and lasted for 13 months.

FPL performed a model verification study by comparing the model prediction to the actual data observed in the field for the 13-month time period from May 2018 to May 2019.

Following the MDC CA, FPL installed a remediation system that consists of 10 groundwater recovery wells (RWS) as extraction wells along the western side of the CCS and one deep injection well that used to discharge the extracted water to the Bolder zone, which is about 3,000 ft deep below the land surface. All these RWS extraction wells are tapping at the bottom of the Biscayne aquifer (corresponding to model layers 10 and 11). The deep injection well (UIC) was not simulated in the model.

After one year of operation of the RWS, the effectiveness of this remediation system was assessed based on the field measure water quality data, the salt removed from the aquifer and the CSEM survey data. At the same time, the flow and solute transport model was verified and revised based on the measured data during the first year of remedial operation.

Model verification results reviewed included:

(1) Modeled versus measured changes in salinity and water levels over the verification period; (2) Modeled versus estimated salt extracted over the verification period; (3) Change of the spatial extent of the hypersaline plume; and (4) Modeled concentration changes versus CSEM-based concentration changes between May 2018 and the end of Year 1 of the RWS operations (May 2019).

4.1 Evaluation Based on Water Quality and Water Level Data Water quality samples were taken during the time period from March 22, 2018 to March 20, 2019. The water quality data measured in March 2018 was used as the baseline condition. The performance of the RWS remediation system was evaluated by comparing the chloride concentration change over the one year period while the RWS was in operation.

FPL indicated (2019) that the evaluations of both analytic and automated monitoring data provide positive findings related to reductions in salinity in some of the hypersaline monitoring wells located west and north of the CCS after Year 1 of the RWS operations. Most significant reductions in the hypersaline plume appear to occur in areas north and northwest of the CCS.

FPL showed the comparison of water quality changes at some of shallow and middle monitoring well screens over the verification period (FPL, 2019: Figure 5.2-1). FPL (2019: p. 3-2) indicated there was a statistically significant declining trend and other factors showing notable decreases in chloride and/or salinity in five monitoring wells located west and north of the CCS during Year 1 of groundwater remediation. However, FPL did not provide supporting statistics as mentioned in their report.

18

Attachment F FPL did not include similar comparison graphics for the deep screens of these monitoring wells. Modeled and observed relative salinity data from the deep screens of five monitoring wells are shown in Figures 4-1 through 4-5. Data from TPGW-12D is not shown because this well is located near the coastal area. Modeled relative salinity data were extracted from the verification simulation. The data shown in these figures do not visually indicate a conclusive downward trend. Actually the data shown indicates the remediation at the deep portion of the aquifer is much less efficient.

Selected chloride concentration measured in March 2018 and March 2019 were shown in Table 4-1. A complete data set can be found in Table 3.3-1 of FPLs report (FPL 2019). The change over the year was computed and the trends were shown in the table as well and a trend analysis is shown in Table 4-2. To better see the trend visually, the uptrends (i.e. the chloride concentration increased) are marked as yellow and the downtrends (i.e. the chloride decreased) are marked as the green color, based on a rather arbitrary and semi-quantitative classification. If the change is less than 1%, it is assumed the change is insignificant and shown as no change.

As shown in Table 4-1 and Table 4-2, more downtrends appear in the shallow and middle sections of the monitoring wells while the water change from most deep well sections indicate an uptrend. This observation suggests the RWS may work successfully as expected between the shallow and middle portions of the aquifer but may not be as effective for the deep portion of the aquifer. This observation is further supported by the data of CSEM surveys and modeling results, which will be discussed later.

The water quality data were taken within a short time period (one year), so it is difficult to draw any long-term conclusions. It is possible that some of these trends may change in the future as the hypersaline plume shrink and retracts due to the RWS as mentioned in the FPL report (2019, p3-5).

FPL provided chloride concentration contour maps for the shallow, middle and deep portions of the aquifer, based on measured water quality data from March 22, 2018 (FPL 2019, Figures 3-2.2 - 3.2.4) (Figures 4-6, 4-7 and 4-8), respectively. Modeled extent of the hypersaline plume is shown in Figures 4-9, 4-10 and 4-11 for model layers 4, 8 and 11 respectively. The spatial extents of the hypersaline water plume shown in these contour maps seem to match the shape of the plume predicted by the flow and solute transport model.

It should be noted that it appears that the data of TPGW-18 were not used when the contour maps were generated. This monitoring well is particularly important because it is probably the only one monitoring well located within but nearby the edge of the plume (Figures 4-9 through 4-11). Based on the water quality data from this well, we may obtain critical information to show whether the plume is retreating eastward as expected especially in deep depths.

4.2 Comparison of Salt Extraction Rates Figure 4-12 shows the comparison of observed and modeled monthly salt removal rates of the RWS wells for the first year. As one step of verification, independently recalculated 19

Attachment F monthly salt extractions by this review is also shown in the figure. The modeled results were obtained by rerunning the FPL flow and solute transport model. The model predicted salt removal rates appear to generally match the measured monthly total salt extracted by the RWS wells, but always lower than the actual amounts based on the pumping rates and salt concentration values.

As also shown in Figure 4-13, the efficiency of those RWS wells is different. Among the 10 groundwater recovery wells, it appears that the efficiency of RWS 1 is the lowest likely due to its location. RWS 1 is the northern most extraction well so it may withdraw water with less salinity from its northern direction. The results shown also suggest that the pumping rates among the RWS wells may be adjusted so the remediation system can be optimized.

Figure 4-14 shows the cumulative total salt extraction by the recovery wells. At the end of first year of operation, a total of 1.84 billion pounds of salt has been expectedly removed from the aquifer based on the model prediction. The actual amount of salt extracted is 2.18 billion pounds, so the model prediction is about 10% less than the actual amount.

4.3 Modeled Extent of the Hypersaline Plume The change in spatial extent of the hypersaline plume is evaluated. The spatial extents of the hypersaline water plume, defined as the relative salt concentration of 1.0, in each of the modeler layers in April 2018 and May 2019 were compared. Figures 4-15 through 4-19 show the comparisons of some of the model layers 1, 4, 8, 9 and 11 respectively. As shown in the figures, some minor retractions of the plume can be observed in shallow depths but the changes in the extent of the plume are insignificant in deep layers as shown in Figures 4-18 and 4-19.

4.4 CSEM Survey Continuous Surface Electromagnetic (CSEM) is an aerial electromagnetic survey that produces a three-dimensional (3D) map of the hypersaline plume based on derived chloride concentration of the aquifer (Enercon, 2016a). Two rounds of CSEM surveys were conducted using the airborne transient electromagnetic (TEM) methods. The first CSEM survey was conducted in early April 2018, as the baseline survey to map the hypersaline plume west and north of the FPL property adjacent to Turkey Point. The second CSEM survey was conducted in late May 2019, 30 days after the first-year remediation operation. The CSEM data from these two rounds of surveys provided one means of verification of the groundwater flow and solute transport model and a progress check of the remedial action plan. It should be noted that the survey covers the area to the western portion of the CCS and some of areas outside the CCS, as shown in Figure 4-20.

The geophysical data were collected using transient electromagnetic (TEM) sounding equipment suspended from an airborne platform flying along individual transect lines over the target area. The CSEM survey measures the bulk resistivity of the ground. For water-saturated materials, bulk resistivity or its inverse bulk conductivity, is principally determined by pore fluid conductivity and porosity. When the pore water chloride ion content is high, bulk conductivity and fluid conductivity have a nearly 1:1 relationship. This allows the measurement of fluid 20

Attachment F conductivity or chloride content from bulk resistivity or conductivity values from the CSEM survey.

Borehole induction logs were conducted at some of TPGW wells located within the CSEM survey area. Quarterly water data from the TPGW monitoring wells were used to develop the relationship for conversion of CSEM resistivity to equivalent groundwater chloride concentration or chlorinity. A regression equation was established by fitting the CSEM resistivity data and measured groundwater samples resistivity:

Water Resistivity = 0.0843 (CSEM Resistivity) 1.1784 (2)

The second regression equation was derived based on the best fitting of the water resistivity and chloride concentration:

Chloride concentration = 2282 (Water Resistivity) -1.3019 (3)

Due to unavailability of the voxel chloride concentration data converted from the CSEM survey at the time of this review, this review of the CSEM data was performed based on visual estimates of concentrations from the maps, showing the extents of hypersaline water plume and changes in the numbers of voxel grid cells (chloride concentration equal to or greater than 19,000 mg/l) before and after the 1 year RWS operation provided by FPL (2019).

The CSEM-derived chlorinity values were interpolated to a 3-D chloride Voxel grid within the survey area as shown in Figure 4-20 which has a uniform horizontal dimension of 100 m by 100 m. The Voxel grid has 14 flat layers and the thickness increases with depth from 1 m for the top layer to 3.9 m for the bottom layer (FPL, 2019: Figure 4.2-1). FPL counted the numbers of the Voxel grid whose chloride concentration is equal to or exceeding 19,000 mg/l and compared the change of the numbers at the beginning and after 1 year of operation. Chloride concentration or porosity was not considered in the volume computation.

Based on the results (as shown in Table 4-3), it seems that the hypersalinity in every CSEM layer (except for Voxel grid layer 4) decreased after one year of RWS operations as compared against 2018 baseline condition. FPL (2019) concluded that the CSEM survey data indicated a 22% reduction in the volumetric extent of the hypersaline groundwater plume west and north of the CCS after one year of the RWS operation. It should be noted that the 22%

refers to the change of the volumetric change of the qualified Voxel grids not the mass reduction of the plume.

Evaluation of the changes of Voxel grid cells from March 2018 to March 2019 reveals:

(1) All layers below Layer 4 show a reduction of the plume volume; There is no explanation to why the CSEM survey indicated an unintuitive increase of chloride concentration in Voxel grid layer 4, corresponding to a depth interval between 10.8 ft to 15.4 ft below the land surface.

21

Attachment F (2) Significant reduction occurred at the middle section of the aquifer; For instance, Layers 8 and 9 only had a reduction of 39% and 42% respectively in the numbers of Voxel grid whose chloride concentration exceeding 19,000 mg/l. A 40% reduction of the plume in one year is very significant.

(3) The bottom Voxel grid layers (12, 13 and 14) had much less significant reduction: For example, Voxel grid layers 13 and 14 only had a reduction of 5% and 6% respectively in the numbers of Voxel grid whose chloride concentration exceeding 19,000 mg/l.

A close comparison of the CSEM survey data from 2018 and 2019 along survey line L105301 (FPL, 2019: Appendix E and Appendix F), as shown in Figure 4-21, indicates the plume may be expanding slightly at the aquifer bottom, indicated by the increase of the red-colored cells along this survey line. These three layers (12 through 14) count for about 35% of the total thickness of the Voxel grid (as shown in Table 4-3).

FPL compared the plume extent defined by 2018 CSEM survey and 2019 CSEM survey.

Figure 4-22 shows their comparisons for the deep portion of the aquifer (Voxel layers 12, 13 and 14). It seems that the plume was retracted eastward in some of the areas (southwest of TPGW-18) in Voxel grid layer 12. The change of the plume extent is much smaller in other areas in Voxel layer 12. A noticeable reversed trend, indicating the plume was actually expanding, can be also observed in many areas in Voxel grid layers 13 and 14.

Based on the CSEM survey data of 2019, the western edge of the plume in Voxel grid layer 14 is about 1,500 m (or 4800 ft) away from the CCS. A 5% plume reduction as observed in the first year operation may not be sufficient to retract the plume to the CCS as require.

It should be noted that the changes shown in Figure 4-22 were just for 1 year between 2018 and 2019. It may need longer time to see more meaningful trends and whether the remediation objectives will be met for the deep portion of the aquifer.

The location of TPGW-18 offers a unique opportunity to verify the Voxel grid and the numerical model. Figure 4-23 shows the chloride concentration of 19,000 mg/l (a) in the Voxel grid layer 12 (FPL 2019: Figure 4.3-8) and (b) the modeled extent of the hypersaline plume (equivalent to chloride concentration of 19,000mg/l) in model layer 10. The Voxel grid layer 12 (depth interval from 64.6 to 75.1 ft) and model layer 10 are selected because they are comparable to the screen of TPGW-18D, which is open between 72 to 79 feet below the land surface. It seems CSEM survey data and modeled salinity are in a general agreement at this location.

Figure 4-24 shows the 2019 CSEM survey data along survey line L102301 (FPL 2019:

Appendix F). Based on the color scales of derived chloride concentration, the chloride concentration is about 20,000 mg/l +/- 2,200 mg/l (assuming 11% of error as reported by FPL) at the time of the survey. At the same time, as shown in Figure 4-5, modeled chloride concentration at this location (TPGW-18D) is 26,730 mg/l while field measured chloride concentration is 24,760 mg/l. The model overestimated the chloride concentration by about 8%

while the 2019 CSEM survey under-estimated the concentration by 10.33%, assuming CSEM 22

Attachment F derived chloride concentration is 22,200 mg/l (20,000+2200) at this location. Inspection of Figure 4-24 also indicates that the chloride concentration based on the CSEM data in the proximity of TPGW-18D is lower (less than 19,000 mg/l), although Figure 4-23(a) indicates this well is located within the plume. Therefore, the CSEM survey data may not be necessarily more accurate than the models prediction at this location.

The location of TPGW-18 is about 4,400 ft from L-31E. Because this morning well (TPGW-18D) is located within the extent but nearby the edge of the plume, it will be interesting to monitor the water quality change at this location. Not only will it provide a reality check to both the CSEM survey data and the numerical model, it will also be an early warning if the remediation system might fail.

Considering the extent of the plume and the progress shown in Figure 4-22 and Table 4-3, which all are based on the CSEM survey data, it is difficult to conclude with confidence that the plume currently in the deep portion of the aquifer would be retracted back to the FPLs property boundary by 2028 based on the model prediction or the CSEM data collected in 2018 and 2019.

It is interesting to compare the volume changes based on the numerical model to the CSEM survey data, although it should be mentioned that FPL suggests the modeled chloride concentration may not be as reliable as the CSEM data (FPL 2019). Table 4-4 is a summary of the numbers of model cells which has a relative salinity value equal to or great than 1 from 2018 to 2019, based on FPLs verification model. In order to compare to Table 4-3, which was based on the CSEM data in the Voxel grid, the number of the models that has relative salinity equal or greater than 1 was counted layer by layer within the CSEM survey area in the same way as presented by FPL in their report (FPL, 2019).

The comparison of Table 4-3 (based on the CSEM data) and Table 4-4 (based on the groundwater modeling results) indicates a similar pattern in terms of the volume change of the plume. As shown in Table 4-4, the greatest plume reduction, about 40%, appears in model layers 4 and 5. This is similar to what was shown in Table 4-3. Towards to the deep portion of the aquifer, the modeling results indicate a reversed trend (i.e. the number of hypersaline model cells increased in model layers 10 and 11 from 2018 to 2019) while the CSEM data showed a small reduction (about 5%) in Voxel grid layers 13 and 14. The reason for the increase of hypersaline cells might be caused by the hypersaline water migrating from the center of the plume located beneath the CCS and/or the gravitational settling down of the hypersaline water from the layers above. This possibility is actually supported by the evidence shown in Figure 4-21.

Based on the changes in the number of cells, either from the groundwater modeling or the CSEM Voxel grid, it seems that the current RWS is able to capture and retract the plume in the shallow and middle portions of the aquifer, but the RWS has very limited impacts to the plume currently at the deep portion of the aquifer.

23

Attachment F 5 Review of Remediation Prediction The groundwater flow and solute transport model (V3) was used by FPL to predict the cleanup of the hypersaline plume in the Biscayne aquifer after the model calibration and verification processes. The simulation time frame is 9 years, from June 2019 to April 2028 for model prediction.

The RWS was assumed to operate according to the design: 1.5 mgd from each of the ten RWS wells, for a total of 15 mgd withdrawal from model layers 10 and 11, at the bottom of the Biscayne aquifer. The total pumping rate remains unchanged for the entire time period.

The initial conditions (heads, salinity and temperature) for the simulation were derived from the results of model verification run (i.e. the first year of RWS operation). Climate conditions (precipitation, evaporation, canal stages, etc.) for 2018-2019, representing a year of relatively stable conditions applied in the verification, were applied repeatedly nine times.

The projected results of operation of the RWS during the 9 years, from 2019 to 2028, were simulated in a similar fashion as the operations of the RWS during the first year as discussed in the model verification in this report.

The prediction model was verified by rerunning the model files provided by FPL. The assumptions for the boundary conditions that was used in prediction seem to be reasonable.

5.1 Salt Extraction The salt extraction rates form each of the RWS wells predicted by the prediction model are shown in Figure 5-1. As shown, the RWS-1 seems to have the lowest efficiency in salt extraction, probably due to its location as the northern most extraction well in the system. The extraction efficiencies for RWS-3 and RWS -2 are better than RWS-1 but still below most of the other extraction wells.

The monthly cumulative salt extraction from all the RWS wells is shown in Figure 5-2. As expected, the extraction rate will gradually decline when the salt concentration in the aquifer decreases with time.

The total cumulative salt extraction from all the RWS wells is shown in the Figure 5-3.

The model result indicates that a total of approximate 14 billion pounds of salt will be expectedly extracted from the aquifer by the RWS wells over the 9-year time period.

5.2 Retraction of the Hypersaline Plume Figures 5-4 through 5-8 show the plume extents after 0, 3, 5, and 10 years of remediation in model layers 1, 4, 8, 9, and 11 respectively. Year 0 refers to the beginning of the RWS operation, i.e. March 2018. It should note that Year 10 refers to the end of 9 year model prediction run since the first year was already simulated in the verification model, and the 24

Attachment F results (heads, salt concentration and temperature) were used as the initial conditions for the prediction model. The results of model predictive simulation indicate the hypersaline plume, defined as relative salinity of 1.0, in the Biscayne aquifer stops westward migration west of the CCS after Year 3 in layers 1 through 8. These figures indicate that the hypersaline water plume will be retracted eastward to the FPLs property boundary west of the CCS up to model layer 8.

The model predicts the northward migration will be halted in 3 years and the plume up to model layer 5 will be reacted to the FPLs property line north of the CCS.

However, the results shown in Figures 5-7 and Figure 5-8 also indicate there is little change of the plume extent in the deep portion of the aquifer west of the CCS. Based on the top elevation of model layer 9 at TPGW-2 (about 56 ft below the land surface), so the plume in about one third to one half of the aquifer thickness would not be cleaned to meet the remediation objectives specified per the MDC CA.

FPL argued that the deep model layers (9-11) may not belong to the Biscayne aquifer based on the hydraulic conductivity. FPL (2019, p.5-7) suggested the minimum hydraulic conductivity for classification as the Biscayne Aquifer (1,000 ft/day). This review does not agree with FPLs opinion regarding the definition of the Biscayne aquifer. Fish and Stewart state The Biscayne aquifer, as used in this report, is defined as the part of surficial aquifer system in southeastern Florida composed of (from the land surface downward) the Pamlico Sand, Miami Oolite, Anastasia Formation, Key Largo Limestone, and Fort Thompson Formation (all of Pleistocene age), and continuous, highly permeable beds of the Tamiami Formation of Pliocene and late Miocene age where at least 10 ft of the section is highly permeable (a horizontal hydraulic conductivity of about 1,000 ft/day or more). (Fish and Stewart, 1991, p 12). In reviewers opinion, A horizontal hydraulic conductivity of 1,000 ft/day is clearly used in the context by Fish and Stewart to define what is a highly permeable section rather than what is the Biscayne aquifer.

Fish and Stewart showed the USGS monitoring well (G-3321: ID:252506080212801),

located nearby L-31E between RWS-3 and RWS-4, indicates the base of the Biscayne aquifer is about 106 ft below the sea level (Fish and Stewart, 1991: Figure 16 and Cross-section E-E). This is consistent with the bottom elevation of FPLs groundwater model at this location (approximately -101 ft).

Tetra Tech (2016a), the groundwater model developer, reported that whereas Alternative 3D is effective at removing mass and pulling the hypersaline plume eastward in much of the aquifer, the impact of this alternative on the deepest part (model layers 10 and 11) of the Biscayne aquifer is much more subtle. FPL also states The Biscayne aquifer was divided into 11 layers in the numerical model. (FPL 2019: p. G3-5).

It should note that, as shown in the figures, the extent of hypersaline water plume retracts back to the CCS boundary only in model layers 1 through 5 north of the CCS, while MDC CA requires FPL to demonstrate a statistically valid reduction in the salt mass and volumetric extent of hypersaline water (as represented by chloride concentrations above 19,000 mg/l) in groundwater west and north of FPLs property without creating adverse environmental 25

Attachment F impacts. According to the MDC CA, the objective of this remediation action is to retract the hypersaline plume back to the FPLs property line in 10 years without creating any adverse impacts, whether or not the plume is within the Biscayne aquifer or within the flow high permeable zone.

FPL also argued that the modeled plume is contradicted by the 2019 Year 1 CSEM survey result that shows net reductions in the volume of the plume in the equivalent layers to model layers 9, 10, and 11. As discussed in previous section of this report, the CSEM survey data showed only 5% to 6% reduction of the plume in at the bottom of the aquifer (Voxel grid layers 13 and 14) after the 1st year, so it would be unusual to expect the plume completely retracted back to the CCS boundary according to this slow reduction rate. Based on the reviewers experiences, it is common that an extraction system in a groundwater remediation site, like the one implemented by FPL at this site, will have a higher efficiency at its early stage of operation because the solute concentration is higher.

The discrepancy in the plume extents between the CSEM survey and the groundwater model should be resolved. The CSEM data has been used to model calibration (Tetra Tech, 2016a). It is obvious that the initial salt concentration used in the numerical model should be revised based on the measured water equality data and CSEM survey data if the CSEM data is proven to be a better representation of the water quality in the deep portion of the aquifer.

Since the remediation system design and prediction are basically relying on numerical models, this review believes that it is important to determine the reason for the discrepancy between the CSEM data and modeling results, and fix the problems in either CSEM inversion or groundwater flow and solute transport model or the both. Longer term water quality monitoring data from the deep well screens may be helpful.

5.3 Capture Zone Analysis FPL has performed a forward particle tracking analysis to delineate the capture zone of the remediation system (FPL, 2019). FPL performed a capture zone analysis based on particle tracking method with the prediction model and FPL showed that the RWS wells prevents water from the CCS from moving westward through the RWS wells as represented by the coalescing of capture radii between the recovery wells (2019).

A backward particle tracking analysis was performed in this review using the USGS code MODPATH (Pollock, 1996). Twenty particles were placed around the each of the RWS wells in both layer 10 and layer 11. The particle tracking pathlines were shown in Figure 5-9. As shown in the figure, most of the water captured by the wells comes from the CCS area.

Capture zone analysis reveals that the capture zone of the RWS wells are relatively limited to the CCS area and some of the water coming from the cooling canals. The horizontal extents of the capture zone are limited to the close proximity of the extraction wells in model layers 10 and 11, likely due to the nature of relatively low hydraulic conductivity present in these layers. This design seems to be capable of catching the hypersaline plume in the proximity of the wells in the shallow and middle portions of the aquifer but not sufficient to 26

Attachment F capture the entire plume, especially in the deep portion of the aquifer. This may also explain why the remediation efficiency is low in deep layers.

The capture zone analysis conducted by this review supports the findings of FPL (2019) that it is clear that there is an extensive capture zone to the east and west in layer 8, despite this layer not being pumped. In contrast, separate capture zones surround each of the RWS wells in layer 10, indicating these wells obtain only a small portion of their water laterally from layer 10. This conclusion also leads to the concern regarding whether the remediation system is capable of meeting to the objectives specified in the MDC CA.

Based on the groundwater modeling results, it appears that the current RWS wells are effective only to shallow and middle portions of the Biscayne aquifer but these wells do not have much impacts to the hypersaline plume sitting in the deep portion of the aquifer.

Therefore, it would be difficult to foresee if the current RSWS will be capable of retracting the hypersaline plume back to FPLs property boundary in the deep portion of the aquifer within 10 years.

5.4 Water Levels and Drawdown in the Surficial Aquifer Figure 5-10 shows modeled water levels (ft, NGVD) in the surficial aquifer (model layer

1) at the end of 10 years of remediation (April 2028).

Model predicted drawdown at the surficial aquifer (model layer 1) is computed as:

Drawdown = h0-h1 (4) where h0 is the heads predicted at the end of first stress period in the verification model, which represents the initial heads prior to the operation of RWS wells; h1 is the head at the end of the 10 year remediation period, extracted from the last stress period of the prediction model results.

The results shown in Figure 5-11 indicate that the adverse hydraulic impact from the remediation is insignificant to the area. This is expected because all the RWS wells pump from the aquifer bottom in model layers 10 and 11. The canal system existing at the site area will further reduce the impact from the RWS well operation.

5.5 Comments on the FPLs Recommendations for Model Improvements FPL made three recommendations for the improvements of groundwater model that was used as a predicting tool (FPL 2019):

(1) The hydraulic conductivity of the lower model layers: This review does not think the problem is that the calibrated hydraulic conductivities of the lower model layers are so low that they do not qualify these layers to be part of the Biscayne aquifer as stated by FPL (FPL, 2019: p. 5-9). A more important issue is that if the hydraulic conductivities applied in the model are real and representative to the field 27

Attachment F conditions whether or not these layers should be called a part of the Biscayne aquifer. Model calibration should be based on the available field data and adjustment of the input parameters should be made within physically reasonable ranges. It is true that any field data could have errors, so sensitivity analysis and uncertainty analysis could be performed to establish and improve the confidence levels towards the modeling predictions.

The input of both horizontal and vertical hydraulic conductivities for the deep model layers should be revisited and more aquifer performance tests may be necessary.

FPL indicated that the RWS wells were turned off and on twice (from Dec. 28, 2018 to Jan. 18, 2019 and from Feb. 28, 2019 to Mar. 14, 2019) as the RWS drawdown assessment (FPL, 2019). The drawdown data collected during the shut-off evens might be applicable as aquifer performance tests and model calibration, particularly for the deep portion of the aquifer.

(2) The model indicates a high connection of RWS extraction to the CCS resulting from the estimated high vertical conductivity of the dredged CCS canals.

The close match between modeled and measured the salt removal rates during the first year operation may suggest the mass contribution from the CCS was modeled reasonably. The canal bottom hydraulics can be checked by the RWS shut-off tests mentioned above as well as a part of model calibration. If there is high connection of RWS and the CCS canals, the salt concentration specified in the canals should be checked against field conditions.

(3) A dry climate may affect the capture zones:

The sensitivity (SEN 2) seems to support FPLs conclusion. This review agrees in general with FPLs suggestion that the CCS salinity and climate conditions should continue to be monitored. This review does not expect the climatic conditions should significantly affect the effectiveness of the RWS.

(4) Besides the three suggestions of future model improvement from FPL, this review believes that the initial salt concentration might need to be revisited.

Firstly, the initial salt concentration at locations where the water quality data were collected should be adjusted based on the field data so the error from the previous simulation would not be carried into the following step simulation.

Second, the initial extents of the hypersaline plume: FPL suggested the plume extents in deep layers were overestimated in the model, compared to the CSEM data. If the CSEM data are validated, the initial salt concentration in the model should updated by the CSEM data wherever the CSEM data available.

28

Attachment F 6 Sensitivity Analysis Sensitivity analysis of model input parameters is performed by varying the input parameters from its calibrated values (Anderson et al., 2015) as a way to assess the uncertainty of modeling results.

Two sensitivity simulations were made by FPL to test the effectiveness of some of the assumptions made in the prediction model or a proposed extraction well at the UIC location (FPL, 2019). Additional sensitivity runs were performed by GTI in this review to further assess the sensitivity of some other model input parameters pertaining to the salt transport.

6.1 FPL Sensitivity Simulations 6.1.1 FPL Sensitivity Simulations 1 (SEN1)

In the prediction model discussed earlier in this report, the boundary conditions were set up based on the climate conditions of 2018-2019 which turned out to be a relatively dry year. In the prediction discussed in the previous section, the climate conditions for 2012-2013 were cycled for 9 years. Since the climate conditions from 2019 to 2028 are basically unknown, a wetter climate condition may be tested out to see the response of the RWS system. The first sensitivity simulation uses the same configuration as the predictive simulation, except it uses precipitation, evapotranspiration, canal stages, and the CCS stage from the 2012-2013 timeframe.

Model predicted water levels in the surficial aquifer (Model layer 1) under the climate conditions of 2012-2013 are shown in Figure 6-1. As expected, the modeled water levels are higher due to a wetter condition. And some of the model areas are flooded by more than a half foot of water.

The modeled drawdown under the wetter climate is shown in Figure 6-2. As expected, higher water table rebound is shown due to high groundwater recharge.

Predicted salt concentrations at selected model layers (1, 4, 8 and 11) are shown in Figures 6-3 through 6-6, respectively. They appear to be similar to those shown in previous section under the prediction run. However, it should be noted that one small area located west of the CCS area in Layer 8 (Figure 6-5) has predicted relative salt concentration greater than 1.0.

This is likely caused by the higher groundwater recharge from the wetter climate condition so the RWS wells receive more water from the shallow portion of the aquifer.

Overall, the impact of wetter conditions to the model prediction is insignificant.

Comparison of cumulative salt extraction in 9 years based on this sensitivity run to the cumulative salt extraction predicted in the baseline simulation run indicates the total cumulative salt extraction under wetter climate condition is only slightly lower than the run under 2018-2019 conditions (Figure 6-7).

29

Attachment F 6.1.2 FPL Sensitivity Simulation 2 (SEN2)

The second sensitivity simulation that FPL conducted was to assess whether pumping additional 3 mgd from beneath the CCS could accelerate or improve the attainment of MDC CA objectives. In this sensitivity simulation that FPL has performed, the UIC well, designed for discharging water from other 10 extraction wells, was assumed to pump at 3 mgd from model layers 10 and 11 beneath the CCS in combination with RWS operation at 15 mgd for a 8.5 year duration, from the seventh month of the 1st year to the end of the 9th year. Unlike other 10 extraction wells, where the pumping rates were proportionally distributed based on the layer transmissivity, the pumping rate of 1.5 mgd was equally split between model layers 10 and model 11 for the proposed extraction well located at the UIC.

Modeled water levels (ft, NGVD) and drawdown (ft) in the surficial aquifer are shown in Figures 6-8 and 6-9, respectively. They are similar to the results from the prediction run.

Figures 6 6.13 show the extents of the hydpersaline pume in model layers 1, 4, 8 and 11, respectively, after 9 years of operation (equivalent to Year 10) based on the second sensitivity simulation run. They appear to be comparable to those from the simulation under baseline conditions.

It should be noted that there is a small area in the surficial aquifer (Layer 1), just west to the CCS as shown in Figure 6-10, where the relative salt concentration is as high as 1.3. It is not clear what caused this concentration anomaly in the surficial aquifer. Figure 6-11 shows modeled the special extent of the hypersaline plume, represented by the relative salinity equal to 1.0 contour in the higher High Flow Zone (model layer 4). Figure 6-12 shows the spatial extent of modeled hypersaline plume in the lower High-Flow-Zone (model layer 8). Figure 6-13 shows the modeled spatial extent of the hypersaline plume at the aquifer bottom (Layer 11).

Comparison of cumulative salt extraction in 9 years based on this sensitivity run to the cumulative salt extraction predicted in the baseline simulation run indicates the total cumulative salt extraction under wetter climate condition is only slightly lower than the run under 2018-2019 conditions (Figure 6-14). It should be noted that the salt extracted by the proposed UIC well is not included in the cumulative salt extraction shown in Figure 6-14.

The results of the simulation are not significantly different from the baseline condition (as shown above). FPL concluded the results were likely for two reasons: (1) the additional withdrawal at the UIC is not located in the most concentrated part of the hypersaline plume beneath the CCS; and (2) the removal of hypersaline water from beneath the CCS accelerates more leakage from the CCS.

6.2 Additional GTI Sensitivity Runs Sensitivity analysis is often used to test the sensitivity of model prediction to the input parameters. A number of the input parameters pertain to solute transport simulation, such as dispersivity, used in the models are not actually based on field measurements but estimated from literature review and determined by model calibration. Sensitivity analysis is helpful in identifying the sensitive input parameters and assessing the uncertainty of the model.

30

Attachment F 6.2.1 Courant numbers Courant number is a dimensionless number often used in solute transport modeling. It is used to define the ratio between numerical dispersion and physical dispersion. For the finite-difference scheme, increasing the Courant number also leads to an increase in numerical dispersion. The higher the Courant number, the higher ratio of numerical dispersion to the physical dispersion (Zheng and Bennett, 1995).

FPL used a Courant number equal to 1 in the steady state calibration and increased the value to 10 in all of the transient models, including the verification model and prediction models. No discussion was provided for the increase. Two additional prediction simulation runs were performed, one run with Courant number of 1 and the other run with Courant number of 20, to test the sensitivity of model predictions to this model parameter.

Model-predicted total cumulative salt extraction amounts from these two sensitivity analysis runs are compared to the results from the baseline run with Courant number equal to 10, as shown in Figure 6-15. The results shown in Figure 6-15 indicate the model results are not very sensitive to the Courant number used in the model.

6.2.2 Sensitivity of Dispersivities Longitudinal dispersivity (aL) is one of the input parameters for solute transport. It is a property of the porous medium describing dispersive transport in the flow direction.

Transverse dispersivity (aT) is a similar property of the porous medium that describes the dispersive transport normal to flow direction. Vertical dispersivity (aV) is also a property of porous medium describing the dispersive transport in the vertical direction.

Although dispersivities are the property of porous media, it is difficult and expensive to measure them in the field. Common practice is to estimate them based on literature and model calibration. FPL derived the value of longitudinal dipsersivity and used it in all the models.

Values of transverse and vertical dispersivity ratios appeared to be selected as commonly used the solute transport modeling as recommended in MT3DMS (Zheng and Wang 1998). These ratios were not calibrated and their sensitivities to the modeling prediction were not tested.

FPL used a longitudinal dispersivity value (aL) of 11.2777 ft in all their models. Two additional prediction simulation runs were made, one run with longitudinal dispersivity of 22.5554 ft and the other run with a longitudinal dispersivity of 5.6389 ft, to test the sensitivity of model predictions to this model parameter.

Parameters aT and aV are defined as the ratio of transverse dispersivity and vertical dispersivity to the longitudinal dispersivity, respectively. The values that used in the baseline model and upper and lower ranges tested are shown in Table 6-1.

The results, modeled total cumulative salt extraction over 9 years from 2019 to 2018, from these sensitivity simulation runs are shown in Figures 6-16 through 6-18 respectively.

31

Attachment F 6.3 Summary of Sensitivity Analysis Simulations Figure 6-19 shows the range of model predictions with consideration of all the sensitivity analysis runs. The least salt extraction efficient scenario appears in SEN1. The most predicted salt extraction scenario occurs when the Courant number is lowered from 10 to 1.

To assess the sensitivity of these parameters, cumulative salt extraction is used as an indicator. Table 6-2 shows the modeled cumulative salt extraction resulted from each of the sensitivity analysis runs from 2019 to 2028. The last row in this table is a comparison, as a percentage of changes, of salt extraction from each scenario to the baseline run which is presented as FPLs original prediction model. The total salt extraction predicted by the baseline run is 13.994 billion lbs. and the least case is 13.672 billion lbs. (2.3% less) while the most case is 14.582 billion lbs. (4.21% more).

In general, the model prediction is consistent and is in a generally close consensus among all the sensitivity analysis runs. The modeling results appear not to be sensitive to the input parameters tested in this this review.

32

Attachment F 7 Evaluation of FPL Conclusions and Recommendations Below are the major findings presented by FPL in its one year remediation operation report (FPL 2019):

1. The RWS remediation objectives are being met; GTI Comment: Based on the information summarized below,this review does not agree with FPLs statement The RWS remediation objectives are being met. The groundwater flow and solute transport model has gone through a number of model calibration and verification processes. The model predicts that hypersaline water plume will be retracted back to the FPLs property line west of the CCS from the plumes current position from in the shallow and middle portions of the Biscayne aquifer (model layers 1 through 8) under the existing RWS. To north of the CCS, the model predicts the plume will track to FPLs property line only in the shallow portion of the aquifer (model layers 1 through 5).

The MDC Consent Agreement defines the objective of the RWS system as to retract the hypersaline plume back to FPLs property lines, from both the west and north directions. The MDC CA does not specifically limit the remediation objectives to the shallow or middle portions of the aquifer so this review believes that the remediation objectives shall be applied to the entire hypersaline plume outside the CCS.

Therefore, according to the model predictions and the CSEM data analysis presented, the current RWS will unlikely fulfill the remediation objectives within the 10 years timeframe in pursuit of the MDC CA.

FPL has argued that the deep portion of the aquifer modeled (model layers 9-11) may not be a part of the Biscayne aquifer due to their relatively low hydraulic conductivity (less than 1,000 ft/day): The calibrated hydraulic conductivities of the lower model layers are so low that they do not qualify these layers to be part of the Biscayne Aquifer, according to the USGS definition of the Biscayne aquifer that the aquifer has a hydraulic conductivity of 1,000 ft./day or greater (Fish and Stewart, 1991)

(FPL 2019, p 5-9). This review does not agree with FPLs opinion regarding the definition of the Biscayne aquifer. Fish and Stewart state The Biscayne aquifer, as used in this report, is defined as the part of surficial aquifer system in southeastern Florida composed of (from the land surface downward) the Pamlico Sand, Miami Oolite, Anastasia Formation, Key Largo Limestone, and Fort Thompson Formation (all of Pleistocene age), and continuous, highly permeable beds of the Tamiami Formation of Pliocene and late Miocene age where at least 10 ft of the section is highly permeable (a horizontal hydraulic conductivity of about 1,000 ft/day or more). (Fish and Stewart, 1991, p 12). In this reviewers opinion, A horizontal hydraulic conductivity of 1,000 ft/day is clearly used in the context by Fish and Stewart to define what is a highly permeable section rather than what is the Biscayne aquifer.

33

Attachment F Fish and Stewart showed the USGS monitoring well (G-3321:

ID:252506080212801), located nearby L-31E between RWS-3 and RWS-4, indicates the base of the Biscayne aquifer is about 106 ft below the sea level (Fish and Stewart, 1991:

Figure 16 and Cross-section E-E). This is consistent with the bottom elevation of FPLs groundwater model at this location (approximately -101 ft).

Tetra Tech (2016a), the groundwater model developer, reported that whereas Alternative 3D is effective at removing mass and pulling the hypersaline plume eastward in much of the aquifer, the impact of this alternative on the deepest part (model layers 10 and 11) of the Biscayne aquifer is much more subtle. FPL also states The Biscayne aquifer was divided into 11 layers in the numerical model. (FPL 2019: p.

G3-5).

FPL has also argued that the larger extent of the hypersaline water plume in deep model layers (9 to 11) as predicted by the model may not be real as contrary to the results of CSEM survey data. The CSEM survey data showed only 5% to 6% reduction of the plume in at the bottom of the aquifer (Voxel grid layers 13 and 14) after the 1st year, so it would be unusual to find the plume completely retracted back to the CCS boundary according to this slow reduction rate. Based on the reviewers experiences, it is common that in a groundwater remediation site, an extraction system, like the one implemented by FPL at this site, will have a higher efficiency at its early stage of operation because the solute concentration is higher.

PFLs statement regarding the CSEM data versus modeling results could be true and the initial salt concentration in the model might be incorrect. If this is the case, however, the models predictions become questionable. It is necessary to confirm whether the CSEM survey data, as an indirect measurement of salt concentration, or the model input data is a better representation of the field conditions. If the CSEM data is proven to be more reliable and the numerical model input (salt initial salt concentration) needs to be corrected, the model should be revised and recalibrated based on the CSEM data and more realistic predictions should be obtained from the revised model.

Although the hypersaline water plume in the shallow and middle portions of the aquifer is expectedly retracted back to FPLs property line west and the shallow portion of the plume north of the CCS in 10 years, the hypersaline plume left in the lower model layers (9-12) appears likely to remain as a source of pollution and the salt will likely diffuse back to the layers above once the RWS wells cease pumping due to the concentration gradient. Based on the reviewers experience, this is a common phenomenon observed in groundwater pump and treat systems used in contamination remediation sites (Fetter, 1992). Considering the fact that model predicted salt concentration in model layers 8 after 10 years remediation is barely below the target concentration (relative salt concentration =1.0), the hypersaline plume is likely to come 34

Attachment F back to model layer 8 unless the existing RWS wells operate continuously or occasionally or an on-and-off switch schedule (USEPA 1990).

2. During the first year of operations, the RWS removed approximately 5 billion gallons of hypersaline groundwater containing over 2 billion pounds of salt from the Biscayne Aquifer without creating adverse environmental impacts; GTI Comment: Concurred.

3. Analytical and automated groundwater monitoring data indicate a statistically significant trend showing notable decreases in chloride and/or salinity in multiple groundwater monitoring wells located west and north of the CCS; GTI Comment: FPL did not provide associated statistics to support the conclusion.

Without statistical support, FPLs conclusion is likely only applicable to the shallow and middle sections of the aquifer.

4. CSEM methodology has been statistically verified to assess the volumetric extent of the hypersaline groundwater plume with an accuracy of +/-2% and spatial extent of the 19,000mg/L isochlor in each survey layer to an accuracy of +/-11%;

GTI Comment: FPL did not provide associated statistics to support the conclusion, and the CSEM survey-derived Voxel grid chloride concentration data was not available at the time of this review. Therefore, the claimed accuracy could not be verified.

5. CSEM comparison to the 2018 baseline data depict a 22% +/- 2% reduction in volume of the aquifer occupied by the hypersaline plume. The CSEM survey also shows reductions in the plume along the base of the Biscayne Aquifer in areas where groundwater modeling predicted that little change would occur after 10 years of RWS extraction; GTI Comment: A reduction of 22% in one year was based on the change of the numbers of Voxel grids that had a derived chloride concentration more than 19,000 mg/l.

Although the value seems promising, it does not imply a same reduction rate should be applied to all depths. Actually, the lowest reduction rate in the plume should be considered in order to assess the effectiveness of this remediation system unless either areal extent and/or contaminant mass removed are the criteria. In addition, it might be more meaningful to compute the salt mass reduction in the plume with time.

The plume reduction along the base of Biscayne aquifer is not obvious or conclusive as shown in Figure 4-22. The CSEM survey shows the overall one-year reduction of the Voxel grids in Layer 13 and 14 is 5% and 6% respectively. However, these two layers account for approximately 24.4% of the total aquifer thickness.

6. Modeling depicts the westward expansion of the hypersaline plume will be halted within three years and completely retracted to the FPL property by year 10 in all but the bottom 35

Attachment F layers, which have model-simulated hydraulic conductivities indicating that they are not part of the Biscayne Aquifer; GTI Comment: The FPL statement is partially correct. See discussion in GTI Comment 1 above.

7. The updated calibrated and validated groundwater model continues to show challenges in lower layers 9 to 11, due to model-simulated hydraulic conductivities that are not indicative of the Biscayne Aquifer. Modeled salinity results for these lower layers in Year 1 do not match the measured salinity reductions in these same layers recorded by the CSEM survey. Continued updates and calibration of the model with Year 2 salinity data in the base of the aquifer are expected to improve the model predictions in the lower layers; GTI Comment: This review does not agree with most part of the FPLs statement for the same reasons discussed under Comment #1 in this section. FPL should explain why modeled salinity results for these lower layers in Year 1 from a calibrated and verified model did not match the CSEM data. If FPL is recommending that the model be updated with Year 2 salinity data, then FPL should demonstrate that the data that they will be using are real and representative.

8 Isochlor mapping based on monitoring well data alone represents high error potential and low accuracy and should not be used to assess orientation of the hypersaline groundwater. Isochlor mapping using CSEM data informed by groundwater monitoring data has a much higher spatial and volumetric accuracy and should be used exclusively for assessing progress in meeting remediation objectives. As a result, isochlor mapping based on monitoring data alone will be discontinued; GTI Comment: Water quality data from the monitoring wells may be sparse to generate a detailed and accurate isochlor map, but they are valuable to validate the accuracy of the CSEM survey and groundwater modeling. Correlations between CSEM resistivity and chloride concentration data obtained from wells completed at the various depths of the aquifer must be demonstrated through regression analysis or other acceptable statistical means. If valid data from CSEM surveys are used to prepare isochlor maps, the concentration values from monitoring wells should also be shown on future isochlor maps.

36

Attachment F 8 Summary, Conclusions and Recommendations 8.1 Summary Overall, it appears that a significant amount of work was done to develop these data-intensive, variable density models. The assumptions and approach used during model development seem to be reasonable. Results of sensitivity analyses show consistency in the modeling prediction for the model input parameters and boundary conditions tested. The model development followed the general standard procedures. The model calibration seems reasonable and adequate. The results of model calibration indicate that the model calculated water levels and salinity are in general agreement with the field data. Results of monthly salt extraction from the verification period matched the actual data closely. However, as noticed by FPL, there is a disagreement for the plume extent in the deep portion of the aquifer between the numerical model and the CSEM survey data.

CSEM survey data was used to delineate the hypersaline plume in the aquifer. FPL reported a 22% reduction of the plume volume in the verification period based on the change of Voxel grid cells. However, the retraction of the plume varied greatly with depths, from 40%

at the middle to 5% at the bottom of the aquifer.

This review does not agree with FPLs conclusion that predictive modeling indicates the required objectives with the MDC CA will be achieved. Groundwater modeling prediction shows the RWS remediation system is capable of retracting the plume to the CCS property up to model layer 8 (about 60 ft deep) west of the CCS and up to model layer 5 (about 35-40 feet) north of the CCS at the end of remediation period (2028). Similar modeling prediction was noticed previously by FPL (Tetra Tech, 2016a; FPL 2017; FPL 2019). It is questionable whether the current RWS remediation plan will be able to meet the objectives specified in the MDC CA:

to retract the hypersaline plume back to the FPLs property from both west and north directions of the CCS.

FPL argued that the actual spatial extent of the plume may be smaller than that predicted by the numerical model, based on CSEM survey data. The volumetric change of the plume based on the CSEM data also showed the change in the deep portion (Voxel grid Layers 13 and 14) of the aquifer was small (about -5%) for the first-year of remediation operation. In general, a groundwater remediation system works better at the beginning because of higher concentrations in the aquifer. Therefore, it is questionable the RWS will be able to meet the remediation goals on time as specified in the MDC CA.

FPL compared the plume extent defined by 2018 CSEM survey and 2019 CSEM survey.

Based on their analysis, it seems that the plume was retracted eastward in some of the areas (to southwest of TPGW-18) in the Voxel grid layer 12 (63 to 73 ft deep), but the change of the plume extent is much smaller in other areas in this layer. However, a reversed trend indicating the plume was actually expanding, could be observed in many areas in Voxel grid layer 13 and layer 14 as shown in Figure 4-23. It may need longer time to observe a more definite trend and whether the remediation objectives might be met for the deep portion of the Biscayne aquifer.

37

Attachment F Whether the CSEM survey data is more reliable than the numerical modeling needs to be further verified by the field water quality data. If the CSEM data is proven to better represent the hypersaline water plume, the numerical model should then be updated and revised so a more reliable prediction can be made.

Data from TPGW-18 is useful to validate both the CSEM survey data and the numerical model. Its unique location also provides an opportunity for early detection whether the plume is retracting eastward. The chloride concentration was 25,400 mg/l (about relative salinity 1.3) in March 2019. New water quality data should be available soon to allow a progress assessment of the remediation system after its two years of operation. Extra monitoring wells within the hypersaline plume might be necessary to detect and verify the progress of planned remedial actions.

8.2 Conclusions The following findings and conclusions based on results of the evaluations performed by GTI are summarized below:

(1) FPL developed site-specific variable density groundwater flow and solute transport model using USGS computer code SEAWAT. The three-dimensional model was well designed, calibrated with data from more than 70 years. The model was evaluated with the data collected during the first year operation of the RWS (May 2018 to May 2019). The model predicted salt extraction rates, water levels and salinity appeared to be in acceptable agreement with the data measured in the field. However, as noticed by FPL, there is a disagreement for the plume extent in the deep portion of the aquifer between the model and the CSEM data.

(2) Model predictions indicated the RWS would be capable of retracting the hypersaline plume up to model layer 8 (about 60 feet deep) to FPLs property west of the CCS and up to model layer 5 (about 35 ft deep) north of the CCS.

(3) The model predictions indicate that the existing RWS has little impact to the plume in the deep portion of the Biscayne aquifer (model layers 9 to 11).

(4) CSEM surveys were conducted to delineate the hypersaline plume in the aquifer in April 2018 and May 2019. FPL reported a 22% reduction of the plume volume based on the CSEM survey data. However, the reduction of the plume varied greatly at depths, from 40% plume of volume reduction at the middle to only 5%

to 6 % of plume volume reduction at the bottom of the aquifer.

(5) The CSEM Voxel grid analysis indicated the plume volume within the CSEM survey area was reduced in the shallow and middle portions of the aquifer within the first year of RWS operation. The plume volume analysis based on the numerical model also showed a similar pattern for the shallow and middle portions of the Biscayne aquifer, but indicated the plume in the lower portion of the aquifer may not be remediated as required.

38

Attachment F (6) Both modeling results and the CSEM survey data indicate the RWS is capable of preventing the hypersaline plume from its expansive migration in the Biscayne aquifer.

(7) Both groundwater modeling and the CSEM survey data indicate that the existing RWS has little remedial effect to the hypersaline plume currently in the deep portion of the Biscayne aquifer.

(8) Since the CSEM data cannot be used to predict how fast the hypersaline plume will be retracted, a well-calibrated and validated groundwater model is needed as a reliable prediction tool. It is strongly recommended to verify the chloride concentration interpreted from the CSEM survey using field measured data. The discrepancy between the CSEM data and modeling prediction in the deep portion of the aquifer should be resolved and the numerical model should be revised and recalibrated before the groundwater model can be used as a reliable predictive tool.

(9) Data from TPGW-18 is useful to validate both the CSEM survey data and the numerical model. Its unique location provides an opportunity for early detection whether the plume is retracting eastward.

(10) Extra monitoring well(s) within the hypersaline plume (for example, between TPGW-4 and TPGW-17) might be helpful to detect and verify the progress of planned remediation action.

(11) Based on the information reviewed and analyzed, the current RWS does not appear to be capable of meeting the remediation objectives of retracting the hypersaline plume to the FPLs property per the Miami-Dade County Consent Agreement and Florida Department of Environmental Protection.

8.3 Recommendations The following recommendations are based on the GTI evaluations described above:

(1) FPL should explain why modeled salinity results for these lower layers in Year 1 from a calibrated and verified model did not match the CSEM data. If the CSEM data is proven to be more reliable, the SEAWAT model should be revised, re-calibrated and re-verified as a predictive tool. And all model predictions presented in the first year remediation action annual report (FPL, 2019) should be updated based on the revised model.

(2) It is recommended to verify the chloride concentration interpreted from the CSEM survey using field measured data. The discrepancy between the CSEM data and model prediction in the deep portion of the aquifer should be resolved and the numerical model should be revised and recalibrated before using the model as a predictive tool.

39

Attachment F (3) FPL should explain why the plume volume in Voxel grid layer 4 indicated an unintuitive 140% increase over the first year of RWS operation based on the CSEM data so a higher level of confidence towards the CSEM data can be achieved.

(4) When the SEAWAT model is to be updated with Year 2 salinity data, FPL should demonstrate that the data that they will be using is reasonable and representative.

(5) It would help better understand the RWS performance if a layer-by-layer water source analysis can be performed for each of the RWS wells using a backward particle tracking method.

(6) It might be more meaningful to compute the salt mass reduction in the plume with time by factoring porosity and salt concentration. MDC CA actually requires FPL demonstrate a statistically valid reduction in the salt mass and volumetric extent of hypersaline water.

(7) An optimization analysis is recommended for the extraction wells by varying their pumping rates and schedules so the overall efficiency of the RWS can be maximized and the cost of remediation may be reduced.

(8) It is recommended to continuously monitor the water quality changes at TPGW-18 at its middle and deep screens. TPGW-18 may be useful to validate both the CSEM survey data and the numerical model. Its unique location also provides an opportunity for early detection whether the plume is retracting eastward. The chloride concentration was 25,400 mg/l (about relative salinity 1.3) in March 2019. New water quality data should be available soon to allow a progress assessment of the remediation system after its two years of operation.

(9) Extra monitoring wells open to the deep portion of the aquifer will be helpful and thus recommended to verify the progress of remediation, the CSEM data and numerical models. Suggested locations may be between TPGW-17 and TPGW-4, and TPGW-18 and RWS-5 if applicable.

(10) The drawdown data collected during the shut-off evens might be useful as aquifer performance tests and model calibration, particularly for the deep portion of the aquifer.

(11) The initial salt concentration at locations where water quality data were collected should be adjusted based on the field data so the error from the previous simulation would be carried into the next simulation.

(12) The initial extents of the hypersaline plume: FPL suggested the plume extents in deep layers were overestimated in the model, compared to the CSEM data. If the CSEM data are validated, the initial salt concentration in the model should updated by the CSEM data wherever the CSEM data available.

40

Attachment F (13) To remediate the deep portion of the plume found west the CCS, alternative technologies, such as horizontal wells (van Heest, 2013; U.S. EPA, 2017) may be necessary in addition to the existing RWS wells.

(14) Extra extraction well may be needed to remediate the hypersaline plume found north of the CCS. Caution should be excised to avoid introduced salt water intrusion to the aquifer.

41

Attachment F 9 References Anderson, M.P., W.W. Woessner, R.J. Hunt, 2015. Applied Groundwater Modeling, 2nd edition, 564 p.

Bear, J., 1972. Dynamics of Fluids in Porous Media, Dover Publication Inc., New York, 764 p Academic Press, 564 p.

Chin, D., 2016. The cooling-canal system at the FPL Turkey Point power station From: https://www.miamidade.gov/environment/library/reports/cooling-canal-system-at-the-fpl-turkey-point-power-station.pdf.

Council, G.W. and Richards, C.J., 2008. A Saltwater Upconing Model to Evaluate Wellfield Feasibility. SWIM: 20th Salt Water Intrusion Meeting, Program and Proceedings. June 23-27.

Naples, Florida. Doherty, 2010.

Doherty, J., 2010. PEST: Modelindependent parameter estimation, Watermark Numer.

Comput., Brisbane, Queensl., Australia. [Available at http://www.pesthomepage.org.

Enercon, 2016a. Biscayne Aquifer Performance Testing, Turkey Point Facility. Technical Report prepared for Florida Power and Light. April 1.

Enercon, 2016b. PTN Cooling Canal System Electromagnetic Conductance Geophysical Survey Final Report. Technical report prepared for Florida Power and Light. May 13.

Fetter, C.W., 1992. Contaminant Hydrogeology, McConnin, New York, New York, 458.

Fish, J.E. and M. Stewart, 1991. Hydrogeology of the surficial aquifer system, Dade County, Florida, US Geological Survey, Water-Resources Investigations Report, 90-4108.

Florida Department of Environmental Protection (FDEP). 2016. Consent Order OGC No 16-0241.

June 20, 2016.

FPL, 2017. Turkey Point Cooling Canal System Phase I Remediation Action Plan, Jan. 27, 2017, Attachment 1: Biscayne Aquifer Groundwater flow and solute transport model:

Heterogeneous hydraulic conductivity analysis.

FPL, 2019. FPL Turkey Point Remedial Action Annual Status Report (RAASR), November 15, 2019.

Guo, W. and C. Langevin, 2002. Users Guide to SEAWAT: A computer Program for Simulation of Three-dimensional Variable-Density Ground-Water Flow, USGS Techniques of Water Resources Investigations, Book6, Chapter A7.

42

Attachment F Harbaugh, A.W., E. R. Banta, M. C. Hill, and M. G. McDonald, 2000. MODFLOW-2000, The U.S.

Geological Survey Modular Groundwater-Model User Guide to Modularization concepts and the groundwater flow USGS Open-File Report 00-92.

Hughes J.D. and White, J.T., 2014. Hydrologic Conditions in Urban Miami-Dade County, Florida, and the Effect of Groundwater Pumpage and Increased Sea Level on Canal Leakage and Regional Groundwater Flow. U.S. Geological Survey Scientific Investigations Report 2014-5162, 175 p., http://dx.doi.org/10.3133/sir20145162.

JLA Geosciences, 2010. Geology and Hydrogeology Report for FPL, Turkey Point Plant Groundwater, Surface Water, & Ecological Monitoring Plan. Prepared for Florida Power and Light Company. October 2010.

JLA Geosciences, 2016. Summary of hydraulic conductivity estimation, Technical Memorandum provided to FPL August 2016.

Klein, J.E., and Hull, H., 1978. Biscayne Aquifer, Southeast Florida. U.S. Geological Survey Water-Resources Investigations Report 78-107, 52 p.

Langevin, C.D., Shoemaker W.B., and Guo, W., 2003. MODFLOW-2000, the U.S. Geological Survey Modular Ground-Water Model - Documentation of the SEAWAT-2000 Version with the Variable-Density Flow Process (VDF) and the Integrated MT3DMS Transport Process (IMT): USGS Open-File Report 03-426, 43 p.

Langevin, C.D., Thore, D.T., Dausman A.M., Sukop, M.C., and Guo, W., 2008. SEAWAT Version 4:

A Computer Program for Simulation of Multi-Species Solute and Heat Transport: USGS Techniques and Methods Book 6, Chapter A22, 39 p.

Miami-Dade County (MDC) Department of Regulatory and Economic Resources (RER), Division of Environmental Resources Management (DERM). 2015. Consent Agreement between Miami-Dade County and Florida Power & Light Company. October 7, 2015.

Parker, G.G., Ferguson, G.E., and Love, S.K., 1955. Water Resources of Southeastern Florida, U.S. Geological Survey Water-Supply Paper 1255.

South Florida Water Management District, 2013. Letter from Terrie Bates, Director, Water Resource Division, to Barbara Linkiewicz, FPL & NextEra Energy Resources, Juno Beach, Florida, regarding Units 3 and 4 Post-Uprate Monitoring (reduction of groundwater/surface water monitoring). June 3, 2013.

Tetra Tech, 2016a. A Groundwater Flow and Salt Transport Model of the Biscayne Aquifer, Technical Memorandum provided to FPL June 10, 2016.

Tetra Tech, 2016b. Application of parameter estimation techniques to simulation of remedial alternatives at FPL Turkey Point Cooling Canal System, Technical memorandum provided to FPL, July, 2016.

43

Attachment F U.S. EPA, 1990. Basics of Pump-and-Treat Ground-Water Remediation. Robert S. Kerr Environmental Research Laboratory Office of Research and Development U.S. Environment Protection Agency, Ada, Oklahoma. 66p.

U.S. EPA, 2017. How to Evaluate Alternative Cleanup Technologies for Underground Storage Tank Sites-A Guide for Corrective Action Plan Reviewers.

Van Heest, G., 2013. Horizontal Wells for Groundwater Remediation: How a Technology That Revolutionized the Oil Industry Is Used to Remediate Groundwater. EM Magazine, June 2013.

Zheng, C. and G. Bennet, 1995. Applied Contaminant Transport Modeling - Theory and Practice, Van Nostrand Reihold, New York, New York, 440 p.

Zheng, C. and P. Wang, 1998. MT3DMS - Documentation and Users Guide, U.S. Army Corps of Engineers Waterways Experiment Station Technical Report, 214 p.

44

Attachment F Tables

Attachment F Table 4.1 Measured Chloride concentration (mg/l) in March 2018 (baseline) and March 2019.

Wells 3/1/2018 3/1/2019 Change % Shallow Middle deep TPGW-1S 19400 11000 -8400 -43.30 down TPGW-1M 27700 28300 600 2.17 up TPGW-1D 28500 29100 600 2.11 up TPGW-2S 24800 23800 -1000 -4.03 up TPGW-2M 29500 30600 1100 3.73 up TPGW-2D 31300 31800 500 1.60 up TPGW-4S 2280 1640 -640 -28.07 down TPGW-4M 15100 15200 100 0.66 no change TPGW-4D 14800 16000 1200 8.11 up TPGW-5S 164 142 -22 -13.41 down TPGW-5M 11700 12300 600 5.13 up TPGW-5D 13100 13600 500 3.82 up TPGW-7S 37 36.8 -0.2 -0.54 no change TPGW-7M 40 40.4 0.4 1.00 no change TPGW-7D 3970 3780 -190 -4.79 down TPGW-12S 16500 18900 2400 14.55 up TPGW-12M 20900 24200 3300 15.79 up TPGW-12D 24000 26600 2600 10.83 up TPGW-15S* 20100 12600 -7500 -37.31 down TPGW-15M* 30000 30100 100 0.33 no change TPGW-15D* 28800 29200 400 1.39 up TPGW-17S 24900 24700 -200 -0.80 no change TPGW-17M 29300 28600 -700 -2.39 down TPGW-17D 28600 30300 1700 5.94 up TPGW-18S 14200 7680 -6520 -45.92 down TPGW-18M 25200 24800 -400 -1.59 down TPGW-18D 26400 25400 -1000 -3.79 down TPGW-19S 1830 3330 1500 81.97 up TPGW-19M 26000 22000 -4000 -15.38 down TPGW-19D 26800 24700 -2100 -7.84 down TPGW-L3-18 2030 874 -1156 -56.95 down TPGW-L3-58 31400 32000 600 1.91 up TPGW-L5-18 1290 677 -613 -47.52 down TPGW-L5-58 29500 29300 -200 -0.68 no change Note: No change if Change <=1%;

Attachment F Table 4.2 Evaluation of the 1 st year CCS remediation System Screens Total Uptrend No change Downtrend Shallow 10 3 2 5 Middle 12 4 3 5 Deep 12 8 1 3

Attachment F Table 4.3 Estimated Plume Reduction based on CSEM survey data of 2018 and 2019 Voxel Grid Layer Bottom Depth Layer Thickness Volume Changes from 2018 Baseline Layer from land surface (ft) (ft) (%) (C>19000 mg/l) 1 3.3 3.28 0 2 6.9 3.61 0 3 10.8 3.94 0 4 15.4 4.59 140 5 20.4 4.92 -70 6 25.9 5.58 -42 7 32.1 6.23 -16 8 39 6.89 -39 9 46.6 7.54 -42 10 55.1 8.53 -24 11 64.6 9.51 -30 12 75.1 10.50 -18 13 86.6 11.48 -5 14 99.4 12.79 -6 Note: Information of Layer bottom depth and thickness is based on FPL 2019: Table 4.2-1; Volume changes are based on FPL 2019: Table 4.3-2.

Attachment F Table 4.4 Estimated Plume Reduction based on the Groundwater Modeling Results Days from Layer Layer Layer Layer Layer Layer Layer Layer Layer Layer Layer April 30, 2018 1 2 3 4 5 6 7 8 9 10 11 Total 0 1908 1667 1740 2426 4505 6487 8235 9043 9058 9187 8728 62984 31 1681 1673 1747 1966 4346 6391 8206 8996 9055 9191 8739 61991 61 1726 1661 1739 1663 4099 6347 8157 8952 9048 9195 8747 61334 92 1702 1649 1722 1538 3934 6279 8126 8929 9047 9199 8756 60881 123 1703 1641 1713 1468 3782 6225 8066 8794 9043 9208 8762 60405 153 1602 1633 1696 1345 3530 6083 7955 8708 9036 9209 8771 59568 184 1644 1624 1682 1360 3285 6065 7934 8718 9029 9214 8780 59335 214 1633 1611 1662 1344 3104 6035 7966 8804 9028 9220 8786 59193 245 1610 1604 1653 1267 2937 5987 7966 8830 9028 9225 8794 58901 276 1610 1596 1633 1215 2899 5941 7969 8889 9031 9231 8802 58816 304 1629 1589 1621 1238 2652 5898 7962 8852 9034 9233 8809 58517 335 1652 1579 1606 1427 2801 5819 7854 8701 9037 9238 8813 58527 365 1719 1579 1597 1424 2737 5834 7882 8770 9040 9241 8820 58643 396 1788 1589 1607 1429 2754 5885 7864 8616 9025 9240 8826 58623 Changes of Cells (C/Co>1) -120 -78 -133 -997 -1751 -602 -371 -427 -33 53 98 -4361 Percentage ( %) -6.29 -4.68 -7.64 -41.10 -38.87 -9.28 -4.51 -4.72 -0.36 0.58 1.12 -6.924

Attachment F Table 6-1. Sensitivity analysis of Dispersivities Dispersivity Prediction Upper range Lower Range Longitudinal (ft) 11.2777 22.5554 5.6389 Transverse ratio 0.1 0.5 0.05 Vertical ratio 0.001 0.01 0.0001

Attachment F Table 6-2 Summary of cumulative salt extraction (million lbs)

Years Prediction aLx2 aLd2 aT005 aT05 aV00001 aV001 Cour1 Cour20 SEN1 SEN2 1 1836 1810 1855 1836 1836 1836 1822 1875 1806 1823 1835 2 3591 3542 3627 3591 3592 3592 3562 3695 3531 3577 3578 3 5257 5183 5309 5257 5257 5257 5216 5437 5167 5220 5217 4 6834 6739 6902 6834 6835 6835 6786 7091 6716 6761 6759 5 8352 8235 8434 8351 8353 8353 8295 8683 8207 8236 8232 6 9815 9679 9912 9815 9816 9816 9749 10217 9643 9654 9648 7 11239 11082 11349 11239 11241 11241 11162 11706 11040 11029 11021 8 12629 12452 12753 12630 12631 12631 12539 13159 12403 12368 12360 9 13994 13797 14131 13995 13996 13996 13889 14583 13680 13680 13672

% 1 -1.41 0.98 0.01 0.02 0.02 -0.75 4.21 -2.24 -2.24 -2.30

Attachment F Table 7-1 Measured Chloride Concentration (mg/l) at TPGW wells between 3/20/2018 and 3/20/2019

Attachment F Figures

Attachment F Figure 1-1 Map of the project area Groundwater Tek Inc

Attachment F Figure 1-2 Location of monitoring and remediation wells Groundwater Tek Inc

Attachment F TPGW-1D 1.6 1.5 1.4 Relative Salinity 1.3 Modeled Observed 1.2 1.1 1

2/19/2018 5/30/2018 9/7/2018 12/16/2018 3/26/2019 7/4/2019 Figure 4-1 Comparison of modeled and observed water quality change March 2018 to May 2019 TPGW-1D Groundwater Tek Inc

Attachment F TPGW-2D 1.68 1.66 1.64 Relative Salinity 1.62 Modeled Observed 1.6 1.58 1.56 2/19/2018 5/30/2018 9/7/2018 12/16/2018 3/26/2019 7/4/2019 Figure 4-2 Comparison of modeled and observed water quality change March 2018 to May 2019 TPGW-2D Groundwater Tek Inc

Attachment F TPGW-15D 1.8 1.6 1.4 Relative Salinity 1.2 1

Modeled 0.8 Observed 0.6 0.4 0.2 0

2/19/2018 5/30/2018 9/7/2018 12/16/2018 3/26/2019 7/4/2019 Figure 4-3 Comparison of modeled and observed water quality change March 2018 to May 2019 TPGW-15D Groundwater Tek Inc

Attachment F TPGW-17D 1.62 1.6 1.58 1.56 Relative Salinity 1.54 1.52 Modeled 1.5 Observed 1.48 1.46 1.44 1.42 2/19/2018 5/30/2018 9/7/2018 12/16/20183/26/2019 7/4/2019 Figure 4-4 Comparison of modeled and observed water quality change March 2018 to May 2019 TPGW-17D Groundwater Tek Inc

Attachment F TPGW-18D 1.6 1.4 1.2 Relative Salinity 1

0.8 Modeled Observed 0.6 0.4 0.2 0

2/19/2018 5/30/2018 9/7/2018 12/16/2018 3/26/2019 7/4/2019 Figure 4-5 Comparison of modeled and observed water quality change March 2018 to May 2019 TPGW-18D Groundwater Tek Inc

Attachment F 00 00 40 90 0 0 00 00 14 19 0

00 0 19 2400 Figure 4-6 Chloride isochlor plan view map for shallow intervals Based on the data of March 22, 2018 (FPL 2019: Figure 3.2.2)

Groundwater Tek Inc

Attachment F 0

00 19 0

400 0 00 1400 240 0

900 Figure 4-7 Chloride isochlor plan view map for middle intervals Based on the data of March 22, 2018 (FPL 2019: Figure 3.2.3)

Groundwater Tek Inc

Attachment F 00 00 190 00 0 140 400 240 0

900 Figure 4-8 Chloride isochlor plan view map for deep intervals Based on the data of March 22, 2018 (FPL 2019: Figure 3.2.4)

Groundwater Tek Inc

Attachment F Figure 4-9 Salinity distribution based on monthly calibration model (Layer 4) for March 2018 (FPL 2019 App. G: Fig. G.4-3)

Groundwater Tek Inc

Attachment F Figure 4-10 Salinity distribution based on monthly calibration model (Layer 8) for March 2018 (FPL 2019, App. G: Fig. G.4-4)

Groundwater Tek Inc

Attachment F Figure 4-11 Salinity distribution based on monthly calibration model (Layer 11) for March 2018 (FPL 2019, App. G: Fig. G.4-5)

Groundwater Tek Inc

Attachment F Monthly Salt Extraction Rates 250 Salt Extracted (million lbs) 200 150 GTI 100 FPL Data 50 0

Figure 4-12 Comparison of monthly salt extraction (million lbs) from May 2018 to May 2019 Groundwater Tek Inc

Attachment F Cumulative Sat Extraction 250 RWS1 RWS2 Salt Extraction (million lbs) 200 RWS3 150 RWS4 RWS5 100 RWS6 RWS7 50 RWS8 RWS9 0 RWS10 Apr-18 Jul-18 Oct-18 Feb-19 May-19 Aug-19 Figure 4-13 Total salt extraction by well (million lbs) from May 2018 to May 2019 Groundwater Tek Inc

Attachment F Modeled Total Cumulative Salt Extraction 20 Cumulative Salt Extraction (10^8 lbs) 18 16 14 12 10 8

6 4

2 0

Apr-18 May-18 Jul-18 Sep-18 Oct-18 Dec-18 Feb-19 Mar-19 May-19 Jul-19 Figure 4-14 Monthly total salt extraction from RWS wells (million lbs) from May 2018 to May 2019 Groundwater Tek Inc

Attachment F Figure 4-15 Modeled extent of the hypersaline plume (Layer 1) from May 2018 to May 2019 Groundwater Tek Inc

Attachment F Figure 4-16 Modeled extent of the hypersaline plume (Layer 4) from May 2018 to May 2019 Groundwater Tek Inc

Attachment F Figure 4-17 Modeled extent of the hypersaline plume (Layer 8) from May 2018 to May 2019 Groundwater Tek Inc

Attachment F Figure 4-18 Modeled extent of the hypersaline plume (Layer 9) from May 2018 to May 2019 Groundwater Tek Inc

Attachment F Figure 4-19 Modeled extent of the hypersaline plume (Layer 11) from May 2018 to May 2019 Groundwater Tek Inc

Attachment F Figure 4-20 CSEM Survey Area Groundwater Tek Inc

Attachment F 2018 AEM Chloride Mapping along L105301 2019 AEM Chloride Mapping along L105301 W 7

-4 W-1 G G L-31 TP TP E

The CCS TPGW-17D Figure 4-21 Chloride profiles along flight path L105301 CSEM Survey Groundwater Tek Inc

Attachment F 2018 2019 2018 2019 2018 2019 Layer 12 (19.7-22.9m bls) Layer 13 (22.9-26.4m bls) Layer 14 (26.4-30.3m bls) 0 2000m Figure 4-22.

Change of the plume extent (Chloride 19,000 mg/l) based on CSEM survey in deep layers 2018-2019 Groundwater Tek Inc

Attachment F a b 0 2000m L

g/

m 0 00 19 r ide C hlo Figure 4-23.

Spatial extent of the plume nearby TPGW-18 in FPL Voxel grid L12 and model layer 10 (a) Based on CSEM data (b) Based on modeling results May 2019 Groundwater Tek Inc

Attachment F TPGW-18D Figure 4-24 CSEM survey data along flight path L102301 Prediction model Groundwater Tek Inc

Attachment F 1800 Cumulative Salt Extraction (million lbs) 1600 RWS1 1400 RWS2 1200 RWS3 1000 RWS4 800 RWS5 600 RWS6 400 RWS7 200 RWS8 RWS9 0

0 2 4 6 8 10 RWS10 Years Figure 5-1 Predicted monthly salt extraction (million lbs)

Prediction Model Groundwater Tek Inc

Attachment F 180 160 Monthly Salt Extraction (million lbs) 140 120 100 80 60 40 20 0

0 2 4 6 8 10 Years Figure 5-2.

Predicted total monthly salt extraction (million lbs)

Prediction Model Groundwater Tek Inc

Attachment F 16000 Cumulative Salt Extraction (million lbs) 14000 12000 10000 8000 6000 4000 2000 0

0 2 4 6 8 10 Years Figure 5-3 Predicted total cumulative salt extraction from the RWS wells (million lbs)

Prediction Model Groundwater Tek Inc

Attachment F r

Figure 5-4 Modeled plume extents after 0, 3, 5, and 10 years Layer 1 Groundwater Tek Inc

Attachment F Figure 5-5 Modeled plume extents after 0, 3, 5, and 10 years Layer 4 Groundwater Tek Inc

Attachment F Figure 5-6 Modeled plume extents after 0, 3, 5, and 10 years Layer 8 Groundwater Tek Inc

Attachment F Figure 5-7 Modeled plume extents after 0, 3, 5, and 10 years Layer 9 Groundwater Tek Inc

Attachment F Figure 5-8 Modeled plume extents after 0, 3, 5, and 10 years Layer 11 Groundwater Tek Inc

Attachment F Figure 5-9 Capture zone in plan view Prediction model Groundwater Tek Inc

Attachment F Figure 5-10 Modeled water levels (ft, NGVD) in the surficial aquifer (Layer 1) under wetter conditions Prediction Model Groundwater Tek Inc

Attachment F (ft)

Figure 5-11 Predicted drawdown (in ft) in the surficial aquifer Prediction model Groundwater Tek Inc

Attachment F Figure 6-1 Modeled heads (ft, NGVD) in the surficial aquifer under wetter conditions SEN 1 Groundwater Tek Inc

Attachment F (ft)

Figure 6-2 Model drawdown (ft) in the surficial aquifer under wetter conditions SEN 1 Groundwater Tek Inc

Attachment F Figure 6-3 Predicted extent of the hypersaline plume (Layer 1)

SEN 1 Groundwater Tek Inc

Attachment F Figure 6-4 Predicted extent of the hypersaline plume (Layer 4)

SEN 1 Groundwater Tek Inc

Attachment F Figure 6-5 Predicted extent of the hypersaline plume (Layer 8)

SEN 1 Groundwater Tek Inc

Attachment F Figure 6-6 Predicted extent of the hypersaline plume (Layer 11)

SEN 1 Groundwater Tek Inc

Attachment F 16000 14000 Cumulative Salt Extraction (million lbs) 12000 10000 8000 Prediction 6000 SEN1 4000 2000 0

0 2 4 6 8 10 Years Figure 6-7 Cumulative salt extraction (million lbs)

SEN 1 Groundwater Tek Inc

Attachment F Figure 6-8 Modeled heads (ft, NGVD) in the surficial aquifer with UIC pumping at 3 mgd SEN 2 Groundwater Tek Inc

Attachment F Figure 6-9 Predicted drawdown (in ft) in the surficial aquifer with UIC pumping at 3 mgd SEN 2 Groundwater Tek Inc

Attachment F Figure 6-10 Predicted extent of the hypersaline plume (Layer 1)

SEN 2 Groundwater Tek Inc

Attachment F Figure 6-11 Predicted extent of the hypersaline plume (Layer 4)

SEN 2 Groundwater Tek Inc

Attachment F Figure 6-12 Predicted extent of the hypersaline plume (Layer 8)

SEN 2 Groundwater Tek Inc

Attachment F Figure 6-13 Predicted extent of the hypersaline plume (Layer 11)

SEN 2 Groundwater Tek Inc

Attachment F 16000 Cumulative Salt Extraction (million lbs) 14000 12000 10000 8000 Prediction 6000 SEN2 4000 2000 0

0 2 4 6 8 10 Years Figure 6-14 Predicted cumulative salt extraction (million lbs)

SEN 2 Groundwater Tek Inc

Attachment F 16000 Cumulative Salt Extraction (million lbs) 14000 12000 10000 8000 Prediction 6000 SEN2 4000 2000 0

0 2 4 6 8 10 Years Figure 6-15 Predicted cumulative salt extraction (million lbs)

Sensitivity of Courant numbers Groundwater Tek Inc

Attachment F 16000 Cumulative Salt Extraction (million lbs) 14000 12000 10000 8000 Prediction 6000 aLx2 aLd2 4000 2000 0

0 2 4 6 8 10 Years Figure 6-16 Predicted cumulative salt extraction (million lbs)

Sensitivity analysis of longitudinal dispersivitiy Groundwater Tek Inc

Attachment F 16000 14000 Cumulative Salt Extraction (million lbs) 12000 10000 8000 Prediction 6000 aT005 aT05 4000 2000 0

0 2 4 6 8 10 Years Figure 6-17 Predicted cumulative salt extraction (million lbs)

Sensitivity analysis of transverse dispersivity Groundwater Tek Inc

Attachment F 16000 14000 Salt Extraction (million lbs) 12000 10000 8000 Prediction 6000 aV00001 aV001 4000 2000 0

0 2 4 6 8 10 Years Figure 6-18 Predicted cumulative salt extraction (million lbs)

Sensitivity analysis of vertical dispersivity Groundwater Tek Inc

Attachment F Range of Predictions Prediction 16000 aLx2 Cumuative Salt Extraction (million lbs) 14000 aLd2 12000 aT005 10000 aT05 8000 aV00001 6000 aV001 4000 Cour1 2000 Cour20 0

0 1 2 3 4 5 6 7 8 9 10 SEN1 Years SEN2 Figure 6-19 Range of predicted cumulative salt extraction (million lbs)

Sensitivity Analysis Groundwater Tek Inc

Attachment F 2018 Model Row 158 2019 Location Map Figure 7-1 Comparison of CSEM survey and modeled salinity (along model row 158)

(FPL 2019: Figure 5.3-5b)

Groundwater Tek Inc

Attachment F ATTACHMENT C Peer Review Initial Comments CSEM

Attachment F Summary of Preliminary DERM Consultants CSEM Peer Review Comments

1. As close adherence to the flight lines and flight directions should be made as is feasible within the tolerances of the available instrumentation.

2. Continued use of the same model of SkyTEM TEM instrument (the model 304) is highly recommended for consistency.

3. Given the narrow corridor of data collection that is possible east of the line of pumping wells in the evaporation pond area, it is highly recommended that all sources of electrical current such as power transmitted to well pumps or other power lines passing through the survey area be de-energized during the duration of the survey since this type of noise leads to capacitive coupling in the TEM data.

4. To the extent possible, all movable metallic objects such as roll-off containers, trucks and trailers, and stockpiles of materials should be removed from the survey area to avoid galvanic couplings.

5. To the extent possible, we recommend that a baseline area distant from noise sources and outside the area of influence of the site remediation be chosen as a control area to perform a year to year comparison to independently identify potential systematic differences in the CSEM data collection, instrument calibration, processing. The baseline area would be chosen based on low data residuals and a relatively large degree of spatial variability of electrical resistivity outside the hypersaline plume. An area with minimal temporal variations and level trend lines in groundwater temperature, salinity, conductivity, etc. would be most appropriate.

6. Groundwater sampling and collection of induction conductivity logs in monitoring wells should be done as closely to the dates of the CSEM survey as possible.