Large-Scale Surface Water - Groundwater Exchange Processes in a Karst Aquifer

MISSING IMAGE

Material Information

Title:
Large-Scale Surface Water - Groundwater Exchange Processes in a Karst Aquifer Examples from the Suwannee River Basin
Physical Description:
1 online resource (88 p.)
Language:
english
Creator:
Sutton, James E
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Geology, Geological Sciences
Committee Chair:
Screaton, Elizabeth Jane
Committee Members:
Adams, Peter N
Martin, Jonathan Bowman

Subjects

Subjects / Keywords:
floridan -- groundwater -- hydrology -- karst
Geological Sciences -- Dissertations, Academic -- UF
Genre:
Geology thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Due to increasing demands on fresh water for societal, commercial, and agricultural uses, quantifying water resources is essential for regulating water withdrawals and insuring long term availability. The objective of this study is to understand surface water - groundwater interactions in the upper Floridan aquifer in the Suwannee River basin in north-central Florida.  To achieve this objective, river discharge data was analyzed and a regional scale transient groundwater flow model was developed using a 5 year hydrologic record. The one-layer transient groundwater model was calibrated using groundwater levels in monitoring wells from the Suwannee River Water Management District and provides insight to the hydrologic system that previous steady-state models could not show.   Detailed analysis of river fluxes and the model presented indicate that river losses can occur on the Suwannee River at both extreme low flows and during high stage events. These losses were identified by decreases in volumetric fluxes between gaging stations calculated from the river discharge data and the direction of groundwater velocity vectors generated by the groundwater flow model. River losses at extreme low flows can occur in upper portions of the Suwannee River, whereas river losses during high stage events can occur along the entire Suwannee and Santa Fe Rivers. Observed fluxes indicate that adjacent portions of the Suwannee River can have significantly different patterns of gains and losses. These differences are potentially due to the geometry of the river channel and aquifer.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by James E Sutton.
Thesis:
Thesis (M.S.)--University of Florida, 2013.
Local:
Adviser: Screaton, Elizabeth Jane.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2014-08-31

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0046013:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 LAR G E SCALE SURFACE WATER GROUNDWATER EXCHANGE PROCESSES IN A KARST AQUIFER: EXAMPLES FROM THE SUWANNEE RIVER BASIN By JAMES SUTTON A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFI LLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2013

PAGE 2

2 2013 J ames S utton

PAGE 3

3 To my family and friends for their cons tant support and encouragement

PAGE 4

4 ACKNOWLEDGMENTS I w ould like to thank everyone who has supported and encouraged me throughout my time being a college student. I would like to thank my mother for inspiring me to pursue graduate school and to never give up no matter how stressful or difficult lif e may be. I would like to thank my girlfriend Kaitlyn for listening to me vent, even though at times, she had no idea what I was talking about I'd like to thank my advisor, Liz Screaton, for having patience while guiding me through the modeling and teaching me how to be a better scientist I'd like to thank the Estavelles lab group (especially John Ezell and Amy Brown) for letting me help with field work and reminding me that things aren't always as bad as they seem.

PAGE 5

5 TABLE OF CONTENTS P age ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 6 LIST OF FIGURES ................................ ................................ ................................ .......... 7 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 12 2 BACKGROUND ................................ ................................ ................................ ...... 15 Study Area ................................ ................................ ................................ .............. 15 Previous Numerical Modeling ................................ ................................ ................. 16 3 METHODS ................................ ................................ ................................ .............. 22 Numerical Groundwater Flow Modeling ................................ ................................ .. 23 Model Design ................................ ................................ ................................ .......... 24 Steady State and Transient Recharge ................................ ................................ .... 27 Calculation of Groundwater Exchange with the Suwannee River ........................... 28 Model Limitations ................................ ................................ ................................ .... 30 4 RESULTS ................................ ................................ ................................ ............... 39 Steady State Calibration: Heads, Budget and Flow Patterns ................................ 39 Transient Model ................................ ................................ ................................ ...... 40 Surface Water Groundwater Exchange ................................ ................................ 41 Sensitivity Analysis ................................ ................................ ................................ 43 5 DISCUSSION ................................ ................................ ................................ ......... 71 6 CONCLUSIONS ................................ ................................ ................................ ..... 74 APPENDIX: MONITORING WELL HYDROGRAPHS ................................ ................... 75 LIST OF REFERENCES ................................ ................................ ............................... 86 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 88

PAGE 6

6 LIST OF TABLES Table P age 2 1 Summary of hydraulic conductivity, aquifer thickness, and transmissivity values used in previous models. ................................ ................................ ......... 18 3 1 USGS gaging stations used in the model. ................................ .......................... 31 3 2 Summary of values used for river reach inputs. ................................ .................. 31 4 1 Na mes of SRWMD monitoring wells used for calibration. ................................ .. 44 4 2 Summary of hydraulic conductivity values used in the models. .......................... 44 4 3 T otal gains and losses from r eaches 1 4 during transient time period normalized for length of each reach. ................................ ................................ .. 45 4 4 Results of sensitivity analysis ................................ ................................ ............ 45

PAGE 7

7 LIST OF FIGURES Figure P age 2 1 Location map. ................................ ................................ ................................ ..... 19 2 2 Map showing sub basins of the Suwannee River basin and spring locations within Suwannee County. ................................ ................................ ................... 20 2 3 Potentiometric surface map from September 2009.. ................................ .......... 21 3 1 Location of monitoring wells used for calibration, USGS gaging stations and station for precipitation and evapotranspiration data ................................ .......... 32 3 2 Model grid with river reaches and Cody Scarp. ................................ .................. 33 3 3 Suwannee River stage at USGS gaging stations. ................................ .............. 34 3 4 Ichet ucknee and Santa Fe River stage at USGS gaging stations ....................... 34 3 5 Discharge from USGS gaging stations on Alapaha, Withlacoochee, and Suwannee Rivers ................................ ................................ ............................... 35 3 6 Precipitation at Live Oak and computed unconfined recharge during transient time period ................................ ................................ ................................ .......... 35 3 7 Precipitation and estimated recharge at Live Oak during steady state tim e period. ................................ ................................ ................................ ................ 36 3 8 Relationship between stage and cross sectional area at the USGS Ellavile gaging station. ................................ ................................ ................................ .... 36 3 9 Relationship be tween stage and cross sectional area at the USGS Dowling Park gaging station. ................................ ................................ ............................ 37 3 1 0 Relationship between stage and cross sectional area at the USGS Luraville gaging station. ................................ ................................ ................................ .... 37 3 1 1 Relationship between stage and cross sectional area at the USGS Branford gaging station. ................................ ................................ ................................ .... 38 4 1 Steady state model hydraulic conductivit y (K) zones ................................ ......... 46 4 2 Simulated versus observed heads during steady state time period. ................... 47 4 3 Steady state head residuals at target monitoring wells ................................ ...... 48 4 4 Steady stat e model potentiometric contours and velocity vectors. ..................... 49

PAGE 8

8 4 5 Simulated versus observed heads at target monitoring wells during transient time period. ................................ ................................ ................................ ......... 50 4 6 Map showing root mean square error at all 23 monitoring wells during transient calibration. ................................ ................................ ........................... 51 4 7 Well hydrograph representing a "good" match between observed and computed heads. ................................ ................................ ................................ 52 4 8 Well hydrograph representing a "mediocre" match between observed and computed heads. ................................ ................................ ................................ 52 4 9 Well hydrograph representing a "poor" match between observed and computed heads. ................................ ................................ ................................ 53 4 10 Observed fluxes in r each 1 with corrections for channel storage and groundwater contribution area. ................................ ................................ ........... 53 4 11 Observed fluxes in r each 1 with corrections for channel storage and groundwater contribution area (e xpanded scale). ................................ ............... 54 4 12 Observed fluxes in r each 2 with corrections for channel storage and groundwater contribution area. ................................ ................................ ........... 54 4 13 Observed fluxes in r each 2 with corrections for channel storage and groundwater contribution area (expanded scale). ................................ ............... 55 4 14 Observed fluxes in r each 3 with corrections for channel storage a nd groundwater contribution area. ................................ ................................ ........... 55 4 15 Observed fluxes in r each 3 with corrections for channel storage and groundwater contribution area (expanded scale). ................................ ............... 56 4 16 Observed fluxes in r each 4 with corrections for channel storage and groundwater contribution area. ................................ ................................ ........... 56 4 17 Observed fluxes in r each 4 with corrections for channel storage and groundwater contribution area (expanded scale). ................................ ............... 57 4 1 8 Head difference between well G and Suwannee River at Dowli ng Park and flux at r each 2. ................................ ................................ ................................ .... 57 4 19 Obse rved and simulated fluxes from r each 1. ................................ .................... 58 4 20 Obse rved and simulated fluxes from r each 1 (expanded scale). ........................ 59 4 21 Observed and s imulated f luxes from r each 2. ................................ .................... 60 4 22 Obse rved and simulated fluxes from r each 2 (expanded scale). ........................ 60

PAGE 9

9 4 23 Obse rved and simulated fluxes from r each 3. ................................ .................... 61 4 24 Obse rved and simulated fluxes from r each 3 (expanded scale). ........................ 62 4 25 Obse rved and simulated fluxes from r each 4. ................................ .................... 63 4 26 Obse rved and simulated fluxes from r each 4 (expanded scale). ........................ 64 4 27 Simulated potentiometric contours and velocity vectors from 12/12/ 2007 showing river losses from r each 1 during a low flow period. ............................... 65 4 28 Simulated potentiometric contours and velocity vectors from 1/21/ 2012 showing river losses fr om r each 1 during a low flow period ................................ 66 4 29 Simulated potentiometric contours and velocity vectors from 3/4/2008 showing river losses du ring a high flow period ................................ ................... 67 4 30 Simulated potentiometric contours and velocity vectors from 4/13/2009, showing river losses during a high flow period. ................................ .................. 68 4 31 Simulated potentiometric contours and velocity vectors from 2/3/2010, showing river losses during a high flow period ................................ ................... 69 4 32 Simulated potentiometric contours an d velocity vectors from 7/1/2012, showing river losses during a high flow period ................................ ................... 70

PAGE 10

10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requiremen ts for the Degree of Master of Science LAR G E SCALE SURFACE WATER GROUNDWATER EXCHANGE PROCESSES IN A KARST AQUIFER: EXAMPLES FROM THE SUWANNEE RIVER BASIN By James Sutton August 2013 Chair: Elizabeth Screaton Major: Geology Due to increasing demands on fresh water for societal, commercial, and agricultural uses, quantifying water resources is essential for regulating water withdrawals and insuring long term availability. The objective of this study is to understand surface water groundwater interac tions in the upper Floridan aquifer in the Suwannee River basin in north central Florida. To achieve this objective, river discharge data was analyzed and a regional scale transient groundwater flow model was developed using a 5 year hydrologic record. Th e one layer transient groundwater model was calibrated using groundwater levels in monitoring wells from the Suwannee River Water Management District and provides insight to the hydrologic system that previous steady state models could not show. Detailed analysis of river fluxes and the model presented indicate that river losses can occur on the Suwannee River at both extreme low flows and during high stage events. These losses were identified by decreases in volumetric fluxes between gaging stations calcu lated from the river discharge data and the direction of groundwater velocity vectors generated by the groundwater flow model. River losses at extreme low flows can occur in upper portions of the Suwannee River, whereas river

PAGE 11

11 losses during high stage event s can occur along the entire Suwannee and Santa Fe Rivers. Observed fluxes indicate that adjacent portions of the Suwannee River can have significantly different patterns of gains and losses. These differe nces are potentially due to the geometry of the riv er channel and aquifer.

PAGE 12

12 CHAPTER 1 INTRODUCTION The magnitude of exchange between surface water and groundwater sources can vary spatially and temporally, depending on changes in climate, surface water levels, and aquifer properties. In north central Flor ida, the upper Floridan aquifer has regions of both confined and unconfined conditions which are separated by the Cody Scarp (Upchurch, 2002). At the transition from confined to unconfined conditions, streams such as the Suwannee River can lose water to th e subsurface with relative ease due to the high permeability of the upper Floridan aquifer, which is an eogenetic karst aquifer that underlies much of the southeastern United States. Eogenetic karst aquifers such as the upper Floridan aquifer retain relati vely high primary porosity and permeability compared to telogenetic karst aquifers because of the lack of burial diagenesis (Vacher and Mylroie, 2002). The Suwannee River stage is very dynamic at the transition from confined to unconfined conditions, with stage varying from 8.7 m above sea level at low flow to 19.5 m above sea level between April 2007 and August 2012. Because of the close interaction between the river and the unconfined portion of the aquifer, groundwater levels are also variable. Current issues facing the Suwannee River basin include recent extreme low river and springs flows, as well as recent large floods. In addition spring water quality has been impacted by nitrate pollution, growth of algae, and decreases in biodiversity (Katz, 2004). Downstream of the transition from confined to unconfined conditions, the Suwannee River overall gains water throughout most of the year, but river water can be lost to the aquifer during high flows. Geochemical evidence of river losses during floods

PAGE 13

13 (ba nk storage) has been documented along the Suwannee River at Little River Springs (Crandall et al. 1999; Gulley et al. 2011). In March to April 1996, during period of sustained rainfall, monitoring wells were sampled along the Suwannee River and data showed a decrease in concentrations of calcium, radon, and bicarbonate and an increase in concentrations of dissolved organic carbon, tannic acid and chloride (Crandall et al., 1999). These changes suggested that river water had infiltrated the aquifer. Solution scallops in the Little River Springs cave system provide physical evidence for water flow directed into springs rather than out (Gulley et al., 2011). Gulley et al. (2011) suggested that these river losses into springs contribute to dissolution and condui t development. Ball et al (2012) used discharge data from gaging stations along the Suwannee River to estimate river water losses during flood events, but did not correct losses for changes in channel storage. Despite the dynamic nature of the river a quifer system in the Suwannee River Basin, most of the previous modeling work in the area has focused on steady state si mulations (e.g. Planert, 2007; Schneider, 2008). An exception to this was Grubbs and Crandall (2007) who modeled aquifer river exchanges in the lower Suwannee River basin and found that transient modeling of this area proved to be useful in understanding river aquifer exchanges with changes in groundwater withdrawals from the upper Floridan aquifer and lowered flows in the Suwannee River. More recently, Spellman (2012) simulated the influence of a conduit system at Madison Blue springs on losses of river water during floods. In this project, I analyzed Suwannee River discharge data to better quantify gains and losses and developed a regi onal scale, long term transient groundwater flow model

PAGE 14

14 simulating surface water groundwater exchanges between the upper Floridan aquifer and the Suwannee River in Suwannee County, Florida. Due to the dynamic nature of river aquifer exchanges in karst e nvironments, numerical modeling could be an effective tool to help understand the impact of changes in these exchanges and their effects on aquifer water quality and karstification.

PAGE 15

15 CHAPTER 2 BACKGROUND Study Area North central Florida and much of the s outheastern region of the United States is underlain by the Floridan aquifer system, an Oligocene to Eocene carbonate sequence that has not been deeply buried. As a result, the upper Floridan aquifer has high primary porosity as well as the development of secondary porosity in the form of fractures and dissolution features. The upper Floridan aquifer consists of the Suwannee limestone (10 30 m in thickness) overlying the Ocala limestone (20 80 m in thickness). Above the limestone of the Floridan aquifer sys tem, the Hawthorn Group is a late Oligocene to Pliocene formation that is approximately 0 40 m thick. The Cody Scarp is the erosional edge of the Hawthorn Group, which acts as the confining unit of the upper Floridan aquifer where it exists. There is 0 10 m of undifferentiated Quaternary sediments at the surface. To the west of the Cody Scarp, the upper Floridan aquifer is unconfined and drainage occurs primarily in the subsurface (Figure 2 1). The Suwannee River originates in southern Georgia and flows to the Gulf of Mexico in north central Florida. The study area in this project focuses on Suwannee County within the middle portion of the Suwannee River basin. Suwannee County contains approximately 60 springs with 2 being first magnitude springs (Figure 2 2). River aquifer interactions in the Peacock cave system are currently being studied, and the large scale modeling efforts in this project will provide context for those investigations. A 2009 potentiometric surface map shows the general flow of groundwater in the region to the Suwannee River (Figure 2 3). In the eastern areas of Suwannee County

PAGE 16

16 groundwater, hydraulic heads reach approximately 18 m and decrease to approximately 6 m at the confluence of the Suwannee and Santa Fe Rivers. Average annual precipitation in the area is 1.3 to 1.5 m per year with approximately half of this rainfall occurring in summer months. Precipitation in the summer is usually generated by thunderstorms resulting in intense, localized rainfall. Precipitation occurr ing in fall and winter months is generally associated with tropical storms and the passage of cold fronts and can have less of a localized effect. Previous work (Florea et al. 2007; Ritorto et al., 2009) documented that the fall and winter storms have a mu ch greater contribution to recharge than summer thunderstorms due to lower losses to evapotranspiration. Development in the Suwannee County region is sparse, with no major cities and much of the industry in the area being agriculture related. Marella (200 4) estimated groundwater withdrawals in each county in Florida based on hydrologic data from 2000. The total amount of groundwater withdrawal from the Floridan aquifer system in 2000 in Suwannee County was estimated to be 99,000 m/ day and Columbia County was estimated to be 51,000 m/ day. The total amount of groundwater withdrawal from agricultural practices for Suwannee County was estimated to be 78,000 m/ day and Columbia County was estimated to be 22,000 m/ day. The percentage of groundwater withdra wals for agricultural use out of the total estimated groundwater withdrawals for Suwannee County is 78% and for Columbia County is 42%. Previous Numerical M odeling Bush and Johnston (1988) used a numerical groundwater flow model to investigate the effects of development on the regional Floridan aquifer system. Predevelopment groundwater levels were generated by the numerical groundwater flow

PAGE 17

17 model and were compared to observed 1980 groundwater levels in the Floridan aquifer. Transmissivity values used in t he Bush and Johnston (1988) model in the Suwannee County region range from 76,000 305,000 m / day. Schneider et al (2008) used a three dimensional, five layer steady state numerical model to evaluate hydrogeologic conditions of the Floridan Aquifer Syst em within the Suwannee River Water Management District boundaries and surrounding areas. The model was intended to establish average groundwater levels and provide a tool for the SRWMD to make decisions on future consumptive use permits. The calibrated hyd raulic conductivity for the UFA has a median value of 488 m/ day and a median transmissivity (hydraulic conductivity times aquifer thickness) of approximately 56,000 m/ day. Calibrated values of hydraulic conductivity were approximately 150 m/ day in nort hern Suwannee County, 1675 m/ day in the Peacock Springs region, and 2750 m/ day in the area near the Ichetucknee River. Planert (2007) created a one layer, steady state regional groundwater water flow model simulating surface water groundwater exchange between surface water features and the upper Floridan aquifer under low flow conditions during September 1990. The calibrated transmissivity from the model presented in the study was approximately 140,000 m/ day for the entire Suwannee County region incl uding both confined and unconfined portions of the upper Floridan aquifer. Transmissivity values from aquifer tests conducted in Suwannee County 28,000 m/ day near the Peacock Springs region and 42,000 m/ day just north of the USGS Branford gaging statio n (Planert, 2007).

PAGE 18

18 Although it does not cover the study area, modeling by Grubbs and Crandall (2007) provides one of the few transient models in the Suwannee River basin. Grubbs and Crandall (2007) developed a one layer transient numerical groundwater fl ow model to simulate surface water and groundwater exchange between the Lower Suwannee River and the upper Floridan aquifer. The model used three years of surface water and groundwater data in order to simulate hydrogeologic conditions within the upper Flo ridan aquifer under four water withdrawal scenarios. The model calibration yielded a transmissivity of approximately 13,200 m/ day in the area of the confluence of the Suwannee River and the Santa Fe River. The analysis of river and spring discharge data showed that some portions of the lower Suwannee River receives more groundwater flux than others and this change in flux coincided with large magnitude springs. Groundwater discharge simulated in the study showed that fluxes to the Suwannee River decrease when hydraulic gradient between the river and the aquifer is low and fluxes were increased shortly after flood peaks. A summary of hydraulic conductivity, aquifer thickness, and transmissivity values used in the previous numerical groundwater flow mod els mentioned above and in this study are shown in Table 2 1. Table 2 1 Summary of hydraulic conductivity, aquifer thickness, and transmissivity values used in previous models. Groundwater Flow Model K (m/ day) Aquifer thickness (m ) T (m/day) Bush and Johnston (1988) N/A N/A 76,000 305,000 Grubbs and Crandall (2007) 32 410 13,200 Planert (2007) 460 305 140,000 Schneider et al. (2008) 150 2,750 115 17,000 316,000 Sutton (2013) 248 100 24,800

PAGE 19

19 Figure 2 1. Locati on map with counties labeled; USG S stations used in the model: Alapaha River (red dot), Withlacoochee River (bl ue dot), Suwannee River (green dots), Ichetucknee River (pink dots), Santa Fe River (yellow dot); Peacock cave system (red lines behind Luraville label); and Cody Scarp in Suwannee County. Georgia Cody Scarp Peacock Springs

PAGE 20

20 Figure 2 2. Map showing sub basins of the Suwannee Riv er basin and spring locations (blue triangles) within Suwannee County.

PAGE 21

21 Figure 2 3. Potentiometric surface map from September 2009. Blue lines in feet above sea level with approximate m shown on labels. E astern boundary is shown in red. 15 m 12 m 9 m 6 m

PAGE 22

22 CHAPTER 3 METHODS The modeling focused on the Suwannee River Basin in Suwannee County (Figure 3 1 and 3 2 ). The time period used for the transient simulation was from 10/7/2007 8/23/2012 and was chosen to examine aquifer conditions and exchanges between the Suwannee River and upper Floridan aquifer. This time period contained approximately 6 major fluctuations in river stage and discharge (Figures 3 3 t o 3 5 ). Precipitation during the transient time period totaled approximately 5.88 m and averaged 0.003 m/ day (F igure 3 6) The transient model consisted of 1783 daily stress periods in which recharge and river stage were varied. Daily stage data from the Suwannee, Santa Fe, and Ichetucknee Rivers were collected from 10 gaging stations (Table 3 1) monitored by the United States Geological Survey (USGS) and accessed through their Water Watch Network ( waterwatch.usgs.gov ). Prior to the transient simulation, a steady state simulation was conducted for initial calibration of hydraulic conductivity values and to provide initial conditions for the transient model. The input data for the steady state model included river stage and estimated recharge from 4/23/2007 10/6/2007. T his time period was chosen due to its having the smallest changes in Suwannee River stage. The river stage for the steady state model was averaged over 4/23/2007 10/6/2007 at each USGS gaging station. The average stage at each gaging station was used to interpolate river stages along each river reach. Groundwater levels from the 23 monitoring wells during the steady state period were used to calibrate the model through comparison of simulated and

PAGE 23

23 observed hydraulic heads. Precipitation during the steady s tate time period totaled approximately 0.5 m and averaged 0.003 m/ day (Figure 3 7 ). Numerical Groundwater Flow Modeling The groundwater flow modeling in this study was conducted with MODFLOW, a three dimensional finite difference code created by the U nited States Geological Survey (McDonald and Harbough, 1988) and Groundwater Vistas, a graphical user interface for MODFLOW. MODFLOW uses a finite difference approximation of the partial differential equation governing three dimensional, transient groundwa ter flow in an unconfined aquifer: (3 1) Where The above equation is only applicable for a flow with a constant density and viscosity, fully saturated, and laminar flow obeying Darcy's law: (3 2) Where

PAGE 24

24 Model D esign The model area was bounded in the north and west by the Suwannee River, in the south by the Santa Fe River, and the east by the Ichetu cknee River and the groundwater divide surrounding the Suwannee County region. The groundwater divide on the eastern side of the model was delineated from flow velocity vectors in the Schneider et al. (2008) North Florida model. These boundaries are consis tent with divides interpreted from a p otentiometric surface map from 2009 (Figure 2 3) created by the Suwannee River Water Management District. The steady state and transient models consist of one layer to simulate horizontal flow in the upper Flor idan aquifer. No flow boundaries were assigned to all areas outside the delineated regional groundwater divide derived from the North Florida model. The steady state and transient model grid consists of 50 columns and 50 rows with row dimensions of 1500 m and column dimensions of 1250 m for a total of 1468 active cells. The modeled area was separated into 3 zones of hydraulic conductivity with one representing the confined region, and 2 zones with the unconfined region. There were 2 zones of recharge appl ied to the large scale model: zone 1 representing recharge to the confined upper Floridan aquifer and zone 2 representing recharge to the unconfined upper Floridan aquifer. Storage parameters were separated into 2 zones representing the confined and unconf ined portions of the upper Floridan aquifer. A shapefile of the Suwannee River, acquired from the Florida Department of Environmental Protection, was imported from ArcGIS into Groundwater Vistas to aid in the creation of the river boundaries in the model. A shapefile of monitoring well

PAGE 25

25 coordinates was created in ArcGIS and imported to aid in the placement of head targets used for steady state and transient calibration. A shapefile of Florida's physiographic provinces in north central Florida, primarily the boundary between the Northern Highlands and the Gulf Coastal Lowlands, was used to delineate the boundary between the confined and unconfined upper Floridan aquifer. Modeling of the exchange of groundwater with the Suwannee and Santa Fe Rivers used MODF LOW's River package, which calculates the volumetric exchange of river water and the aquifer as a function of the head gradient and the conductance of the river bed sediments: (3 3) Where River bed conductance riv can be expressed as t he product of two terms: 1) vertical hydraulic conductivity of the river bed sediments divided by the thickness of the sediments and 2) the cross sectional area of flow. River bed conductance is calculated as followed: (3 4) Where

PAGE 26

26 Th e Ichetucknee River, which is an entirely spring fed river, was represented using MODFLOW's Drain package. The Drain package is similar to the River package with the exception that flow from the drain into the aquifer is not allowed. The Drain package calc ulates volumetric exchange between the drain and the aquifer as follows: (3 5) Where The Suwannee River was separated into 5 reaches, the Santa Fe River into 2 reaches, and the Ichetucknee River into 2 reaches based on locations of USGS stream gages. River and drain conductance requires inputs of river bed hydraulic conductivity, sedimen t thickness, width, and depth. r each 1 is the only river reach that flows where the UFA is confined and was given different river bed hydraulic conducti vity and sediment thickness values than the other reaches because of the presence of the Hawthorn Group that would reduce exchange. Reaches 2 9 were given the same values of river bed hydraulic conductivity, sediment thickness, and depth. The river

PAGE 27

27 widths of each reach were determined using an average of 4 measurements along each reach in Google Earth (Table 3 2). Steady State and Transient R echarge The unconfined portion of the upper Floridan aquifer does not have any surface runoff and recharge was calcula ted from precipitation and evapotranspiration with consideration of soil moisture storage. An initial estimated soil moisture content of 0.05 cm was assumed and each day precipitation was added and evapotranspiration was subtracted. If the soil moisture r eached 0, no evapotranspiration was subtracted. If soil moisture exceeded an assumed capacity of 0.1 m, the excess (i.e. the difference in the final soil moisture and 0.1) was considered recharge for that day. Precipitation data was collected and accessed through the Florida Automated Weather Network (FAWN) website ( fawn.ifas.ufl.edu ). Evapotranspiration data was also collected from FAWN (Live Oak, FL station) and was estimated using the Penman Monteith method (Mitchell, 2004). Estimated recharge to the unc onfined upper Floridan aquifer totaled approximately 1.44 m and averaged 0.0008 m/ day d uring the transient time period The confined portion of the upper Floridan aquifer in the Suwannee River Basin has surface runoff and thus has reduced r echarge to upper Floridan aquifer compared to the unconfined portion. Grubbs (1997) estimated recharge to the upper Floridan aquifer using a water budget of drainage basins, chloride mass balance, changes in water levels in wells, and base flow separation analysis. From Grubbs (1997), an average ratio of confined to unconfined rec harge was esti mated to be 0.41. This ratio was used to assign recharge to the confined portion of the upper Floridan aquifer. Computed recharge was applied to the transient model i n each of the 1783 daily stress periods.

PAGE 28

28 Estimated recharge for the steady state model was initially calculated using the above method and averaging the calculated recharge over the steady state period (1.02x10 m/day). However, to match observed fluxes to the Suwannee River in r eaches 2, 3, and 4, it was necessary to adjust the recharge to be 0.000471 m/ day for the unconfined upper Floridan aquifer and 0.000198 m/ day for the confined upper Floridan aquifer. This water budget estimate of steady state re charge is similar to the 0.000483 m/ day estimate used by Planert (2007) for the unconfined upper Floridan aquifer. Calculation of Groundwater Exchange with the Suwannee River River discharge data at river gaging stations on the Alapaha, Withlacoochee, an d Suwannee rivers were used to estimate groundwater fluxes to compare to the simulated fluxes generated by the steady state and transient groundwater flow models. The comparison focused on Suwannee River r eaches 1 to 4 (Figure 3 2 ). The difference in disc harge between downstream and upstream gaging stations w ere used to estimate the flux of water contributed to the river between those two stations, termed river pickup (Grubbs and Crandall, 2007). Computing gro undwater inflow or outflow for r each 1 required subtraction of inflow from Suwannee River tributaries, the Withlacoochee and Alapaha Rivers. Due to a period of missing discharge measurem ents at the Withlacoochee River at Lee gaging station (10/ 1/2009 9/30/2011), fluxes in r each 1 could not be computed during this period. To compare estimated observed fluxes from river discharge data a correction must be made to account for the changes in water storage within the river channel and floodplain. This change in channel storage can be added or subtracted fro m the estimated observed fluxes estimated from river discharge data. LiDAR elevation data

PAGE 29

29 was used to calculate channel volume at various elevations at USGS gaging stations Ellaville, Dowling Park, Luraville, and Branford (i.e. the stations representing re aches 1 4). For reaches 2 4, the average storage of the two stations (start and end of reach) was used. LiDAR elevation data was not available for the USGS White Springs gaging station so the channel storage calculated for the Ellaville station was applied to the entire reach Linear interpolation was used to estimate the change in channel storage with a given stage at each station (Figure 3 8 to 3 11 ). This volume change per day was then added to the observed gains/ losses in each reach. The groundwater flow mo del created in this study simulates fluxes to the Suwannee River from e ast side of the Suwannee River. To compare simulated fluxes from the model to estimated observed fluxes, it was necessary to estimate the proportion of groundwater contributed from each side of the Suwannee River. The correction factor was calculated by estimating the area of groundwater contribution to reaches 1 4 from inside and outside the modeled region using the potentiometric surface map in ArcGIS. Areas within the confined portion of the upper Floridan aquifer were multiplied by 0.41, the ratio of confined to unconfined recharge found for the region by Grubbs (1997). The contribution area estimated for r each 1 was underestimated due to a portion of the groundwater coming from Georg ia which was not represented in the potentiometric surface map. Additionally, the groundwater contribution area was assumed to stay constant through time, where in reality, as groundwater levels in the region change, the groundwater contribution to each re ach would change.

PAGE 30

30 Model L imitations The model presented in this study has several limitations in the way it could be used to represent reality. The modeled area contains many conduit systems that were not represented in the model. However, the scale of th e mapped cave systems is small relative to the scale of the modeled area (Figure 2 1) and these conduit systems would have more of a localized, small scale effect on flow patterns. Another limitation of the model is the lack of pumping wells in the simula tion. Marella (2004) estimated total groundwater withdrawals in Suwannee County in 2000 to be approximately 26.14 million gallons per day or 99,000 m/ day The sum of estimated fluxes to reaches 1 through 4 is approximately 960,000 m/ day making the Mar ella (2004) estimated groundwater withdrawals approximately 10.3% of the total mass balance. In addition, observed river stages were used as inputs into the model which would limit its use for predicting future conditions where river stage would change wit h recharge and water withdrawals. This is in contrast to the Grubbs and Crandall (2007) model which simulated river levels using a coupled surface water groundwater model. The model does not represent changes in river bed conductance during changes in riv er stage. The river conductance values do not incorporate changes in river width with changes in stage. The model retains the same river width through time. As a result, this simplification would result in an underestimate of fluxes between the river and t he aquifer at high stages. There are also several places along the Suwannee River that have dry spring runs (i.e. sloughs) that connect the Suwannee River to inland springs (e.g. Peacock Springs and Luraville Spring) that would expedite the flow of river w ater inland at high stages.

PAGE 31

31 Table 3 1 USGS gaging stations used in the model. # Gaging Station Name Gaging Station Number River 1 Alapaha River near Jennings, FL 2317620 Alapaha 2 Withlacoochee River near Lee, FL 2319394 Withlacoochee 3 Suwannee River at White Springs, FL 2315500 Suwannee 4 Suwannee River at Ellaville, FL 2319500 Suwannee 5 Suwannee River at Dowling Park, FL 2319800 Suwannee 6 Suwannee River at Luraville, FL 2320000 Suwannee 7 Suwannee River at Branford, FL 2320500 Suwannee 8 Blue Hole Spring at Hildreth, FL 2322688 Ichetucknee 9 Ichetucknee River at Dampier's Landing near Hildreth, FL 2322698 Ichetucknee 10 Ichetucknee River at Highway 27 near Hildreth, FL 2322700 Ichetucknee 11 Santa Fe River near Hildreth, FL 2322800 Santa Fe Table 3 2 Summary of values used for river reach inputs. Reach 1 2 3 4 5 6 7 8 9 Width (m) 35.7 72.1 81.7 85.9 83.3 13.0 15.9 38.0 39.4 Sediment T hickness (m) 5 1 1 1 1 1 1 1 1 River K (m/ day) 5 200 200 200 200 200 200 200 200 Depth (m) 3 3 3 3 3 3 3 3 3

PAGE 32

32 Figure 3 1 Location of monitoring wells used for calib ration (red dots), USGS gaging stations (green dots) and Live Oak station for precipitation and evapotranspiration data (yellow triangle). Peaco ck Springs

PAGE 33

33 Figure 3 2 Model grid with river reaches and Cody Scarp labeled. Reach 1 Reach 2 Reach 3 Cody Scarp Reach 4 Reach 5 Reach 6 Reach 7 Reach 8 Reach 9

PAGE 34

34 Figure 3 3 Suwannee River stage at USGS gaging stations: 4/23/2007 8/23/2012. Figure 3 4 Ichetucknee and Santa Fe River s tage at USGS gaging stations: 4/23/2007 8/23/2012.

PAGE 35

35 Figur e 3 5 Discharge from USGS gaging stations on Alapaha, Withlacoochee, and Suwannee Rivers: 4/23/2007 8/23/2012. Discharge i s displayed on a logarithmic scale. Figure 3 6 Precipitation at Live Oak and compu ted unconfined recharge during transient tim e period, 10/7/2007 8/23/2012.

PAGE 36

36 Figure 3 7 Precipitation and estimated recharge at Liv e Oak during steady state time period: 4/23/2007 10/6/2007. Figure 3 8 Relationship between stage and cross secti onal area at the USGS Ellavile gaging station.

PAGE 37

37 Figure 3 9 Relationship between stage and cross sect ional area at the USGS Dowling Park gaging station. Figure 3 10 Relationship between stage and cross sectio nal area at the USGS Luraville gaging station.

PAGE 38

38 Figure 3 11 Relationship be tween stage and cr oss sectional area at the USGS Branford gaging station.

PAGE 39

39 CHAPTER 4 RESULTS Steady State Calibration: Heads, Budget, and Flow P atterns The hydraulic conductivity of 3 zones and the riverbed vertical sediment hydraulic conductivity were a djusted manually to minimize the head residuals at the 23 targe t monitoring wells (Figure 3 1 ). The zones of hydraulic conductivity and their values are shown in Figure 4 1. Initially an additional zone of high hydraulic conductivity was included to repr esent the region surrounding the Peacock Springs cave system. During calibration, the zone of high hydraulic conductivity was changed to that of the neighboring zone of hydraulic conductivity because head residuals in both areas were lowest when the two zo nes were assigned approximately the same value. The difference between observed and simulated heads from the steady state calibration range from 1.04m to 3.10m with the residual at 20 wells less than 1 m and 13 wells less than 0.5 m (Figures 4 2 and 4 3 ). The root mean square error using calibrated values of hydraulic conductivity and river conductances was 0.563 m. The highest head residual in the steady state simulation was at Well V nearest to the Ichetucknee River with observed head exceeding simulat ed head by 3.10 m. Other areas containing high head residuals include Well C in the north west region (0.97 m) ,Wells K and P in the central region near the transition from confined to unconfined (0.86 m, 0.89 m), and Well N just downstream of the Peacock Springs region (0.87 m). Wells near the Suwannee River generally had the lowest head residuals with 6 wells having residuals less than 0.25 m and 4 wells less than 0.1 m. Head residuals near the Suwannee River were most likely low because of the strong con nection between river stage and groundwater levels near the river. The northeast area of the model in the

PAGE 40

40 confined portion have head targets as this area did not contain groundwater data that could be accessed from the Suwannee River Water Management Distr ict's Water Data Portal. The simulated steady state potentiometric surface map (Figure 4 4) shows a general flow of groundwater from the north east region of the modeled region to the southwest towards the Suwannee River. Figure 4 4 shows relative ground water velocity and direction as red arrows The increase in the size of the arrows near the Suwannee River indicate s an increase in groundwater velocity. The maximum arrow size in Figure 4 4 corresponds to a velocity of approximately 1 m/s. The simulated p otentiometric surface map generated by the steady state model (Figure 4 4) is similar to the potentiometric surface map from September 2009 in Figure 2 3. Transient M odel The calibration of the transient model was performed manually by varying aquifer hyd raulic conductivity, the vertical hydraulic conductivity of river bed sediment, and specific yield. Hydraulic conductivities of the 3 zones were adjusted if needed according to the root mean square error of heads at all of the 23 monitoring well targets (F igures 4 5 and 4 6 ). The hydraulic conductivity in zone 1 did not change from the steady state to transient model, zone 2 hydraulic conductivity was lowered by approximately 16%, and zone 3 hydraulic conductivity was lowered by approximately 10% relative to the steady state model. Changing the river bed vertical hydraulic conductivity of the river reaches 1 9 provided no improvement to the root mean square error of head residuals. As a result, river bed hydraulic conductivity values for each reach were uncha nged from the steady state model.

PAGE 41

41 Specific yield for the transient model began with an initial estimate of 0.15 based on literature values (Grubbs and Crandall, 2007) and was adjusted to minimize head residuals at individual monitoring well targets (Figu re 4 7 to 4 9 ). The resulting calibrated value of specific yield for the transient model was 0.09. The root mean square error for all wells using calibrated values of hydraulic conductivity, river hydraulic conductivity, and specific yield was 0.658 m. The roo t mean square error at individual wells show that simulated heads at most wells near the Suwannee River match observed heads the best. Wells G,J, O, and U had the best matches throughout the entire simulation with all having RMSE less than 0.2 m. Wells C, P, R, and V had the poorest matches with all having RMSE greater than 1.0 m. Well V, the well closest to the Ichetucknee River, had the worst match of simulated to observed with an RMSE of 2.53 m. Because the northeastern portion of the model area has be en described as confined, a specific storage value of 1.0x10 6 m 1 (storativity of 1.0x10 4 ) was tested with the transient model, but resulted in very high head residuals at nearby monitoring wells. This suggests that the confining unit of the upper Flori dan aquifer (i.e. the Hawthorn Group) may be perforated with karst features such as sinkholes that allow for the aquifer to have a high storage coefficient like an unconfined aquifer rather than a low storage typical of a confined aquifer. Surface Water Groundwater E xchange Figure s 4 10 to 4 17 show observed pickup along each reach, the corrected gains and losses for the river, and the gains and losses adjusted for the estimated proportion of water contributed from or to the modeled region. Head differences b etween monitoring well G and the Suwannee River at Dowling Park provide a useful

PAGE 42

42 check on the computed fluxes (Figure 4 1 8). Corrected r each 2 fluxes are generally consistent with patterns observed in the head differences. Observed and simulated fluxes fro m reaches 1 4 are compared in Figure s 4 1 9 to 4 26 Simulated fluxes in reach 1 generally underestimated observed fluxes throughout the entire simulation. This reach may receive groundwater contributions from aquifers above the upper Floridan aquifer. Reach 2 had the best match of simulated to observed fluxes throughout the entire simulation with the exception of the April 2009 flood where observed gains and losses are underestimated. Reach 2 also contained false "spikes" in simulated gains and losses for reasons which are currently unknown. Reaches 3 and 4 have better matches than r each 1, but the model appears to underestimate gains during the entire simulation at th ese reaches and s how losses in r each 3 where observations indicated that the river gains water. D uring the March 2008 and February 2010 events, the simulated fluxes show a loss of river water, where the observed fluxes show gains to the river (Figure 4 2 3 and 4 2 4 ) Both data and modeling indicate that during extreme low flow periods in December of 2 007 and February of 201 2 the northern most portion of r each 1 is losing water to the aquifer ( Figure s 4 1 9 4 20 4 2 7 and 4 28 ). The red arrows in Figures 4 2 7 and 4 28 represent direction and velocity of groundwater flow The red arrows adjacent to r each 1 p ointing towards the aquifer indicate the flow of river water into the upper Floridan aquifer from the river The model also shows the changes in potentiometric surface and flow patterns following increases of stage of the Suwannee River with the largest ch anges occurring in March 2008, April 2009, February 2010, and July 2012 (Figures 4 2 9 to 4 32 ). During these time periods, water is lost from all 4 simulated

PAGE 43

43 reaches of the Suwannee River. The potentiometric contours during these high flow periods show that ri ver water entering the aquifer can travel until the gradient driving water into the aquifer meets the background head gradient. The contours also show that head gradients during high flow events tend to be the greatest in the western most portion of the mo del (near r eaches 2 and 3) and are least in t he southern most portion (near r each 4). Sensitivity Analysis A sensitivity analysis was performed on simulated groundwater heads and fluxes given changes to the major model parameters (i.e. riverbed hydraulic conductivity, aquifer hydraulic conductivity, specific yield, and recharge). The root mean square error (RMSE) of gro undwater heads and fluxes from r eaches 2 4 were calculated after making changes to the various model parameters and compared to the RMSE of groundwater heads and fluxes using the calibrated model parameters. The results of the sensitivity analysis are shown in Table 4 4 The sensitivity analysis shows that groundwater heads are most sensitive to increases in aquifer hydraulic conductivity and decrease in specific yield. Groundwater fluxes in r each 2 were most sensitive to an increase in specific yield and recharge. Groundwater fluxes in r each 3 were approximately equally sensitive to changes in rive rbed hydraulic conductivity in r each 2 4, spec ific yield, and r echarge. Groundwater fluxes in r each 4 were most sensitive to an increase in aquifer hydraulic conductivity, decrease in specific yield, and less sensitive to changes in rive rbed hydraulic conductivity in r each 1.

PAGE 44

44 Table 4 1 Names of SR WMD monitoring wells used for calibration. Well Name (Full) Well Name (Short) Notes 11232006 A 11420006 B 21231001 C 21322008 D 21335001 E 21516001 F 31105006 G Data collection stops at 1/2012 31232001 H 31232002 I Data collection st ops at 8/2011 31305005 J 31335002 K 41112005 L 41223004 M Data collection stops at 8/2009 41227001 N 41329001 O 41402002 P Only 1 data point between 10/2007 and 10/2008 51201007 Q 51311001 R 51405002 S 51428004 T 61301007 U 61401003 V 61434006 W 61512010 X 61521005 Y Table 4 2. Summary of hydraulic conductivity values used in the models. K: Zone 1 (m/ day) K: Zone 2 (m/ day) K: Zone 3 (m/ day) K: River reach 1 (m/ day) K: River reaches 2 9 (m/ day) Steady st ate 100 310 335 5 200 Transient 100 260 300 5 200

PAGE 45

45 Table 4 3. Total gains and loss es from r eaches 1 4 during transient time period normalized for length of each reach. Reach 1 2 3 4 Total gains 7.46E+03 6.56E+03 2.59E+04 3.07E+04 Total losses 4.2 6E+03 1.06E+04 1.93E+03 5.63E+03 Net 3.19E+03 4.09E+03 2.39E+04 2.50E+04 Losses/ Gains 0.572 1.623 0.075 0.184 Table 4 4. Results of sensitivity analysis. Kv is riverbed hydraulic conductivity, K is aquifer hydraulic conductivity, Sy is specific yield in zones 1 and 2. % of Calibrated RMSE Heads Change in RMSE (Heads) RMSE Flux Reach 2 Change in RMSE (Reach 2) RMSE Flux Reach 3 Change in RMSE (Reach 3) RMSE Flux Reach 4 Change in RMSE (Reach 4) Kv1 500 0.31 0.0008 10.15 0.04 15.99 0.00 13.24 1.31 Kv1 20 0.32 0.0025 10.26 0.06 15.99 0.00 13.23 1.32 Kv2 4 175 0.31 0.0002 11.08 0.88 17.12 1.12 14.55 0.00 Kv2 4 25 0.31 0.0013 11.06 0.87 17.10 1.10 14.58 0.03 K1 3 125 0.38 0.0671 10.33 0.14 16.00 0.01 12.33 2.21 K1 3 75 0.36 0.0500 10.09 0.10 15.99 0.01 14.30 0.25 Sy 166 0.33 0.0160 13.01 2.81 17.58 1.58 13.57 0.98 Sy 55 0.37 0.0576 10.68 0.49 17.38 1.38 16.59 2.04 Recharge 110 0.33 0.0160 11.28 1.08 17.06 1.07 14.29 0.25 Recharge 90 0.33 0.0158 10.92 0.72 17.16 1.16 14.74 0.20

PAGE 46

46 Figure 4 1 Steady state model hydraulic conductivity (K) zones: Zone 1 (blue), Zone 2 (light blue), Zone 3 (purple). Transient K in pa rentheses where changed. River vertical hydraulic conductivity (Kv) at each reach (sh own with red arrows). Kv = 5 m/day Kv = 200 m/day Kv = 200 m/day Kv = 200 m/day Kv = 200 m/day Kv = 200 m/day Kv = 200 m/day Kv = 200 m/day Kv = 200 m/day

PAGE 47

47 Figure 4 2 Simulated versus observed heads during steady state time period.

PAGE 48

48 Figure 4 3 Steady state head residuals at target mo nitoring wells. Blue: Observed heads are greater than simulated heads. Red : Observed heads are less than simulated heads.

PAGE 49

49 Figure 4 4 Steady state model potentiometric contours (in meters) and velocity vectors. Largest vector corresponds to approximately 1 0 m/ day

PAGE 50

50 Figure 4 5 Simulated versus observed heads at target monitoring wells during transient time period.

PAGE 51

51 Figure 4 6 Map showing root mean square error at all 23 monitoring wells during transient calibration.

PAGE 52

52 Figure 4 7 Well hydrograph representing a "go od" match between observed and computed heads. Figure 4 8 Well hydrograph representing a "medioc re" match between observed and computed heads.

PAGE 53

53 Figure 4 9 Well hydrograph representing a "po or" match between observed and computed hea ds. Figure 4 10 Observed fluxes in r each 1 with corre ctions for channel storage and groundwater contribution area.

PAGE 54

54 Figure 4 11 Observed fluxes in r each 1 with corrections for channel storage and groundwater contribution area (expanded scale) Figure 4 12 Observed fluxes in r each 2 with corrections for channel storage and groundwater contribution area

PAGE 55

55 Figure 4 13 Observed fluxes in r each 2 with corrections for channel storage and groundwater contribution area ( expanded scale). Figure 4 14 Observed fluxes in r each 3 with corrections for channel storage and groundwater contribution area.

PAGE 56

56 Figure 4 15 Observed fluxes in r eac h 3 with corrections for channel storage and groundwater contribution area ( expanded scale). Figu re 4 16 Observed fluxes in r each 4 with corrections for channel storage and groundwater contribution area.

PAGE 57

57 Figure 4 17 Observed fluxes in r each 4 with corrections for channel storage and groundwater contribution area ( expanded scale). Figure 4 1 8 Head difference between well G and Suwanne e River at Dowling Park (Blue) and flux at r each 2 (red).

PAGE 58

58 Figure 4 19 Obse rved and simulated fluxes from r each 1.

PAGE 59

59 Figure 4 20 Obse rved and simulated fluxes from r each 1 (expan ded scale).

PAGE 60

60 Figure 4 21 Obse rved and simulated fluxes from r each 2. Figure 4 22 Obse rved and simulated fluxes from r each 2 (expanded scale).

PAGE 61

61 Figure 4 23 Obse rved and simulated fluxes from r each 3.

PAGE 62

62 Figure 4 24 Obse rved and simulated fluxes from r each 3 (expanded scale).

PAGE 63

63 Figure 4 25 Obse rved and simulated fluxes from r each 4.

PAGE 64

64 Figure 4 26 Obse rved and simulated fluxes from r each 4 (expanded scale).

PAGE 65

65 Figure 4 27 Simulated potentiometric contours and ve locity vectors from 12/12/2007 showing riv er losses from r each 1 during a low flow period. Largest vector co rresponds to approximately 1 0 m/ day

PAGE 66

66 Figure 4 28 Si mulated potentiometric contours and v elocity vectors from 1/21/2012 showing riv er losses from r each 1 during a l ow flow period. Largest vector co rresponds to approximately 8 m/ day

PAGE 67

67 Figure 4 29 Simulated potentiometric contours and vel ocity vectors from 3/4/2008 showing river losses during a high flow period. Largest vector corresponds to approx imately 32 m/ day

PAGE 68

68 Figure 4 30 Simulated potentiometric contours and ve locity vectors from 4/13/2009, showing river losses d uring a high flow period. Largest vector corr esponds to approximately 65 m/ day

PAGE 69

69 Figure 4 31 Simulated potentiometric contours and v elocity vectors from 2/3/2010, showing river losses during a high flow period. Largest vector correspo nds to approximately 2 0 m/ day.

PAGE 70

70 Figure 4 32 Simulated potentiometric contours and v elocity vectors from 7/1/2012, showing river losses during a high flow period. Largest vector corresponds to approximately 2 0 m/ day.

PAGE 71

71 CHAPTER 5 DIS CUSSION The high river conductance of reaches 2 9 used in this study (200 m/ day) is similar to that of the adjacent aquifer which suggests that sediments in the river have little effect on the exchanges occurring between the river and the aquifer. The ca librated values of hydraulic conductivity of 300 m/ day in the unconfined upper Floridan aquifer obtained in this study yields a transmissivity of 30,000 m 2 /day. This value is roughly consistent with the transmissivity value in Grubbs and Crandall for the region immediately south of this study area (43,000 m/ day; 2007). Transmissivity values from the model also match transmissivity values in the area from aquifer tests as reported by Planert (28,000 m/ day and 42,000 m/ day; 2007). The hydraulic conduct ivity generated in this study is similar to the value of 150 m/ day used by Schneider (2008) for the northern Suwannee County area where the aquifer is confined, but are significantly less in the southern areas of the county (1675 m/ day near Peacock Sprin gs; 2750 m/ day near the Ichetucknee River). The transmissivity in this study was also significantly less than the value of 140, 000 m/ day used for the Planert (2007) model. Since the hydraulic conductivity used in this study is consistent with aquifer tests performed in the unconfined portion of the modeled area and transmissivity values used by Grubbs and Crandall (2007) in a nearby area, it makes the hydraulic conductivity values calibrated in Schneider (2008) and transmissivity calibrated in Planert (2007) questionable. Transient models provide greater constraints during calibration. A steady state model requires a time period in which the groundwater system is flowing at steady state, that is, the right side of the groundwater flow equation is zero and the

PAGE 72

72 change in storage is negligible. As discovered in this study, choosing an appropriate time period for a steady state model in the Suwannee County area can be difficult due to frequent changes in Suwannee River stage. Schneider (2008) and Planert ( 2007) who also has a difficult time choosing a steady state time period and due to relative frequent fluctuations, chose September 1990 and June 2001 May 2002 respectively. The two previous models (i.e. Planert, 2007); Schneider, 2008) of the upper Flo ridan aquifer in the upper Suwannee River basin have been steady state. However, this study shows that the use of transient modeling provides a more robust estimate of hydraulic conductivity. In addition, transient modeling is vital to illustrate the dynam ic hydrologic conditions of karst aquifers compared to steady state modeling. For example, the simulation of reversed head gradients immediately following intense or prolonged precipitation would not be visible from steady state models, but will have an im pact on water quality in the short term and karstification in the long term. The corrected fluxes from r eaches 1 4 combined with the model results provides insight into the patterns of gains and losses of water from the Suwannee River from the time perio d of 10/7/2007 8/23/2012. Reaches 2, 3, and 4 show approximately 8 times of major losses through the time of the simulation with the maximum observed loss occurring in April 2009. The gains and losses to each reach in Table 4 3 show that r eaches 1, 3, and 4 had a net gain of groundwater from t he upper Floridan aquifer, but r each 2 lost more river water than it gained from the aquifer. Reach 2 has the least amount of gains and the greatest amount of losses per meter of aquifer compared to all the other reach es. Based on the potentiometric surface map (Figure 2 3), much of the groundwater flowin g from north to south bypasses r each 2 during low flow. Also seen in

PAGE 73

73 the simulated potentiometric contours (Figures 4 2 7 and 4 28 ), the area adjacent to r each 2 appears to have the lowest head gradient compared to the other reaches. As seen in the simulated contours from the March 2008 and April 20 09 events, head g radients near r ea ch 2 appear to be greater than r eaches 1, 3, and 4. The observed fluxes show that r each 3 is the only portion of t he Suwannee River in this study to gain water during the 2008 and 2010 events. In contrast, the simulation shows losses during all of the high flow events. Furt her investigation is needed at r each 3 to determine the reason for thes e gains during high flow events and why the model fails to simulate the gains. A possible solution include s locally occurring recharge during these events and flow to that portion of the river from the western side T his study assumed gains and losses from the Suwannee River to each reach were proportional to the contribution area, but in reality this may not be true. River losses at high flows allows for the infiltration of water that is undersaturated with respect to calcite to enter the upper Floridan a quifer, causing the dissolution of carbonate rock (Gulley et al., 2011, 2013). According to the observed fluxes and the simulation in this study, the upper Floridan aquifer adjacent to reaches 2, 3, and 4 would all be susceptible to karstification due to t he magnitude and frequency of river losses at these locations during high flows. Although previous studies (i.e. Gulley, 2013) has shown that water in the Suwannee River at low flows is likely saturated with respect to calcite, people using private wells in these areas should monitor their water quality before consuming.

PAGE 74

74 CHAPTER 6 CONCLUSIONS The corrected flux data and the transient simulation in this study show that exchanges between the upper Floridan aquifer and the Suwannee River can vary through t ime and could occur during both high flows and extreme low flows. The corrections for channel storage made it possible to calculate groundwater gains and losses and compare to model results. The loss of water from the Suwannee River to the upper Floridan a quifer, as shown in the data and the model in this study, indicates the locations that are most vulnerable to water quality changes during high flows and where increased rates of carbonate dissolution could be occurring. Due to the net loss of river water from r each 2 to the upper Floridan aquifer during low and high flows, people using private wells adjacent to the Suwannee River in these areas should be cautious of the quality of their drinking water. The use of transient modeling over multi year time scales in the Suwannee River basin has the potential to be an effective tool to understand the dynamic river aquifer exchanges between the Suwannee River and the upper Floridan aquifer. Steady state modeling of this area may be useful in changes in ground water flow paths over long time periods (such as the movement of a regional groundwater divide), but would not illustrate the changes in flow paths on a smaller time scale. Future modeling studies of the area could look at river aquifer exchanges on longer time scales (decades) to examine how exchange has varied with development or a smaller spatial scale by simulating caves systems ( e.g. MODFLOW's Conduit Flow Processes) and their effects on flow paths and river gains/ losses.

PAGE 75

75 APPENDIX MONITORING WELL HYDR OGRAPHS

PAGE 76

76

PAGE 77

77

PAGE 78

78

PAGE 79

79

PAGE 80

80

PAGE 81

81

PAGE 82

82

PAGE 83

83

PAGE 84

84

PAGE 85

85

PAGE 86

86 LIST OF REFERENCES Ball, C., Martin, J.B., Screaton, E.J., 2012. Effec ts of antecedent hydrogeologic conditions on flood magnitude and recharge to the Floridan Aquifer in north central Florida. Journal of Undergraduate Research. 13 (2), 1 10 Bear, J., 1972. Dynamics of Fluids in Porous Media, New York, Dover Publications Inc. Bud d, D.A., Vacher, H.L., 2004. Matrix permeability of the confined Floridan Aquifer, Fl orida USA. Hydrogeology Journal 12 (5) 531 549 Budd, D.A., Vacher, H.L., 2002. Facies Control On Matrix Permeability In The Upper Floridan Aquifer, West Central Florida: Imp lications To Diffuse Flow. Karst Waters Special Publication. 7, 14 24 Crandall, C.A., Katz, B.G., Hirten, J.J., 1999. Hydrogeochemical evidence for mixing of river water and groundwater during high flow conditions, lower Suwannee River basin, Florida, US A. Hydrogeolog y Journal. 7, 454 467 Grubbs, J.W., 1998. Recharge Rates to the Upper Floridan Aquifer in the Suwannee River Water Management District, Florida. 97 4283, 1 36 Grubbs, J.W., Crandall, C.A., 2007. Exchanges of Water between the Upper Floridan Aquifer and the Lower Suwannee and Lower Santa Fe Rivers, Florida. 1656 C, 1 93. Gulley, J., Martin, J.B., Screaton, E.J., Moore, P.J., 2011. River reversals into karst springs: A model for cave enlargement in eogenetic karst aquifers. Geological Soci ety of America Bulletin. 123, 457 467 Gulley, J., Martin, J.B., Spellman, P., Screaton, L., 2013. Dissolution in a variably confined carbonate platform: effects of allogenic runoff, hydraulic damming of groundwater inputs, and surface groundwater exchange at the basin scale. Earth Surface Processes and Landforms., 1 14 Katz, B.G., Dehan, R.S., Hirten, J.J., Catches, J.S., 1997. Interactions Between Ground Water And Surface Water In The Suwannee River Basin, Florida. Journal Of The American Water Resources Association. 33 (6) 1237 1254 Knighton, D., 1998. Fluvial Forms & Processes: A New Perspective, Great Britain, Hodder Arnold. Mitchell, M.W., 2004. Evaluation of the agricultural field scale irrigation requirement simulation (AFSIRS) in predicting golf cours e irrigation requirements with site specific data, Gainesville, Fla, University of Florida 1 51

PAGE 87

87 Moore, P.J., Martin, J.B., Screaton, E.J., 2009. Geochemical and statistical evidence of recharge, mixing, and controls on spring discharge in an eogenetic karst aquifer. Journal of Hydrology. 376, 443 455 Motz, L.H., Dogan, A., 2004. North Central Florida Active Water Table Regional Groundwater Flow Model. SJ2005 SP16, 1 141 Palmer, A.N., 1991. Origin and morphology of limestone caves. Geological Society of Americ a Bulletin. 103, 1 21 Planert, M., 2007. Simulation of Regional Ground Water Flow in the Suwannee River Basin, Northern Florida and Southern Georgia. 2007 5031, 1 50 Randazzo, A.F., Jones, D.S., 1997. The Geology of Florida, Gainesville, FL, University of Florida. Ritorto, M., Screaton, E.J., Martin, J.B., Moore, P.J., 2009. Relative importance and chemical effects of diffuse and focused recharge in an eogenetic karst aquifer: an example from the unconfined upper Floridan aquifer, USA. Hydrogeology Jour nal. 17, 1687 1698 Schneider, J.W., Upchurch, S.B., Chen, J., Cain, C., 2008. Simulation of Groundwater Flow in North Florida and South Central Georgia 1 92 Screaton, E.J., Martin, J.B., Ginn, B., Smith, L., 2004. Conduit Properties and Karstification in the U nconfined Floridan Aquifer. Ground Water. 42 (3) 338 346 Shoemaker, W.B., O'Reilly, A.M., Sepulveda, N., Williams, S.A., Motz, L.H., Sun, Q., 2004. Comparison of Estimated Areas Contributing Recharge to Selected Springs in North Central Florida by Using Multi ple Ground Water Flow Models. 03 448, 1 31 Sophocleous, M., 2002. Interactions between groundwater and surface water: the state of science. Hydrogeology Journal. 10, 52 67 Vacher, H.L., Mylroie, J.E., 2002. Eogenetic karst from the perspective of an equiva lent porous medium. Carbonates Evaporites. 17, 182 196. White, W.B., 2002. Karst hydrology: recent developments and open questions. Engineering Geology. 65, 85 105 White, W.B., 1969. Conceptual Models for Carbonate Aquifers. Ground Water. 7 (2) 180 186

PAGE 88

88 BIOGRA PHICAL SKETCH James Sutton is a fifth generation resident of Palmetto, FL, a small coastal town approximately 30 miles south of Tampa, FL. His interest in natural science was strong from a young age and grew stronger with time. He rece ived his Bachelor of Science degree in Environmental Science from the University of Florida in 2011 and wanted to continue his education in hydrologic sciences in UF's Department of G eological Sciences He plans to use his knowledge of environmental science and hydrology to help pre serve Florida's ecosystems and water resources.