Changes in Pumping-Induced Groundwater Quality Used to Supply a Large-Capacity Brackish-Water Desalination Facility, Collier County, Florida: A New Aquifer Conceptual Model

: Brackish-water reverse osmosis (BWRO) desalination facilities are designed to treat feedwater within a ﬁxed range in salinity. If the salinity and ion concentrations of the feedwater rises above the maximum design concentrations, then the plant may ultimately fail. BWRO plants typically use groundwater as a feedwater source. Prior to the process design, a detailed groundwater assessment is made to characterize the source aquifer system and to develop a solute-transport model that is used to project the changes in water quality over the expected useful life of the facility. Solute transport-modeling performed for the Collier County (Florida) South BWRO facility, which was designed to produce 30,303 m 3 /d with an expansion to 75,758 m 3 /d, used an aquifer system conceptual model that assumed upwards migration over time of brackish waters with higher salinities into the production zones. This conceptual model is typical of how most BWRO systems developed in the United States operate. The original solute transport model predicted a range of increases in dissolved chloride concentrations over a 20-year period from a low of 5 mg/L/yr, a mid-range of 35 mg/L/yr, and a high range of 85 mg/L/yr. Actual data collected over a 11- to 13.5-year period showed that the dissolved chloride concentration average of the feed water decreased by 16 mg/L/yr. The original conceptual model was found to be inaccurate in that it suggested an upwards recharging system, whereas downward leakage (or perhaps lateral migration) of fresher water appears to be occurring in the system. This is an example of a long-term solute-transport model audit, which is rarely performed, in which a new conceptual model was found to be applicable to an aquifer system used to feed a BWRO facility. of the initial and alternative possible conceptual models. This research is very important to gaining fundamental knowledge on how best to coordinate the design of wellﬁelds in relationship to BWRO process design.


Introduction
Freshwater shortages plague many regions of the world, particularly in high-population growth areas and arid regions [1]. While conventional sources of water supply are often quite limited, including fresh surface water bodies and fresh groundwater sources, brackish groundwater is abundant in many areas, especially the United States [2]. Brackish water is defined by the U.S. Geological Survey to have a total dissolved solids (TDS) concentration lying between 1000 and 10,000 mg/L [2]. The lower range of brackish-water, between TDS concentrations of 1000 and 3000 mg/L, can replace some uses of freshwater, particularly for irrigation of salt-tolerate plants and some industrial applications [3,4]. The U.S. Environmental Protection Agency defines underground sources of drinking water to have a TDS of less than 10,000 mg/L [5]. Brackish water, however, cannot be used directly as drinking water based on the drinking water quality standards [5].

Hydrogeology of the Aquifer System
The Collier County South BWRO facility was originally designed to c duction wells with individual well capacities of 3818 m 3 /d (700 gpm). Te Hawthorn Aquifer System-Zone I and four wells tap the Lower Hawthor ditional production wells, 24 completed in Hawthorn Aquifer Zone I and the Lower Hawthorn Aquifer, were subsequently added for a plant capaci The hydrogeology of the production wellfield area is summarized in thorn Aquifer Zone 1 is a brackish-water aquifer lying within the Interm System [32]. Zone 1 is confined from the overlying Surficial Aquifer Syst confining unit lying within the Peace River Formation of the Hawthorn G and is confined below a semi-confining unit lying at the base of the Interm System. The overlying aquifers contain freshwater and the underlying Lo Aquifer contains brackish water. The Lower Hawthorn Aquifer is the upp in the Floridan Aquifer System [33,34]. Note that the aquifer system show of stratigraphic variability, and a minor aquifer does occur within the conf arating the two aquifers ( Figure 3).

Hydrogeology of the Aquifer System
The Collier County South BWRO facility was originally designed to contain 14 production wells with individual well capacities of 3818 m 3 /d (700 gpm). Ten wells tap the Hawthorn Aquifer System-Zone I and four wells tap the Lower Hawthorn Aquifer. Additional production wells, 24 completed in Hawthorn Aquifer Zone I and three wells in the Lower Hawthorn Aquifer, were subsequently added for a plant capacity expansion.
The hydrogeology of the production wellfield area is summarized in Figure 2. Hawthorn Aquifer Zone 1 is a brackish-water aquifer lying within the Intermediate Aquifer System [32]. Zone 1 is confined from the overlying Surficial Aquifer System by a semiconfining unit lying within the Peace River Formation of the Hawthorn Group at the top and is confined below a semi-confining unit lying at the base of the Intermediate Aquifer System. The overlying aquifers contain freshwater and the underlying Lower Hawthorn Aquifer contains brackish water. The Lower Hawthorn Aquifer is the uppermost aquifer in the Floridan Aquifer System [33,34]. Note that the aquifer system shows some degree of stratigraphic variability, and a minor aquifer does occur within the confining unit separating the two aquifers ( Figure 3).
Characterization of the Hawthorn Aquifer Zone 1 was reported in CDM [32]. The aquifer consists of porous limestone with a moderate degree of heterogeneity. Test drilling for geologic characterization of the Hawthorn Zone 1 aquifer found that it occurred between 91.5 and 137 m below surface. The thickness of the aquifer ranged from 30.5 and 45.7 m. An aquifer performance test was conducted from which transmissivity, storativity, and leakance values of 248 to 1242 m 2 /d, 1 × 10 −4 to 1 × 10 −5 , and 2.2 × 10 −3 d −1 , respectively, were obtained. The dissolved chloride concentration in the aquifer ranged from 1800 to 2800 mg/L with an approximate TDS concentration ranging from 3350 to 4940 mg/L. The approximate TDS values were obtained from chloride data based on the relationship in this aquifer system [21].   Characterization of the Hawthorn Aquifer Zone 1 was reported in CDM [32]. The aquifer consists of porous limestone with a moderate degree of heterogeneity. Test drilling for geologic characterization of the Hawthorn Zone 1 aquifer found that it occurred between 91.5 and 137 m below surface. The thickness of the aquifer ranged from 30.5 and 45.7 m. An aquifer performance test was conducted from which transmissivity, storativity, and leakance values of 248 to 1242 m 2 /d, 1 × 10 −4 to 1 × 10 −5 , and 2.2 × 10 −3 d −1 , respectively, were obtained. The dissolved chloride concentration in the aquifer ranged from The Lower Hawthorn Aquifer consists of porous limestone and dolomite units that exhibit considerable heterogeneity [32]. The aquifer occurs between 198 and 244 m below surface with a relatively uniform thickness of 46 m. Due to the extreme heterogeneity, the aquifer has a range in measured transmissivity of 124 to 3229 m 2 /d. The measured storativity is near 1 × 10 −4 , and the leakance was estimated to be about 7 × 10 −3 d −1 .
Based on the geology, aquifer testing results, and groundwater modeling, it was demonstrated that the pumping of the upper aquifer will have little impact on the overlying freshwater aquifer system because of the confinement and the stark contrast in water quality across the confining unit. If the potentiometric surface in the aquifers is all converted to freshwater heads, the potentiometric surface of the Hawthorn Aquifer Zone 1 containing brackish water is 3 to 5 m above the potentiometric surfaces of the Lower Tamiami Aquifer and the water-table aquifer, which both contain freshwater. No potentiometric surface data are available from the Sandstone Aquifer that overlies Hawthorn Aquifer Zone 1. The measured leakance value for Hawthorn Aquifer Zone 1 is therefore predominantly unidirectional and directed upward from the Lower Hawthorn Aquifer. The leakance measured in the Lower Hawthorn Aquifer is likely bi-directional with some skewing toward bottom-upward water movement during a pumping condition as there is a lesser degree of confinement at the base compared to the top of the aquifer. The head differential between the Lower Hawthorn Aquifer and the overlying Hawthorn Aquifer Zone 1 shows an upward directed difference of 0.3 to 1.2 m. Groundwater leakage into Lower Hawthorn Aquifer from below would tend to cause an increase in salinity. Data collected during construction of the injection well disposal system at the BWRO plant site indicate that the dissolved chloride concentration in Middle Confining Unit is up to 5500 mg/L, which is equivalent to a TDS of 10,000 mg/L.
Groundwater solute-transport modeling was conducted on the combined two-aquifer system using the SEAWAT code [35]. With a pumping rate at 41,666 m 3 /d divided between the two aquifers, the maximum drawdown in Hawthorn Aquifer Zone 1 was estimated to be about 15 m in the center of the cone of depression and maximum drawdown in the Lower Hawthorn Aquifer was simulated to be 12 m in the center of the cone of depression. Based on a 20-year period of operation at a pumping rate of 41,666 m 3 /d, the dissolved chloride concentration was projected to increase from the initial dissolved chloride concentration of 2600 mg/L to an average of 3500 mg/L [32]. This range of values equates to a TDS concentration range of 4622 to 6000 mg/L. The calculated chloride and TDS concentrations are considered best-estimated projections based on measured aquifer hydraulic data. The results of a sensitivity analysis indicated that the chloride concentration could rise as high as 4250 mg/L and TDS as high as 8100 mg/L. The groundwater modeling also included pumping of the North Collier Regional Water Treatment Plant Wellfield, which provides feedwater to another BWRO facility.

Design of the Brackish-Water Treatment Facility and Water Quality Treatment Limitations
The South Collier County BWRO facility was originally designed and constructed to meet a potable water production capacity of 30,303 m 3 /d (8 MGD). The projected conversion efficiency of the plant is about 80%, which means that of the raw water that is pumped into the BWRO plant, 80% produced becomes potable water and the remaining water is concentrate (brine). Therefore, the feedwater source must supply about 41,667 m 3 /d (11 MGD) to produce the desired quantity of potable water. Eleven production wells were constructed to meet the feedwater demand, each producing 3788 m 3 /d (1 MGD). The plant was brought online in 2004. An expansion of the plant to bring the capacity to 75,758 m 3 /d (20 MGD) was completed and brought into production in 2007. The expanded wellfield contained a total of 41 production wells with 35 in Hawthorn Aquifer Zone 1 and 6 production wells in the Lower Hawthorn Aquifer. The wellfield capacity was in excess of the minimum feedwater pumping rate required of 101,023 m 3 /d (26.67 MGD) as excess well capacity has been built into the system as backup to account for well or well pump failure. The locations of the production wells are given in Figure 4 and the production well    The BWRO plant was designed to treat a feedwater with a maximum TDS of between 8000 and 9000 mg/L with a corresponding dissolved chloride concentration of between approximately 4000 and 4900 mg/L [36]. If the salinity of the feedwater becomes higher than the ability of the BWRO process to treat the raw water, a major modification of the process would be required. The treatment process upgrade could include higher pressure pumps, new membranes, and an increase in raw water capacity caused by a reduced conversion rate. An upgrade of the treatment system to handle higher salinity water might increase the footprint of the plant and increase the concentrate flow to the injection wells.

Water Use and Quality Data Collection
Collier County Utilities monitors the pumping rate and dissolved chloride concentration in each production well. The monthly totalized pumping data for all production wells and the monthly measured dissolved chloride concentration were obtained. These data were graphically analyzed to assess the changes in pumping (bar graphs) and dissolved chloride concentration. Additionally, the combined feedwater into the BWRO plant is monitored continuously for conductivity, which can be converted to TDS. The data for the years 2018 to 2020 were compiled for comparison to the individual well water salinity changes.

Statistical Analysis of Data
The trends in dissolved chloride changes were assessed by a regression analysis of the dataset for each production well. The R 2 and p-value were calculated for each regression to evaluate statistical significance. The regression lines were then used to project changes in dissolved chloride concentration into the future at 5, 10, 20 and 40 years for the current design pumping rate of 56,818 m 3 /d. The projected change in water quality was then evaluated in comparison to the maximum design capacity of the BWRO plant. Many wells that tap Hawthorn Aquifer Zone 1 show a declining trend. The projected trends from the operational data were compared to the results of previously performed solute-transport modeling. The statistical analysis does not take into consideration any future increase in plant capacity. This type of analysis would require a new, expanded solute-transport model to assess expanding aquifer pumping impacts on water quality.

Changes in Salinity in Wells Used for Water Supply for 13 Years of Operations in Hawthorn Aquifer Zone 1
Data obtained from 35 production wells tapping Hawthorn Aquifer Zone 1 were graphed to show the monthly aggregated pumpage and the dissolved chloride concentration ( Figure 5). There is considerable variation in the R 2 values, which define the tightness of the data fit to the trend line. R 2 values range from 7 × 10 −6 to 0.1393. The very low R 2 values are indicative of extreme scatter in the dissolved chloride data, which is likely caused by inaccuracies in the laboratory method used to measure dissolved chloride concentrations and operator error, along with the variable pumping schedules in the wellfield. In total, 43% of the wells have a p-value > 0.05 and 57% wells gave a value of <0.05. The p-values below 0.05 are considered to indicate that the regression is statistically significant, whereas trends with higher values are not considered significant.
The period of record for the dissolved chloride data is given in Table 2. Note that many of the wells had data on monthly pumpage, while other wells did not for the early period of record. The starting and ending dissolved chloride measurements for all the Hawthorn Aquifer Zone 1 wells are given in Table 3 along with the estimated annual rate of change based on these measurements. Note that the period of record is not the same for all production wells, because a larger number of wells was required after the BWRO plant was expanded. To normalize the data, the annual rate of change in the dissolved chloride concentration was calculated. The linear regression analysis for all the production wells show that the intercept of the y-axis does not correlate well with the starting measurement for dissolved chloride concentrations due to the error within the measurements (Table 3). Dissolved chloride concentration increased in only three wells and decreased in 32 wells based on the linear regression. The average rate of change using the starting and ending dissolved chloride concentration was 17 mg/L/yr. concentration was calculated. The linear regression analysis for all the production wells show that the intercept of the y-axis does not correlate well with the starting measurement for dissolved chloride concentrations due to the error within the measurements (Table 3). Dissolved chloride concentration increased in only three wells and decreased in 32 wells based on the linear regression. The average rate of change using the starting and ending dissolved chloride concentration was 17 mg/L/yr.

40-Year
Proj. This number is corrected for the period of record differences between a maximum of 13.5 years and a minimum of 11.17 years. 2 This is the lowest number possible for reduction in dissolved chloride concentration based on the overlying aquifer chemistry.

Changes in Salinity in Wells Used for Water Supply for 13 Years of Operation in the Lower Hawthorn Aquifer
The wellfield produces feedwater from six Lower Hawthorn Aquifer wells. Monthly pumpage and measured dissolved chloride concentrations are shown in Figure 6. The period of record of dissolved chloride data for each of the Lower Hawthorn Aquifer production wells is given in Table 4. There are some differences in the length of time the wells have been operated. A key observation is that that the changes in the dissolved chloride concentrations over the period of record are minimal (Table 5). Of the six wells, four show a downward trend in dissolved chloride concentration and two show an upward trend. The largest increase is in well RO-10S at about 101 mg/L per year. The largest decrease in concentration occurred in well RO-8S, which was 35 mg/L per year. The overall change in dissolved chloride concentration over the full operating period is a slight increase of 1 mg/L per year based on the starting and ending measurements during the period of record. Note that the intercept of the y-axis that defines the beginning of the trend line does not match the starting dissolved chloride measurement. This is caused by the laboratory error and the changes in pumping rates over time. The scatter plots of the dissolved chloride concentrations with time show very low R 2 values. Only half (50%) of the linear regression-based trend lines meet the test of statistical significance as indicated by p-values of less than 0.05. Of the six wells, four show a downward trend in dissolved chloride concentration and two show an upward trend.

40-Year
Proj.  The dissolved chloride concentration in the Lower Hawthorn Aquifer is also projected to decrease over time, but to a lesser degree than Hawthorn Aquifer Zone 1. The average decrease in concentration is 1.6% in 5 years, 3.0% in 10 years, 4.4% in 20 years, and 7.4% in 40 years. Again, the 5-and 10-year projections are considered reliable.

RO
If it is assumed that all the production wells are pumped at equal rates into the future at the pumping rate occurring at the end of 2020, then there will be a continued decline in dissolved chloride concentration. The dissolved chloride vales based on the data for all 41 production wells would be 2650 mg/L in 5 years, 2576 mg/L in 10 years, 2433 mg/L in 20 years, and 2163 mg/L in 40 years. As pumping is increased to meet future increased demand, the overall feedwater could decrease or increase based on the conceptual model used to explain the changes.

Comparison of the Groundwater Solute-Transport Model Results versus the Actual Changes in Water Quality
An initial solute transport model was conducted to assess salinity changes at a pumping rate of 41,667 m 3 /d, which was sufficient to produce the desired 30,303 m 3 /d of potable water [27]. The model was developed using the SEAWAT code [35] and water quality changes were projected over a 20-year period (Figure 7). It was assumed the initial wellfield would have eight wells tapping the Hawthorn Aquifer Zone 1 and three wells tapping the Lower Hawthorn Aquifer. The most probable change in water quality expected occurred within the shaded area in Figure 7. The modeling projected an increase in dissolved chloride concentration from the starting value of 2600 to the best estimate of 3300 mg/L in the line within the blue shaded area. This equates to a dissolved chloride increase of about 35 mg/L/yr (TDS of 63 mg/L/yr) over 20 years. Note that the various sensitivity analysis runs produce two sets of possible dissolved chloride concentration outcomes with the first being 3500 mg/L at the top of the shaded area (i.e., best estimate range) or as high as 4250 mg/L for more unfavorable hydrogeological conditions. The range of projections consider the variability in the aquifer hydraulic coefficients, the variability of water quality within the two aquifers (Figures 8 and 9), and the interaction during pumping of the aquifers. Therefore, the highest rate of pumping-induced change in dissolved chloride concentration would be 1650 mg/L in 20 years or 82.5 mg/L/yr. Note that a no-flow boundary was assumed to be at the top of Hawthorn Aquifer Zone I within the confining bed separating it from the Sandstone Aquifer. This fits with the conceptual model, which accurately describes most of the wellfields used in Florida for BWRO source water (Figure 1).

Conceptual Model of the Aquifer System Based on the Changes in Salinity
The conceptual model for the South Collier BWRO wellfield is different compared to the typical conceptual model used to evaluate most aquifers utilized to produce feed water in Florida and in other parts of the United States [19][20][21][22][23][24][25]. Nearly all the BRWO plants wellfields in the United States pump brackish water from a single aquifer. The South Collier County Wellfield has unique hydrogeology that allows water to be pumped from two different aquifers that have some degree of interaction based on their leaky nature ( Figure  10). In most of the Florida BWRO wellfields, the recharge to the production aquifer is from the bottom upwards. The aquifers underlying the production aquifer contain higher salinity water, which tends to cause a long-term increase in salinity (Figure 1). The Lower Hawthorn Aquifer does behave in this manner to a degree in that 40% of the production wells show an upward trend. However, the aquifer also interacts with the overlying Hawthorn Aquifer Zone 1, which tends to contain less saline water. Hawthorn Aquifer Zone 1 contains fresher water towards the northeast (Figure 8), which corresponds with the regional flow direction as defined in CDM [32]. Since the pumping occurs in both aquifers the upwards movement of higher salinity water into the Lower Hawthorn Aquifer is offset by the production of less saline water from Hawthorn Aquifer Zone 1. In addition, the upper aquifer does capture some of the lower salinity water from the regional flow.
In the case of Hawthorn Aquifer Zone 1, some recharge of the aquifer moves downward across the upper confining unit from the Sandstone Aquifer. Therefore, the conceptual model that fits the water-quality behavior of the production wells involves the interaction between a lower high salinity source, the Suwannee Aquifer System, and a freshwater source, the Sandstone Aquifer, with additional interaction across the confining unit between the two production aquifers. It appears that the influx of freshwater into the twoaquifer system from the top is the predominant factor controlling long-term change in water quality. Therefore, the suggested conceptual model is as shown in Figure 10. This is a new conceptual model not found in any previous evaluation of BWRO source-water aquifers. Based on the data collected from the production wells, the annual rate of change in dissolved chloride concentration in Hawthorn Aquifer Zone 1 is an average decrease of about 17 mg/L/year and the Lower Hawthorn Aquifer average increase is 1 mg/L/yr. The combined change based on relative aquifer use is a 16 mg/L/yr decline. The operational data indicate that the water quality is very stable and occurs just below the lower range of the shaded area of the solute-transport model projection. However, the trend in concentration is generally down and not up.
In general, the modeling results predicted a relatively minor increase of dissolved chloride concentration change over a 20-year period and this seems to be reasonably accurate. However, the model over-predicted the increase of chloride concentration at this wellfield. The field data indicate the dissolved chloride concentration is decreasing in 85% of the wells.
A possible explanation of the discrepancy between the model prediction and actual water quality data may be that the model under-estimated the leakance from above, from the Sandstone Aquifer, which has a low chloride concentration. The model had five layers, and the topmost layer is Hawthorn Zone 1 aquifer. Per MODFLOW/SEAWAT convention [37], the model had a no-flow boundary condition above this aquifer by default. If this default assumption is invalid, the aquifer would receive some recharge, likely of better quality, from the overlying Intermediate Aquifer System (Sandstone Aquifer). Another possibility is that fresher water within Hawthorn Aquifer Zone 1 is being captured and causing some reduction is the concentration (Figure 8). Hawthorn Aquifer Zone 1 contains freshwater water to the north of South BWRO wellfield. This option does not seem to be supported by the water quality data, which shows rather uniform decrease in chloride concentration. If the capture mechanism would be the predominant process, then the northeast part of the wellfield would have the highest reductions in concentration. Therefore, the actual chloride concentration may become more stable than the trends predicted by the model.

Conceptual Model of the Aquifer System Based on the Changes in Salinity
The conceptual model for the South Collier BWRO wellfield is different compared to the typical conceptual model used to evaluate most aquifers utilized to produce feed water in Florida and in other parts of the United States [19][20][21][22][23][24][25]. Nearly all the BRWO plants wellfields in the United States pump brackish water from a single aquifer. The South Collier County Wellfield has unique hydrogeology that allows water to be pumped from two different aquifers that have some degree of interaction based on their leaky nature ( Figure 10). In most of the Florida BWRO wellfields, the recharge to the production aquifer is from the bottom upwards. The aquifers underlying the production aquifer contain higher salinity water, which tends to cause a long-term increase in salinity (Figure 1). The Lower Hawthorn Aquifer does behave in this manner to a degree in that 40% of the production wells show an upward trend. However, the aquifer also interacts with the overlying Hawthorn Aquifer Zone 1, which tends to contain less saline water. Hawthorn Aquifer Zone 1 contains fresher water towards the northeast (Figure 8), which corresponds with the regional flow direction as defined in CDM [32]. Since the pumping occurs in both aquifers the upwards movement of higher salinity water into the Lower Hawthorn Aquifer is offset by the production of less saline water from Hawthorn Aquifer Zone 1. In addition, the upper aquifer does capture some of the lower salinity water from the regional flow.

Implications of the Groundwater Quality Data for the Future Operation of the Desalination Facility
A key aspect of this research shows that the design of the South Collier County BWRO plant is sufficiently robust to be able to treat the feed water produced from the wellfield using a two-aquifer system over its projected lifespan. The plant can treat raw water with a TDS concentration of 8000 to 9000 mg/L or a dissolved chloride concentration of 4400 to 5000 mg/L. Based on the monitored water quality in the two aquifers, and the monitored blended feedwater for two recent years (Figure 11), the BRWO plant will be able to treat the raw groundwater for its 20-year life cycle without undergoing any major changes in process design. During the 2019-2020 period, the average dissolved chloride concentration of the composite water into the plant was about 2585 mg/L or a TDS concentration of about 4700 mg/L. Based on the well data projection, the dissolved chloride concentration at 20 years into the future will likely be close to 2400 mg/L or a TDS concentration of about 4360 mg/L.  In the case of Hawthorn Aquifer Zone 1, some recharge of the aquifer moves downward across the upper confining unit from the Sandstone Aquifer. Therefore, the conceptual model that fits the water-quality behavior of the production wells involves the interaction between a lower high salinity source, the Suwannee Aquifer System, and a freshwater source, the Sandstone Aquifer, with additional interaction across the confining unit between the two production aquifers. It appears that the influx of freshwater into the two-aquifer system from the top is the predominant factor controlling long-term change in water quality. Therefore, the suggested conceptual model is as shown in Figure 10. This is a new conceptual model not found in any previous evaluation of BWRO source-water aquifers.

Implications of the Groundwater Quality Data for the Future Operation of the Desalination Facility
A key aspect of this research shows that the design of the South Collier County BWRO plant is sufficiently robust to be able to treat the feed water produced from the wellfield using a two-aquifer system over its projected lifespan. The plant can treat raw water with a TDS concentration of 8000 to 9000 mg/L or a dissolved chloride concentration of 4400 to 5000 mg/L. Based on the monitored water quality in the two aquifers, and the monitored blended feedwater for two recent years (Figure 11), the BRWO plant will be able to treat the raw groundwater for its 20-year life cycle without undergoing any major changes in process design. During the 2019-2020 period, the average dissolved chloride concentration of the composite water into the plant was about 2585 mg/L or a TDS concentration of about 4700 mg/L. Based on the well data projection, the dissolved chloride concentration at 20 years into the future will likely be close to 2400 mg/L or a TDS concentration of about 4360 mg/L.

Implications for the Future Operation of the BWRO Plant
Since the plant has been designed during the expansion to treat a TDS range of 8000 to 9000 mg/L, the projected reduction of the TDS in the future will allow the plant to operate without any modification over its life expectancy. In fact, the plant should be able to increase the blend of raw water into the finished water, therefore reducing future water treatment cost.

Conclusions
Most brackish-water reverse osmosis (BWRO) desalination plants use a groundwater source that tends to increase in salinity with time based on aquifer hydraulics and the mechanism of recharge during pumping. The reason that the salinity of the feedwater increases is the use of a single, leaky aquifer into which water recharges by upwards movement from a deeper aquifer that contains higher salinity water. The key operational issue is that the TDS of the feedwater must not exceed the ability of the RO process design to treat the water to potable standards during the life-expectancy of the facility.
The Collier County Florida South BWRO plant has a capacity of 75,758 m 3 /d, which requires about 101,023 m 3 /d of raw water to be pumped from an aquifer system based an approximate 70% conversion rate. The wellfield taps two different aquifers that contain brackish water. The original groundwater solute-transport model predicted a moderate to low rate of dissolved chloride increase with a mid-range of 35 mg/L/yr based on the conventional upward recharge conceptual model. The high range of change was 85 mg/L/yr over a 20-year period, while the low range was 5 mg/L/yr. The model was constructed with a no flow boundary placed at the top of the first brackish-water aquifer with no freshwater inflow simulated to occur. The actual operational data show a decrease in

Milliseimons
Feed Water Conductivity Mid Day Swing Figure 11. Conductivity of the feed water for 2019 and 2020. The mean is near 9400 millisiemens, which is an estimated TDS of 4700 mg/L or a dissolved chloride value of about 2585 mg/L.

Implications for the Future Operation of the BWRO Plant
Since the plant has been designed during the expansion to treat a TDS range of 8000 to 9000 mg/L, the projected reduction of the TDS in the future will allow the plant to operate without any modification over its life expectancy. In fact, the plant should be able to increase the blend of raw water into the finished water, therefore reducing future water treatment cost.

Conclusions
Most brackish-water reverse osmosis (BWRO) desalination plants use a groundwater source that tends to increase in salinity with time based on aquifer hydraulics and the mechanism of recharge during pumping. The reason that the salinity of the feedwater increases is the use of a single, leaky aquifer into which water recharges by upwards movement from a deeper aquifer that contains higher salinity water. The key operational issue is that the TDS of the feedwater must not exceed the ability of the RO process design to treat the water to potable standards during the life-expectancy of the facility.
The Collier County Florida South BWRO plant has a capacity of 75,758 m 3 /d, which requires about 101,023 m 3 /d of raw water to be pumped from an aquifer system based an approximate 70% conversion rate. The wellfield taps two different aquifers that contain brackish water. The original groundwater solute-transport model predicted a moderate to low rate of dissolved chloride increase with a mid-range of 35 mg/L/yr based on the conventional upward recharge conceptual model. The high range of change was Water 2021, 13,1951 23 of 24 85 mg/L/yr over a 20-year period, while the low range was 5 mg/L/yr. The model was constructed with a no flow boundary placed at the top of the first brackish-water aquifer with no freshwater inflow simulated to occur. The actual operational data show a decrease in the average chloride concentration of the feedwater of about 16 mg/L/yr over the past 11 to 13.5 years and regression analysis suggests that the decrease will continue to occur. The key finding is that the uppermost confining unit is sufficiently leaky to allow some downward movement of freshwater to cause a slight downward trend in chloride concentration.
This case study produced a new conceptual model for evaluation of a two-aquifer wellfield with complex vertical water movement induced by pumping. This is a rare case where a groundwater solute-transport model could be audited to assess if the model is accurate over a long time-period and what changes in model structure would be required to make it more accurate.