<%BANNER%>

Spatial and Temporal Dynamics of Chlorophyll and Nutrients in the Suwannee River and Estuary, Florida, USA

Permanent Link: http://ufdc.ufl.edu/UFE0043848/00001

Material Information

Title: Spatial and Temporal Dynamics of Chlorophyll and Nutrients in the Suwannee River and Estuary, Florida, USA
Physical Description: 1 online resource (163 p.)
Language: english
Creator: Krzystan, Andrea M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: chlorophyll -- cusum -- nutrients -- regression -- river -- suwannee -- trend
Forest Resources and Conservation -- Dissertations, Academic -- UF
Genre: Fisheries and Aquatic Sciences thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The ability to differentiate between human induced changes and natural variations in water quality is a key element of effective management. An important step in distinguishing these sources of variation involves characterizing the temporal dynamics of specific water quality parameters, which can be accomplished through analysis of long-term datasets. Two long-term datasets were used in this study to examine water quality dynamics within a system of concern: the Suwannee River and its estuary. The first dataset was used to assess the presence of increasing trends in total nitrogen concentrations along the length of the Suwannee River between 1989 and 2010 using cumulative sum (CUSUM) analyses. CUSUM analyses were applied to different sections of the river delineated based on a moving split-window boundary detection analysis. Results demonstrated increasing total nitrogen concentrations in each of the delineated sections of the river across the study period. However, fine-scale differences in CUSUM statistics highlighted spatial differences in the influence of hydrogeology, land use, groundwater quality, and climatic events on nitrogen dynamics, which suggested the need for an adaptive management approach. A second long-term dataset was used to develop time series and regression models that described relationships between chlorophyll-a concentrations and salinity, temperature, light availability, and concentrations of total nitrogen and total phosphorus within the lower Suwannee River and its estuary from 1998 to 2010. To increase the power to detect relationships, stations were grouped according to results of multivariate ordination, and models were developed using pooled data from 1999-2008. The predictive power of each model was evaluated using data from 1998, 2009, and 2010. As expected, final model parameters varied among groups of stations distributed in the river, oyster reef, and nearshore areas of the system; however, color and total phosphorus concentrations always explained significant amounts of variation in chlorophyll-a values. Chlorophyll-a concentrations exhibited different relationships to model covariates at the station level, although observations at the group level ordinated together. Models for groups of stations fit overall trends at all sites with adjusted R-squared values ranging from 0.34 to 0.72, but the accuracy of predictions differed among sites within each group. Differences in spatial relationships were attributed to variation in the river plume, which can influence top-down grazing pressure or bottom-up influences, such as reduced light availability, due to color. Results of both analyses highlight the overall complexity of the Suwannee system and emphasize the need to consider multiple sources and scales of variation when attempting to distinguish natural variation from human activities leading to impairment or improvement.
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 Andrea M Krzystan.
Thesis: Thesis (M.S.)--University of Florida, 2011.
Local: Adviser: Frazer, Tom K.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2013-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043848:00001

Permanent Link: http://ufdc.ufl.edu/UFE0043848/00001

Material Information

Title: Spatial and Temporal Dynamics of Chlorophyll and Nutrients in the Suwannee River and Estuary, Florida, USA
Physical Description: 1 online resource (163 p.)
Language: english
Creator: Krzystan, Andrea M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: chlorophyll -- cusum -- nutrients -- regression -- river -- suwannee -- trend
Forest Resources and Conservation -- Dissertations, Academic -- UF
Genre: Fisheries and Aquatic Sciences thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The ability to differentiate between human induced changes and natural variations in water quality is a key element of effective management. An important step in distinguishing these sources of variation involves characterizing the temporal dynamics of specific water quality parameters, which can be accomplished through analysis of long-term datasets. Two long-term datasets were used in this study to examine water quality dynamics within a system of concern: the Suwannee River and its estuary. The first dataset was used to assess the presence of increasing trends in total nitrogen concentrations along the length of the Suwannee River between 1989 and 2010 using cumulative sum (CUSUM) analyses. CUSUM analyses were applied to different sections of the river delineated based on a moving split-window boundary detection analysis. Results demonstrated increasing total nitrogen concentrations in each of the delineated sections of the river across the study period. However, fine-scale differences in CUSUM statistics highlighted spatial differences in the influence of hydrogeology, land use, groundwater quality, and climatic events on nitrogen dynamics, which suggested the need for an adaptive management approach. A second long-term dataset was used to develop time series and regression models that described relationships between chlorophyll-a concentrations and salinity, temperature, light availability, and concentrations of total nitrogen and total phosphorus within the lower Suwannee River and its estuary from 1998 to 2010. To increase the power to detect relationships, stations were grouped according to results of multivariate ordination, and models were developed using pooled data from 1999-2008. The predictive power of each model was evaluated using data from 1998, 2009, and 2010. As expected, final model parameters varied among groups of stations distributed in the river, oyster reef, and nearshore areas of the system; however, color and total phosphorus concentrations always explained significant amounts of variation in chlorophyll-a values. Chlorophyll-a concentrations exhibited different relationships to model covariates at the station level, although observations at the group level ordinated together. Models for groups of stations fit overall trends at all sites with adjusted R-squared values ranging from 0.34 to 0.72, but the accuracy of predictions differed among sites within each group. Differences in spatial relationships were attributed to variation in the river plume, which can influence top-down grazing pressure or bottom-up influences, such as reduced light availability, due to color. Results of both analyses highlight the overall complexity of the Suwannee system and emphasize the need to consider multiple sources and scales of variation when attempting to distinguish natural variation from human activities leading to impairment or improvement.
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 Andrea M Krzystan.
Thesis: Thesis (M.S.)--University of Florida, 2011.
Local: Adviser: Frazer, Tom K.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2013-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043848:00001


This item has the following downloads:


Full Text

PAGE 1

1 SPATIAL AND TEMPORAL DYNAMICS OF CHLOROPHYLL AND NUTRIENTS IN THE SUWANNEE RIVER AND ESTUARY, FLORIDA, USA By ANDREA MARIE KRZYSTAN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILL MENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Andrea Marie Krzystan

PAGE 3

3 To my family

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank the members of my supervisory committee for their supp ort and their willingness to share their knowledge and advice. Dr. Chuck Jacoby was a constant ally and I am grateful for his insight into research, scientific writing, and statistics and for the time he has invested into this study and my development as a w to place my research into the bigger picture and his passion has been contagious. Dr. Ed Phlips provided valuable insight into the complexities of the Suwannee system and also showed me the light er side of phytoplankton ecology. Finally, I thank Dr. Tim Fik, whose courses inspired the work comprising this study for demonstrating the importance of thinking outside of the box and keeping things in perspective. I would also like to thank members of the Frazer lab, past and present, for collecting the majority of the water quality data used in this study. In particular I would like to thank Darlene some great memories. Finally, I thank ev eryone in the Fisheries and Aquatic Sciences Program and throughout the university who have assisted me throughout my graduate studies.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ ........ 1 1 LIST OF ABBREVIATIONS ................................ ................................ ........................... 14 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 17 2 NITROGEN TRENDS IN THE SUWANNEE RIVER, FLORIDA ............................. 20 Rationale ................................ ................................ ................................ ................. 20 Methods ................................ ................................ ................................ .................. 22 Study Area ................................ ................................ ................................ ........ 22 Data Descriptio n ................................ ................................ ............................... 23 Trend Analyses ................................ ................................ ................................ 24 Results ................................ ................................ ................................ .................... 26 Boundary Detection ................................ ................................ .......................... 26 Trend Analyses ................................ ................................ ................................ 26 Discussion ................................ ................................ ................................ .............. 27 Trends in Total Nitrogen ................................ ................................ ................... 27 CUSUM Analyses ................................ ................................ ............................. 31 Summary ................................ ................................ ................................ .......... 33 3 CHLOROPHYLL DYNAMICS IN THE LOWER SUWANNEE RIVER AND ESTUARY ................................ ................................ ................................ ............... 40 Rationale ................................ ................................ ................................ ................. 40 Methods ................................ ................................ ................................ .................. 42 Study Area ................................ ................................ ................................ ........ 42 Data Sources ................................ ................................ ................................ .... 44 Trend Analyses ................................ ................................ ................................ 46 Regression Analyses ................................ ................................ ........................ 48 Overview ................................ ................................ ................................ .... 48 Data management ................................ ................................ ..................... 49 Ordinary least squares regressions ................................ ........................... 52 Ti me series models ................................ ................................ .................... 55 Final model evaluation ................................ ................................ ............... 57 Results ................................ ................................ ................................ .................... 58

PAGE 6

6 Trend Analyses ................................ ................................ ................................ 58 Test selection ................................ ................................ ............................. 58 Trends ................................ ................................ ................................ ........ 58 Regression Analyses ................................ ................................ ........................ 59 Data management ................................ ................................ ..................... 59 Model descriptions ................................ ................................ ..................... 62 Model fit ................................ ................................ ................................ ..... 64 Model evaluation ................................ ................................ ........................ 66 Discussion ................................ ................................ ................................ .............. 68 Trend Analyses ................................ ................................ ................................ 68 Regression Analyses ................................ ................................ ........................ 70 Model interpretation ................................ ................................ ................... 70 Evaluation of modeling methods ................................ ................................ 75 Summary ................................ ................................ ................................ .......... 79 4 CONCLUSIONS ................................ ................................ ................................ ... 122 APPENDIX DESCRIPTIVE STATISTICS ................................ ................................ ................ 125 Chlorophyll a ................................ ................................ ................................ ......... 125 Total Nitrogen ................................ ................................ ................................ ....... 130 Total Phosphorus ................................ ................................ ................................ .. 135 Water Color ................................ ................................ ................................ ........... 140 Water Temperature ................................ ................................ ............................... 145 Salinity ................................ ................................ ................................ .................. 150 LIST OF REFERENCES ................................ ................................ ............................. 155 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 163

PAGE 7

7 LIST OF TABLES Table page 2 1 Sampling site details ................................ ................................ ........................... 35 3 1 Supplementary data sources ................................ ................................ .............. 81 3 2 Details and results of trend analyses for total nitrogen by station ....................... 82 3 3 Details and results of trend analyses for total phosphorus by station ................. 83 3 4 Details and results of trend analyses for chlorophyll a by station ....................... 84 3 5 Details and results of trend analyses for color by station ................................ .... 84 3 6 Details and results of trend analyses for salinity by station ................................ 84 3 7 Analysis of variance for model used to estimate missing color values ............... 85 3 8 Analysis of variance for model used to estimate missing salinity values ............ 85 3 9 Analysis of variance for model used to estimate missing total nitrogen values .. 86 3 10 Analysis of variance for model used to estimate missing total phosphorus values ................................ ................................ ................................ ................. 86 3 11 Analysis of variance for model used to estimate missing mean daily discharge values for the Gopher River dataset ................................ ................... 86 3 12 Analysis of variance for group model for stations 1 and 2 ................................ .. 87 3 13 Coefficient estimates for group model for stations 1 and 2 ................................ 88 3 14 Results of Box tests for autocorrelation for station residuals extracted from group models ................................ ................................ ................................ ...... 88 3 15 Analysis of variance for station 4 model ................................ ............................. 89 3 16 Coefficient estimates for station 4 model ................................ ............................ 89 3 17 Analysis of variance for group model for stations 3, 5, and 6 ............................. 89 3 18 Coefficient estimates for group model for stations 3, 5, and 6 ............................ 90 3 19 Analysis of variance for group model for stations 7, 8, 9, and 10 ....................... 90 3 20 Coefficient estimates for group model for stations 7, 8, 9, and 10 ...................... 91

PAGE 8

8 3 21 Root mean squared error values for individual station models calculated for each year ................................ ................................ ................................ ............ 92 A 1 Descriptive statistics for chlorophyll a ( g L 1 ) at station 1 ................................ 125 A 2 Descriptive statistics for chlorophyll a ( g L 1 ) at station 2 ................................ 125 A 3 Descriptive statistics for chlorop hyll a ( g L 1 ) at station 3 ................................ 126 A 4 Descriptive statistics for chlorophyll 1 ) at station 4 ................................ 126 A 5 Descriptive statistics for chlorophyll 1 ) at station 5 ................................ 127 A 6 Descriptive statistics for chlorophyll 1 ) at station 6 ................................ 127 A 7 Descriptive statis tics for chlorophyll 1 ) at station 7 ................................ 128 A 8 Descriptive statistics for chlorophyll 1 ) at station 8 ................................ 128 A 9 Descriptiv e statistics for chlorophyll 1 ) at station 9 ................................ 129 A 10 Descriptive statistics for chlorophyll 1 ) at station 10 .............................. 129 A 11 1) at station 1 .............................. 130 A 12 1 ) at station 2 ............................... 130 A 13 1 ) at station 3 ............................... 131 A 14 1 ) at station 4 ............................... 131 A 15 1 ) at station 5 ............................... 132 A 16 1 ) at station 6 ............................... 132 A 17 1 ) at station 7 ............................... 133 A 18 1 ) at station 8 ............................... 133 A 19 1 ) at station 9 ............................... 134 A 20 1 ) at station 10 ............................. 134 A 21 1 ) at station 1 ......................... 135 A 22 1 ) at st ation 2 ......................... 135 A 23 1 ) at station 3 ......................... 136 A 24 Descriptive statistics for total phosphoru 1 ) at station 4 ......................... 136

PAGE 9

9 A 25 1 ) at station 5 ......................... 137 A 26 Descriptive statistics for 1 ) at station 6 ......................... 137 A 27 1 ) at station 7 ......................... 138 A 28 Descriptiv 1 ) at station 8 ......................... 138 A 29 1 ) at station 9 ......................... 139 A 30 1 ) at station 10 ....................... 139 A 31 Descriptive statistics for color (PCU) at station 1 ................................ .............. 140 A 32 Descriptive statistics for color (PCU) at station 2 ................................ .............. 14 0 A 33 Descriptive statistics for color (PCU) at station 3 ................................ .............. 141 A 34 Des criptive statistics for color (PCU) at station 4 ................................ .............. 141 A 35 Descriptive statistics for color (PCU) at station 5 ................................ .............. 142 A 36 Descriptive sta tistics for color (PCU) at station 6 ................................ .............. 142 A 37 Descriptive statistics for color (PCU) at station 7 ................................ .............. 143 A 38 Descriptive statistics for color (PCU) at station 8 ................................ .............. 143 A 39 Descriptive statistics for color (PCU) at station 9 ................................ .............. 144 A 40 Descriptive statistics for color (PCU) at station 10 ................................ ............ 144 A 41 Descriptive statistics for water temperature (C) at station 1 ............................ 145 A 42 Descriptive statistics for water temper ature (C) at station 2 ............................ 145 A 43 Descriptive statistics for water temperature (C) at station 3 ............................ 146 A 44 Descriptive statistics for wa ter temperature (C) at station 4 ............................ 146 A 45 Descriptive statistics for water temperature (C) at station 5 ............................ 147 A 46 Descriptive statist ics for water temperature (C) at station 6 ............................ 147 A 47 Descriptive statistics for water temperature (C) at station 7 ............................ 148 A 48 Descripti ve statistics for water temperature (C) at station 8 ............................ 148 A 49 Descriptive statistics for water temperature (C) at station 9 ............................ 149

PAGE 10

10 A 50 Descriptive statistics for water temperature (C) at station 10 .......................... 149 A 51 Descriptive statistics for salinity (ppt) at station 1 ................................ ............. 150 A 52 Descriptive statistics for salinity (ppt) at station 2 ................................ ............. 150 A 53 Descriptive statistics for salinity (ppt) at station 3 ................................ ............. 151 A 54 De scriptive statistics for salinity (ppt) at station 4 ................................ ............. 151 A 55 Descriptive statistics for salinity (ppt) at station 5 ................................ ............. 152 A 56 Descript ive statistics for salinity (ppt) at station 6 ................................ ............. 152 A 57 Descriptive statistics for salinity (ppt) at station 7 ................................ ............. 153 A 58 Descriptive st atistics for salinity (ppt) at station 8 ................................ ............. 153 A 59 Descriptive statistics for salinity (ppt) at station 9 ................................ ............. 154 A 60 Descriptive statisti cs for salinity (ppt) at station 1 0 ................................ ........... 154

PAGE 11

11 LIST OF FIGURES Figure page 2 1 Map of s ampling sites and the extent of the Suwannee River Basin .................. 35 2 2 Moving split window boundary detection results ................................ ................. 36 2 3 CUSUM statistics and mean annual total nitrogen concentrations f or station 1 from 1989 t o 2010. ................................ ................................ ............................ 36 2 4 CUSUM statistics and mean annual total nitrogen concentrations for station s 2, 3, and 4 from 1989 to 2010 ................................ ................................ ............. 37 2 5 CUSUM statistics and mean annual total nitrogen concentrations for station s 5, 6, and 7 from 1989 to 2010 ................................ ................................ ............. 38 2 6 CUSUM statistics and mean annual total nitrogen concentrations for station s 8, 9, 10, a nd 11 from 1989 to 2010 ................................ ................................ ..... 39 3 1 Map of the Suwannee River basin with the Withlaco ochee Alapaha River, Upper Suwannee River, Suwannee River and Santa Fe River sub basin s illustrated ................................ ................................ ................................ ............ 81 3 2 Map of w ater qua lity sampling sites with the river, oyster reef, and nearshore regions illustrated ................................ ................................ ............................... 82 3 3 Two dimensional ordination of samp ling stations based on range transformed values of temperature, salinity, color, and concentrations of total nitrogen and total phosphorus.. ................................ ................................ .......... 87 3 4 Time series plot of chlorophyll a concentration 1 ) for stations 1 and 2 ..... 93 3 5 Time series plot of chlorophyll 1 ) for stations 3, 5, and 6 ................................ ................................ ................................ ......................... 93 3 6 Time series plot of chlorophyll 1 ) for station 4. ................ 94 3 7 Time series plot of chlorophyll 1 ) for stations 7, 8, 9, and 10 ................................ ................................ ................................ ................ 94 3 8 Time series plot of total nitrogen concentrations ( g L 1 ) for stations 1 and 2 .... 95 3 9 1 ) for stations 3 5, and 6 ................................ ................................ ................................ ......................... 95 3 10 Time series plot of total nitrogen 1 ) for station 4 ................ 96

PAGE 12

12 3 11 Time series plot of total nitrogen 1 ) for stations 7, 8, 9, and 10 ................................ ................................ ................................ ................ 96 3 12 Time series plot of total phosphorus concentrations ( g L 1 ) for stations 1 and 2 ................................ ................................ ................................ ......................... 97 3 13 1 ) for stations 3, 5, and 6 ................................ ................................ ................................ .................. 97 3 14 Time series plot of total phosphorus 1 ) for station 4 .......... 98 3 15 1 ) for stations 7, 8, 9, and 10 ................................ ................................ ................................ ............ 98 3 16 Time series plot of color (PCU) for stations 1 and 2 ................................ ........... 99 3 17 Time series plot of color (PCU) for stations 3, 5, and 6 ................................ ...... 99 3 18 Time series plot of color (PCU) for station 4. ................................ .................... 100 3 19 Time series plot of color (PCU) for stations 7, 8, 9, and 10 .............................. 100 3 20 Time series plot of water temperature (C) for stati ons 1 and 2 ....................... 101 3 21 Time series plot of water temperature (C) for stations 3, 5, and 6. .................. 101 3 22 Time series plot of water temperatu re (C) for station 4 ................................ ... 102 3 23 Time series plot of water temperature (C) for stations 7, 8, 9, and 10 ............. 102 3 24 Time series plot of sa linity (ppt) for stations 1 and 2 ................................ ......... 103 3 25 Time series plot of salinity (ppt) for stations 3, 5, and 6 ................................ .... 103 3 26 Time series plot of sa linity (ppt) for station 4 ................................ .................... 104 3 27 Time series plot of salinity (ppt) for stations 7, 8, 9, and 10 .............................. 104 3 28 Time series plot of mean daily discharge (m 3 s 1 ) at the Gopher River gauge station ................................ ................................ ................................ ............... 105 3 29 Time series plots of monthly observed and fitted values of log 10 transformed chlorophyll a concentrations from models for stati on 1 and station 2 ............... 106 3 30 Time series plots of monthly observed and fitted values of log 10 transformed chlorophyll a concentrations from models for station 4 ................................ ..... 107

PAGE 13

13 3 31 Time series plots of monthly observed and fitted values of log 10 transformed chlorophyll a c oncentrations from models for station 3, station 5, and station 6 ................................ ................................ ................................ ....................... 108 3 32 Time series plots of monthly observed and fitted values of log 10 transformed chlorophyll a co ncentrations from models for station 7, station 8, station 9, and station 10 ................................ ................................ ................................ ... 110 3 33 Time ser ies plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 1 and for holdout samples from 1998, 2009, and 2010 ................................ ................................ ................................ 112 3 34 Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 2 and for holdout samples from 1998, 2009, and 2010 ................................ ................................ ................................ 113 3 35 Time series plots of hindcasts and forecasts of log 10 transf ormed chlorophyll a concentrations for station 3 and for holdout samples from 1998, 2009, and 2010 ................................ ................................ ................................ 114 3 36 Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a conce ntrations for station 5 and for holdout samples from 1998, 2009, and 2010 ................................ ................................ ................................ 115 3 37 Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 6 a nd for holdout samples from 1998, 2009, and 2010 ................................ ................................ ................................ 116 3 38 Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 4 and for holdout samples fro m 1998, 2009, and 2010 ................................ ................................ ................................ 117 3 39 Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 7 and for holdout samples from 1998, 2009, and 2010 ................................ ................................ ................................ 118 3 40 Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 8 and for holdout samples from 1998, 2009, and 2010 ................................ ................................ ................................ 119 3 41 Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 9 and for holdout samples from 1998, 2009, and 2010 ................................ ................................ ................................ 120 3 42 Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 10 and for holdout samples from 1998, 2009, and 2010 ................................ ................................ ....................... 121

PAGE 14

14 LIST OF ABBREVIATION S AIC Akaike Informatio n Criterion C degrees Celsius CUSUM Cumulative sum CWA Clean Water Act h Hours km Kilometers L Liters m Meters m 3 Cubic meters mg Milligrams g Mi cro grams NCDC National Climatic Data Center OLS Ordinary least squares PCU Platinum cobalt units ppt Parts per thousand RMSE Root mean squared error s Seconds SED Squared Euclidean distance SRWMD Suwannee River Water Management District TN Total nitro gen TMDL Total Maximum Daily Load UFA Upper Floridan a quifer USGS United States Geological Survey y Years

PAGE 15

15 Abstract o f Thesis P resented to the Graduate School of the University o f Florida in Partial Fulfillment of the Requirements for t he D egree o f Master of Science SPATIAL AND TEMPORAL DYNAMICS OF CHLOROPHYLL AND NUTRIENTS IN THE SUWANNEE RIVER AND ESTUARY, FLORIDA, USA By Andrea Marie Krzystan December 2011 Chair: Thomas K. Frazer Major: Fisheries and Aquatic Sciences The ability to d ifferentiate between human induced changes and natural variations in water quality is a key element of effective management. An important step in distinguishing these sources of variation involves characterizing the temporal dynamics of specific water qual ity parameters which can be accomplished through analysis of long term datasets. Two long term datasets were used in this study to examine water quality dynamics within a system of concern: the Suwannee River and its estuary. T he f irst dataset was used t o assess the presence of increasing trends in total nitrogen concentrations along the length of the Suwannee River between 1989 and 2010 using cumulative sum (CUSUM) analyses CUSUM analyses were applied to data from diff erent sections of the river delinea ted based on a moving split window boundary detection analysis. Results demonstrated increasing total nitrogen concentrations in each of the delineated sections of the r iver across the st udy period. However, fine scale differences in CUSUM statistics highl ighted spatial differences in the influence of hydrogeology, land use, groundwater quality, and climatic events on nitrogen dynamics, which suggested the need for an adaptive management approach

PAGE 16

16 A second long term dataset was used to develop time series a nd regression models that described relationships between chlorophyll a concentrations and salinity, temperature, light availability, and concentrations of total nitrogen and total phosphorus within the lower Suwannee River and its estuary from 1998 to 201 0. To increase the power to detect relationships, stations were group ed according to results of multivariate ordination, and models were developed using pooled data from 1999 2008. The predictive power of each model was evaluated using data from 1998, 2009 and 2010. As expected, final model parameters varied among groups of stations distributed in the river, oyster reef, and nearshore areas of the system; however, color and total phosphorus concentrations always explained significant amounts of variation i n chlorophyll a values. Chlorophyll a concentrations exhibited different relationships to model covariates at the station level, although observations at the group level ordinated together. Models for groups of stations fit overall trends at all sites with adjusted R squared values ranging from 0. 34 to 0.7 2 but the accuracy of predictions differed among sites within each group. Differences in spatial rela tionships were attributed to variation in the river plume, which can influen ce top down grazin g pressure or bottom up influences, such as reduced light availability, due to color. Resu lts of both analyses highlight the overall complexity of th e Suwannee system and emphasize the need to consider multiple sources and scales of variation when attempti ng to distinguish natural variation from human activities leading to impairment or improvement.

PAGE 17

17 CHAPTER 1 INTRODUCTION The ecological health and integrity and human use of aquatic ecosystems depends on water quality. This importance is highlighted by e xamp les of negative impacts from degraded water quality including drinking water contamination (Alam et al. 2007), loss of revenue from fisheries closures (Mallin et al. 2002; Macfarlane 2006) and altered ecosystem structure and function (Duarte 1995; Cloern 2 001; Wazniak et al. 2007). To address such impacts, managers in the United States initially have focused on controlling threats to surface water quality at their source. At a national level, the Clean Water Act (C WA) was implemented in the 1970 s to regulat e the point source discharge of industrial pollutants (e.g., outfall pipelines). The scope of the act was broaden ed in subsequent year s to consider potential physical and biological impacts as well as threats from non point sources e.g., urban and agricu ltural runoff, which are not confined to a single point of introduction (Norgart 2004). In addition to outlining federal management guidelines, the CWA also provides minimum requirements for water quality management by state governments. More specifically, each state must define designated uses for all water bodies; identify waters not being adequately protected; and apply Total Maximum Daily Loads (TMDLs) to reverse impairment A water body that cannot serve its designated use because of poor water quality (e.g., criteria for a given parameter are not satisfied) is listed as impaired. Once impairment is verified through sampling TMDLs are developed to specify the maximum level of a pollutant that can be assimilated without causing harm and policies are im plemented to reduce loads to this level with the aim of restor ing the ability of the impaired waterway to meet its designated use (Norgart 2004; USEPA 2006).

PAGE 18

18 An example of a system currently being managed according to CWA guidelines for threats to water q uality is the Suwannee River in Florida. Several segments of the river of a failure to meet state water quality standards for nutrients or other parameters (FDEP 2002). Nutrient impairment is of particular concern because of its potential to lead to eutrophication. Eutrophication, an increase in the supply of organic matter to a system, can result in ecosystem wide impacts on biogeochemical and trophic processes by causing habitat loss, dec reas ing biodiversity, and degrad ing water quality especially through form ation of anoxi c or hypoxi c zones (Rosenberg 1985; Nixon 1995; Paerl 1999; Cloern 2001). Thus far, managers have address ed impairment in the Suwannee system with TMDL s and non regulat ory measures including Best Management Practices for agriculture (Norgart 2 004; USEPA 2006). A more recent regulatory initiative is the development of numeric nutrient criteria, which will establish upper limits for nutrient concentrations expe cted to pr event adverse impacts to water quality and wildlife (FDEP 2010). To establish effective numeric criteria, managers will need to understand nutrient sources and dynamics within each system as well as how changing nutrient levels manifest as biological resp onses. More specifically, managers will need to be able to distinguish natural variation in nutrient levels from variation due to anthropogenic inputs in order to designate attainable magnitude, duration, and frequency components for criteria. Although dis tinguishing these sources of variation is complex, a first step

PAGE 19

19 involves characterizing trends in nutrients and biological responses from long term water quality datasets.

PAGE 20

20 CHAPTER 2 NITROGEN TRENDS IN T HE SUWANNEE RIVER, F LORIDA Rationale Nitrogen is a nu trient vital to all life, and its biological availability is determined by the nitrogen cycle. The nitrogen cycle has been altered by human activities, such as fertilizer production, fuel combustion, and crop and animal agriculture (Galloway et al. 1995; V itousek et al. 1997; Carmargo and Alonso 2006). These activities have accelerated since the industrial revolution, resulting in increased levels of nitrogen in many terrestrial and aquatic ecosystems across the globe (Howarth 2008). Increasing nitrogen loa ds to aquatic systems have the potential to negatively impact human health and ecosystems. One health concern is methemoglobinemia a condition induced by high nitrate levels in drinking water that affects the ability of blood cells to transport oxygen (Ca rmargo and Alonso 2006). A primary concern for ecosystems is eutrophication, an increase in the supply of organic matter to a system Eutrophication can have ecosystem wide impacts and lead to decreased biodiversity by altering biogeochemical cycles, creat ing hypoxia or anoxia, and decreasing light penetration, which leads to loss of vegetated habitats (Rosenberg 1985; Nixon 1995; Paerl 1999; Cloern 2001) Although eutrophication can occur naturally, many incidences have been linked to human activities, inc luding detrimental eutrophication in the Wadden Sea, the Black Sea, and the Gulf of Mexico (De Jonge et al. 1996 ; Humborg et al. 1997 ; Turner et al. 2007). These systems share a common characteristic, the influence of large rivers, which is not unexpected because large rivers deliver substantial nutrient loads to many coastal ecosystems (Billen and Garnier 1997). For example, the northern Gulf of Mexico

PAGE 21

21 received 91% of its nitrogen load and 88% of its phosphorus load from the Mississippi and Atchafalaya Riv ers during 1972 1993 (Turner et al. 2007). Nutrients transported to nutrients in coastal systems with strong riverine influences need to account for sources and dynamic s of nutrients not only at the coast, but also along the river. An example of a system currently being managed as a watershed is the Suwannee River in Florida. Several segments of the Suwannee River have been declared as impaired because they did not mee for nutrients (FDEP 2002). To address this impairment, managers have implemented a multi faceted approach involving regulatory and non regulatory measures. Regulatory measures include Total Maximum Daily Loads, which e stimate the maximum level of a pollutant that can be assimilated by the water body without causing harm, and non regulatory measures include Best Management Practices for agriculture (Norgart 2004; USEPA 2006). A recent regulatory initiative invol ves the development of numeric nutrient criteria, which will provide upper limits to nutrient levels within a system to prevent adverse impacts to water quality and wildlife (FDEP 2010). To establish effective numeric criteria, managers will need to unders tand nutrient sources and dynamics within each system. More specifically, managers need to be able to distinguish natural variation in nutrient levels from variation due to anthropogenic inputs in order to designate attainable magnitude, duration, and freq uency components for each criterion. With this need in mind, the present study characterizes spatial and temporal variation in total nitrogen concentrations in the Suwannee River using a combination of

PAGE 22

22 trend and spatial analysis techniques. Cumulative sum (CUSUM) analyses will indicate if and when total nitrogen concentrations at eleven stations have increased from 1989 to 2010. Trend analyses will focus on different sections of the river as delineated with a moving split window boundary detection techniqu e. These results will provide information about differences in trends along the length of the river that will help managers identify areas in need of increased attention or customized management approaches. Methods Study Area The focus area for this study was the main reach of the Suwannee River within Florid a This portion of the river is 394 km long, has an average annual discharge of 300 m 3 s 1 and drains a n 11 0 00 km 2 watershed ( Katz et al. 1997; Wolfe and Wolfe 1985). The watershed as a whole is rela tively pristine with numerous forested, wetland, and protected areas; however, anthropogenic land uses such as crop and animal agriculture and phosphate mining are present (Katz et al. 1997; Bledsoe and Phlips 2000) As the river meanders towards the Gulf of Mexico, variations in geology, topography, and hydrology result in chang ing water chemistry and ecology (Wolfe and Wolfe 1985) Managers use these differences to divide the river into three reaches: the Upper, Middle, and Lower Suwannee. The Upper Suwan nee provides important spawning habitat for the Gulf Sturgeon and has some karst features like sinkholes, limestone outcrops, and springs such as White Springs, Suwannee Springs, and Ellaville Springs. Water in the Upper Suwannee River is highly colored, acidic, and typically has low nutrient concentrations. As the Upper Suwannee transitions into the Middle Suwannee near Ellaville at Highway 90, increased groundwater input from the

PAGE 23

23 Upper Floridan Aquifer via 62 mapped springs results in clearer, more alkal ine water with increased nutrient concentrations. The Middle Suwannee channel is 79 150 m wide bordered by swamp and bottomlan d hardwood plant communities, and inhabited by West Indian manatees during winters The r iver widens to 240 300 m in its Lower re ach which runs from Fanning Springs to the Gulf of Mexico where it splits into two passes. Relative to the Middle reach, there are fewer springs in the Lower Suwannee but major springs like Manatee and Fanning Springs do contribute to high nutrient level s. The Lower reach of the river is tidally influenced up to the Gopher River confluence and its bottom is characterized by exposed limestone, coarse sand, and sandy mud. Data Description Raw data consisted of monthly concentrations (mg L 1 ) of total Kjeld ahl nitrogen, ammonia, and nitrate nitrite for 1989 2010. Data were acquired from the State of Florida STORET database ( http://www.dep.state.fl.us/water/sto ret ) for 11 stations ( Table 2 1; Figure 2 1). For each station and sampling event, total nitrogen (T N) concentrations were calculated as the sum of concentrations of total Kjeldahl nitrogen, ammonia, and nitrate nitrite. Monthly TN concentrations were aggregated into annual arithmetic means to reduce serial autocorrelation and to minimize the influence of short term variability on results of test s for trend s Annual means were calculated using all values for a given year although some years had missing values or multiple sampling events within a given month.

PAGE 24

24 T rend Analyses Cumulative sum (CUSUM) control charts ( Page 1954 ) and a local boundary detection technique were used to identify spatial and temporal patterns in annual mean TN concentrations in the Suwannee River from 1989 to 2010. Spatial differences in temporal trends were evaluated by delineating sections of the river using a moving split window method and then comparing CUSUM results among sections. Sections were delineated based on dissimilarities in annual mean TN concentrations at adjacent sites. D ue to the relatively small number of sites d if ferences were determined using a moving window of 2 sites, the smallest size possible, and the differences were quantified using squared Euclidean distances (SEDs ). Calculations of SEDs using Equation 2 1 were based on mean TN concentrations, x 1 and x 2 fo r each year, t at adjacent sample sites, S 1 and S 2 across T = 22 y ( Fortin and Dale 2005) 2 1 Ten SED measures were calculated and plotted against pairs of adjacent sites to determine where boundaries occurred along the river. Boundaries were visualized as peaks relative to other SED values. High, narrow peaks were interpreted as sharp boundaries, whereas low, wide peaks were interpreted as gradual boundaries (Fortin and Dale 2005). Stations located bet ween boundaries of either type were grouped into the same section for CUSUM analyses. CUSUM analyses were used to determine if annual mean TN concentrations at each station in the previously identified sections increased relative to background levels durin g the 22 y study period Background levels were defined as the arithmetic mean of annual mean values within a section for 1989 2000. These years were selected

PAGE 25

25 because time series plots of annual mean TN concentrations appeared relatively stable during that period. For each station, annual mean values were converted into standardized deviations by s ubtracting the appropriate reference mean from the annual mean value and then dividing this quantity by the standard deviation of annual mean values within the se ction. S tandardized values were used to derive CUSUM statistics ( CS t ) for each year by taking the maximum of two values: 0 or ( CS t 1 + z t k ), where CS t 1 wa s the CUSUM statistic for the previous time period ( CS 1 = 0) z t wa s the standardized annual mean nutrient concentration for the current year and k was a term allowing for some variation around the background value, which was equal to one half the section standard deviation (i.e. equal to 0.5 for the standardized data) To determine if observed change s were significant, a threshold parameter, h was estimated using Equation 2 2 based on k = 0.5 and the ARL (i.e., the average run length between false alarms ) defined as 200 ( Rogerson 2006) 2 2 The ARL value was chosen to obtain a probability of 0.005 for the occurrence of false alarm s (i.e., where the threshold wa s exceeded due to random chance, not a true change in the process). Finally, CUSUM statistics were visualized as a function of each year for each station along with the threshold value, h to determine i f and when annual mean TN concentrations had changed relative to background concentrations for each section

PAGE 26

26 Results Boundary Detection The moving split window bou ndary detection method indicated three sharp boundaries along the study area based on differences in observed TN concentrations at adjacent sites (Figure 2 2). Based on these boundaries, four sections were delineated for the CUSUM analyses: station 1; stat ions 2, 3, and 4; stations 5, 6, and 7; and stations 8, 9, 10, and 11. Trend Analyses Time series plots of annual mean TN concentrations had generally similar patterns at all stations with maximum values occurring during 2003, 2004, or 2005 (Figures 2 3 to 2 6); however, CUSUM results did exhibit some spatial variability across sections. For the most upstream section, station 1, the CUSUM statistic was non zero, but below the threshold during 1994 1999 and 2003, and it exceeded the threshold the first time in 2004 and then steadily increased through 2010 (Figure 2 3). In the next section (stations 2, 3, and 4), the first non zero CUSUM statistics at all sites occurred during 1991 1992, and the threshold was first exceeded one year earlier than at station 1, in 2003 (Figures 2 4A to 2 4C). Furthermore, CUSUM statistics for all three stations increased steadily from 2003 to 2010, although the rate of increase slowed between 2005 and 2006 (Figures 2 4A to 2 4C). Like stations 2, 3, and 4, CUSUM statistics for st ations 5 and 7 in the next section were first non zero during 1991 1992 (Figures 2 5A and 2 5C). However, non zero CUSUM statistics were not observed at station 6 until 2001 2002 (Figure 2 5B). The first significant increase in annual mean TN concentration s was observed at stations 5, 6,

PAGE 27

27 and 7 in 2003, and CUSUM statistics increased steadily through 2010, with only a small change in the rate of increase during 2008 (Figures 2 5A to 2 5C). Similar to stations 2, 4, 5, and 7, the first non zero CUSUM statisti cs were observed in 1991 for stations 8, 10, and 11 in the most downstream section (Figures 2 6A, 2 6C, and 2 6D). However, the first non zero statistic was not observed until 1999 at station 9 (Figure 2 6C). The threshold was exceeded from 2003 to 2010 for stations 9, 10, and 11 but it was first exceeded in 1999 at station 8, the earliest significant increase at any of the stations. Patterns of significant CUSUM statistics at stations 9 and 11 were similar, with a steady increase through 2007, a slight leveling in 2008, and then an increase through 2009 and 2010 (Figures 2 6B and 2 6D). Significant CUSUM statistics at station 8 exhibited a more exaggerated S shape due to a relatively slow increase in values from 1999 to 2002 (Figure 2 6A). Finally, signi ficant CUSUM statistics at station 10 increased at a relatively constant rate from 2003 to 2010. Discussion Trends in Total Nitrogen Results of the CUSUM trend analyses show ed periods of increased TN concentrations in each section of the Suwannee River by the end of the 22 y study period. These results agree with previously documented trends in nitrogen in the river (e.g., increases of 0.02 1 y 1 between 1977 and 1997; Pitman et al. 1997), and they also parallel observed increases in concentrations of nitrate nitrogen discharged from springs (Katz et al. 1999) Nitrate nitrogen concentrations have increased in recent decades from 0.1 1 to 5.0 1 in some springs, with the trend being attributed to human activities, such as agriculture and wastew ater discharge (Katz et al. 1999). Nitrogen from springs and

PAGE 28

28 other groundwater sources can impact surface water quality because springsheds high connectivity between surf ace and ground waters (Katz et al. 1997; Crandall et al. 1999). Other sources of nitrogen loads to the Suwannee River include atmospheric deposition, runoff from swamps or wetlands, and point sources, such as wastewater outfalls (Upchurch et al. 2007; Br ow n et al. 2008). The relative impact of different nitrogen sources varies along the length of the river due to changes in hydrology, geology, and land use. The contribution of groundwater to the base flow of the Suwannee increases in the Middle and Lower Su wannee as the number of springs and other karst features increases (Crane 1986; Grubbs 1997). The Middle Suwannee also is dominated by agricultural land use, which contributes to increased nitrogen loading through leaching and runoff of animal waste and fe rtilizers (Pitman et al. 1997; Cabrera 2004). Because of increasing inputs from springs and increased nitrogen concentrations in groundwater toward the mouth of the Suwannee, higher concentrations of nitrogen and stronger trends in nitrogen were expected t o occur at stations in the downstream sections of the river. However, stations 2, 3, and 4 in the second most upstream section of the river had the highest mean TN concentration, 1.33 m g L 1 among all the sections across the whole study period. Section 2 may not have followed expected spatial TN trends due to the limited influence of groundwater in this region. As part of the Upper Suwannee, this section is coincident with a confined p ortion of Upper Floridan aquifer (UFA). Where the UFA is confined, interactions between groundwater and surface water are limited and water quality primarily is affected by local factors like land use (Crandall et al. 1999; Upchurch

PAGE 29

29 et al. 2007). Furthermo re, groundwater along most of section 2 does not flow towards the Suwannee River (Katz et al. 1997). These hydrogeologic characteristics combined with historic land uses (e.g., wastewater discharges, livestock farms, and paper mills) result in relatively h igh nutrient load s that are less dependent on groundwater than loads in other parts of the system (Hand et al. 1990 in Katz et al. 1997). Despite these unexpected results in spatial patterns of TN concentrations along the river, trends in annual mean TN co ncentrations were indeed strongest in the two downstream sections as indicated by the relative magnitude of their CUSUM statistics. CUSUM statistics for section 4 (stations 8, 9, 10, and 11) ranged from 26 to 41 relative to the 1989 2000 reference value an d exhibited the largest, positive cumulative deviations of all sections w ith non zero values (Figure 6). The trend was particularly strong at station 8 where a significant increase above the reference value was detected in 1999, the earliest positive dev iation for any station (Figure 2 6A). The trend at Station 8 may result from elevated nitrogen concentrations delivered by nearby second magnitude springs (Poe and Rock Bluff) and the Santa Fe River, which joins the Suwannee River just upstre am of this station In addition to the spatial variation in the influences of hydrology, geology, and land use, temporal variation in climatic conditions and agricultural practices can influence groundwater contributions and nitrogen loads to surface water bodies. For example, Katz and Bohlke (2000) noted that seasonal trends in groundwater nitrate concentrations in the Middle Suwannee sub basin followed patterns in fertilizer application. On the other hand, climatic events drive more complex interactions b etween ground and surface waters and the processes that influence nitrogen dynamics.

PAGE 30

30 During high rainfall events, increased water levels in the Suwannee can cause reverse flows for some springs (Giese and Franklin 1996 a ). Reduction in spring inputs, along with the potential for dilution and denitrification of nitrate in groundwater, results in less nitrate nitrogen loading to the Suwannee during periods of high flow (Katz et al. 1997). Where groundwater influences are less significant, increases in TN conc entrations are likely because increased rainfall generates runoff containing organic nitrogen from swamps, wetlands areas or agricultural operat ions (Copeland 2009). In contrast, periods of drought are characterized by higher relative contributions from sp rings ( G iese and Franklin 1996b). As the proportion of groundwater withi n the river increases, increased nitrate concentrations are possible depending, in part, on the age of the groundwater (Upchurch et a l. 2007; Copeland e t al. 2009). Discharged groundwa ter tends to be older during low flow periods, which can introduce groundwater from deep sources like the Avon Park formation that has high concentrations of organic ni trogen (Copeland et al. 2009). With the potential for both increased groundwater contrib ution s and higher nitrogen concentrations during periods of low rainfall, associated increases in riverine nitrogen levels often are observed (Katz et al. 1997). Such complex interactions between groundwater, surface water, and climatic conditions make it difficult to interpret patterns in TN concentrations at timescales from hours to months or seasons within years. In fact, Copeland et al. (2009) both documented variable responses for different components of TN during a single time period and observed that variation in flow can mask trends in nitrogen concentrations. With these difficulties in mind, the present study demonstrated the utility of evaluating trends on a longer timescale.

PAGE 31

31 The use of annual mean concentrations highlighted patterns that can be i nterpreted more easily. For example, annual mean TN concentrations at all stations (except 1 and 8) showed significant cumulative deviations from their respective 1989 2000 means that began in 2003 and continued through 2010. The timing of this increa se coincided with above normal rainfall in 2003 (Verdi et al. 2006 in Copeland et al. 2009). Above normal rainfall also was experienced during the 1997 98 El Nio event; however, significant increases in annual mean TN concentrations were not apparent in t he CUS UM analyses during these years. The discrepancy between trends following these time periods may be attributable to climatic conditions in the years prior to each period of high rainfall. More specifically, an extended drought occurred from 1999 to 20 03 while the 1997 98 El Nio event was preceded by a weak La Nina event in 1994 1995 (characterized by low rainfall) and then above normal rainfall conditions in 1996 (Crandall et al. 1999; Copeland et al. 2009). During the drought from 1999 to 2003, more nitrogen could have been stored within the soils, so when heavy rainfall occurred during 2003, a greater quantity of nitrogen was released causing a large r increase in nitrogen throughout the system. Similar patterns of higher than expected nitrogen concen trations were observed in British rivers after a drought in 1975 76 (Burt et al. 2009). Although the increase after 2003 is apparent in plots of annual mean TN concentrations for stations in the Suwannee River, short term events like tropical storms may ha ve masked trends calculated from monthly or more frequent data (Burt et a l. 2009; Copeland et al. 2009). CUSUM A nalyses Trends in annual mean TN concentrations were identified readily using CUSUM methods. CUSUM analyses have traditionally been used to moni tor industrial

PAGE 32

32 processes, but more recently, they have been applied to environmental data to detect changes in indicators of fishery stocks and to examine trends in water quality data (Gibbons 1999; Mesnil and Petigas 2009). CUSUM analyses are especially u seful for environmental analyses because they are robust to missing values, relatively simple to interpret, sufficiently flexible to meet the desired objectives (e.g., changing the sensi tivity to false alarms), and designed to detect changes from backgroun d noise in datasets where the type of change (e.g., linear or exponential) is unknown (Manly and MacKenzie 2003; Petigas 2009; Tam 2009). Another advantage of CUSUM analyses is the potential to examine relationships between variables by comparing them over the same time period. Standardized CUSUM statistics in particular can be used to compare trends in variables that vary widely in magnitude thereby providing insight into complex system dynamics (see Petigas 2009 and Briceo and Boyer 2010). This type o f analysis would be a logical follow up to the current study and could be used to explore how nitrogen dynamics relate to dynamics of potential nitrogen sources (e.g., fertilizer application) or potential biological responses (e.g., algal blooms). Although CUSUM analyses can be effective in many applications, the ability to obtain reliable results is dependent on meeting assumptions of the test, and the results are sensitive to the user defined reference value. For example, reference values usually are deri ved from a set of historical data for the process of interest. When the reference dataset encompasses a short period of time, variability may be underestimated and results may contain more violations than actually may have occurred. When too long a

PAGE 33

33 period of record is used, real trends may be contained within the reference dataset, and important changes may go undetected (Tam 2009). One way to account for this limitation is to conduct a CUSUM analysis in two stages: i ) select a reference period based on exp ert knowledge and carry out the CUSUM analyses to determine if there are any extreme values during the selected time p eriod then ii ) remove any extreme values and re calculate the reference value for subsequent analyses (Mertens et al. 2008). Another poten tial complication with CUSUM analyses of environmental time series data is the presence of serial or spatial dependence since independence of sample points is a key assumption. Where dependence is an issue, the spatial or temporal resolution of sample poin ts can be adjusted or analyses can be modified to account for dependence (see Manl y and MacKenzie 2000 and 2003). Summary This study demonstrated increasing TN concentrations in the Suwannee River between 1989 and 2010. During the same time period, elevate d nitrogen levels also have been documented in many aquatic systems around the world, and these increases have been linked to human activities (Galloway et al. 1995; Vitousek et al. 1997; Carmargo and Alonso 2006; Howarth 2008). Where nitrogen concentratio ns are high, adverse impacts to ecosystems and human health can occur. To mitigate impacts from nitrogen pollution, managers need to have an understanding of nitrogen sources, factors influencing nitrogen dynamics, biological responses to increased nitroge n levels, and factors that moder ate these biological responses. The influences of hydrogeology, land use, groundwater quality, and climatic events on nitrogen dynamics in the Suwannee River were highlighted through the application of CUSUM trend analyses w ithin a spatial context. This information will help managers

PAGE 34

34 develop comprehensive strategies to address concerns about nitrogen loads. First, the importance of hydrogeology was emphasized as observed trends differed from trends predicted solely on the bas is of differences in geography along the river. This result suggests that management approaches will need to be adapted for different parts of the Suwannee Basin depending, in part, on whether or not the UFA is confined or unconfined. Where the river ov erl ies confined parts of the UFA managers should target run off and promote land uses that do not deliver nitrogen. In areas where groundwater is more influential, nitrogen dynamics are complex and still not well understood; therefore, managers will benefit from additional research on groundwater quality and dynamics. Finally, the potential influence of long term climatic events remains an important consideration because managers will not be able to control these cycles and their impact may mask trends in nit rogen concentrations in the river. The relationships among nitrogen concentrations, biological responses, and mitigating factors were not addressed by this study. However, future efforts can use experimental approaches multiple CUSUM analyses or empir ical models to investigate how nitrogen trends manifest as trends in biological indicators, such a s chlorophyll a, and how the negative impacts of these responses can be mitigated by other factors, such as water residence time. In fact, data from the lower Suwannee River and the associated estuarine waters support an empirical approach.

PAGE 35

35 Table 2 1. Sampling site details River miles are measure from an origin point in Alligator Pass (Tillis 2000) and latitude and longitude are in decimal degrees (WGS 84). S ite Location River mile Latitude Longitude 1 N ear Benton, FL 191 .0 30.508115 82.716569 2 Below White Springs 171.2 30.325833 82.738611 3 At Suwannee Springs 150 .0 30.395000 82.936111 4 At Ellaville below US 90 127.3 30.376944 83.180278 5 A t Dowlin g Park 113 .0 30.244722 83.249722 6 A t Luraville 89.3 30.098889 83.171944 7 A t Branford 76.2 29.955556 82.927778 8 At Rock Bluff 57 .0 29.791111 82.924444 9 Near Wilcox 34.5 29.591389 82.937222 10 A 17 .0 29.399167 83.022778 1 1 A t Gopher River 5.7 29.328056 83.103056 Figure 2 1. Sampling sites and the extent of the Suwannee River Basin (green).

PAGE 36

36 Figure 2 2 Moving split window boundary detection results displaying differences between adjacent stations based on observed mean annu al nitrogen concentrations Figure 2 3. CUSUM statistics and mean annual total nitrogen concentrations (red line) for station 1 from 1989 to 2010. Horizontal dashed line represents CUSUM threshold value of 3.49.

PAGE 37

37 Figure 2 4. CUSUM statistics and m ean annual total nitrogen concentrations (red line) for A) station 2, B) station 3, and C) station 4 from 1989 to 2010. Horizontal dashed line represents CUSUM threshold value of 3.49.

PAGE 38

38 Figure 2 5. CUSUM statistics and mean annual total nitrogen concentr ations (red line) for A) station 5, B) station 6, and C) station 7 from 1989 to 2010. Horizontal dashed line represents CUSUM threshold value of 3.49.

PAGE 39

39 Figure 2 6. CUSUM statistics and mean annual total nitrogen concentrations (red line) for A) station 8, B) station 9, C) station 10, and D) station 11 from 1989 to 2010. Horizontal dashed line represents CUSUM threshold value of 3.49.

PAGE 40

40 CHAPTER 3 CHLOROPHYLL DYNAMICS IN THE LOWER SUWANNE E RIVER AND ESTUARY Rationale Coastal and estuarine systems are highl y productive, contain diverse habitats, and support commercial and recreational fisheries (Raabe et al. 2007). However, many estuaries have been negatively impacted by human activities in adjacent watersheds ( NRC 2000). Nutrient enrichment from land based sources is of particular concern due to the potential for increased delivery or production of organic matter, i.e., eutrophication (Nixon et al. 1996; Howarth 2008). Eutrophication can lead to increased occurrences of high chlorophyll a concentrations, tox ic algal blooms, hypoxia, or anoxia, which can decrease biodiversity and alter food web dynamics ( Rosenberg 1985; Nixon 1995; Paerl 1999; Cloern 2001 ). The extent of impacts from eutrophication varies among systems based on differences in hydrology, basin morphology, and biological factors ( Bricker et al. 2003; Painting et al. 2007 ). Thus, effective nutrient management requires knowledge of how these regulating factors affect biological responses to increased nutrient delivery. River influenced estuarine sy stems are particularly difficult to manage and susceptible to nutrient enrichment because rivers deliver substantial nutrient loads and contribute to complex hydrodynamics ( Billen and Garnier 1997 ). For example, the northern Gulf of Mexico received 91% of its nitrogen load and 88% of its phosphorus load from the Mississippi and Atchafalaya Rivers during 197 3 1993 ( Turner et al. 2007). Because these river introduced nutrients originate from all parts of the watershed, efforts to manage nutrients in coastal s ystems with strong riverine influences need to account for both upstream influences and coastal impacts.

PAGE 41

41 An example of a system currently being managed as a watershed is the Suwannee River in Florida. Several segments of the Suwannee River and its estuary standards for nutrients (FDEP 2002). To address this impairment, managers have implemented a multi faceted approach involving regulatory and non regulatory measures. Regu latory measures include Total Maximum Daily Loads, which estimate the maximum level of a pollutant that can be assimilated by the water body without causing harm, and non regulatory measures include for example, Best Management Practices for agriculture ( Norgart 2004; USEPA 2006). A recent regulatory initiative involves the development of numeric nutrient criteria, which will provide upper limits to nutr ient levels within a system designed to prevent adverse impacts to water quality and wildlife (FDEP 2010 ). To establish effective numeric criteria, managers will need to understand nutrient sources and dynamics within each system and how these dynamics are manifested as biological responses More specifically, managers need to be able to distinguish natural variation in nutrients and their associated biological responses from variation due to anthropogenic inputs This information will help managers designate attainable magnitude, duration, and frequency components for each criterion. A first step towards s eparating these sources of variation is the analysis of long term datasets, which can provide the opportunity to place short term events in the context of long term patterns, allow for the assess ment of processes th at occur over a long timescale, and s uggest potential sources of variation by highlighting patterns (Burt et al. 2010). To this end, the present study utilizes a long term dataset for the lower

PAGE 42

42 Suwannee River and its estuary to support more effective management of nutrients and their associat ed biological impacts. More specifically, this study provides insight into the spatial and temporal dynamics of water quality in the system by characterizing relationships between chlorophyll a, nutrients, and other w ater quality parameters. Chlorophyll a dynamics were th e focus of this study because chlorophyll a is a common proxy for phytoplankton biomass and it can also be used as an indicator of eutrophication (Bricker et al. 2003) Other parameters that are potentially related to chlorophyll dynamics across the system also were considered including causal parameters such as nutrient concentrations and regulating environmental variables like discharge, water color and temperature, salinity, and wind First, l ong term patterns in each variable were eval uated using trend analyses. R elationships between chlorophyll a and other parameters were then characterized using time series and regression modeling Results provide insight into the dynamics of and relationships between chlorophyll a and other water q uality parameters in the Suwannee River and its estuary and results also highlight the importance of considering spatial and temporal variation when managing water quality in estuarine systems Methods Study Area T he Suwannee River a major feature of sou thern Georgia and north central Florida, originates in the Okefenokee Swamp and terminates in a coastal plain estuary comprising a network of tidal channels and salt marsh habitat along the Gulf Coast of Florida (Bales et al. 2006; Orlando et al. 1993 ). Th is river exhibits blackwater characteristics, is 394 km long, has an average annual discharge of approximately 300 m 3 s 1 and meanders through a watershed draining roughly 28,600 km 2 (Wolfe and

PAGE 43

43 Wolfe 1985). The watershed as a whole is relatively pristine with numerous forested, wetland, and protected areas; however, anthropogenic land uses such as phosphate mining and crop and animal agriculture are present (Bledsoe and P hlips 2000; Katz et al. 1997). The Suwannee River basin has a karstic topography chara cterized by featur es like sinkholes and springs. The high porosity of this landscape results in a dynamic exchange between groundwater features like the Upper Floridan Aquifer and surface water s. In addition, several tributaries such as the Alapaha River, the northern Withlacoochee River, and the Santa Fe River are connected to the Suwannee River (Katz et al. 1999, 2001). The drainage patterns of these hydrologically significant tributaries are used to divide the Suwannee River basin into management units k nown as sub basins. Within Florida, there are five suc h divisions: the Alapaha sub basin, the Withlacoochee sub basin, the Upper Suwannee sub basin, the Santa Fe sub basin, and the Lower Suwannee sub basin (Katz et al. 1997). The focus area for this study the lower Suwannee River and e stuary, extends north from Cedar Key, Florida to Horseshoe Beach, Florida and it is part of the Lower Suwannee sub basin (Figure 3 1). The river and estuary are ecologically diverse, sustain both commercial and sport fisher ies, and provide critical habitat for endangered species such as the Gulf of Mexico sturgeon and the Florida manatee (Raabe et al. 2007). These benefits can be affected by variations in water quality within the system, which are driven by regional and loca l climate patterns, freshwater withdrawal for drinking water or irrigation, groundwater input from springs, and wind events (Bales et al. 2006; Quinlan 2003; Raabe et al. 2007).

PAGE 44

44 Data Sources The primary dataset was a thirteen year record of water quality d a ta for the lower Suwannee River and estuary covering 1998 to 2010 (Frazer, unpublished data) Parameters extracted from this dataset were water te mperature, salinity, color, and concentrations of total nitrogen, total phosphorus, and chlorophyll a. Supple mental data acquired from various sources provided information about mean daily d ischarge, wind direction, color, climate indices, and extreme weather events (Table 3 1). Sampling for core water quality parameters was conducted monthly at ten fixed station s distributed among the river, oyster reef, and nearshore zones of the system (Figure 3 2 ). Measurements for water temperature (C) and salini ty (ppt) were taken in situ at 1 m using a Y ellow S prings, I nc. data sonde (Model: 600R) Two surface water samples were collected during each sampling event for analysis at the laboratories of the Fisheries and Aquatic Sciences Program at the University of Florida. First, whole water samples were collected to determine concentrations of total nitrogen ( 1 ) and tot al phosphorus ( 1 ). Then, a second water sample was collected and a known volume from this sample was filtered through a 47 mm glass fiber filter. The filter was placed over dessicant and frozen for later processing at the laboratory. Processing consis ted of pigment extraction in ethanol and measurement of chlorophyll a concentration ( 1 ) using spectrophotometry. An acidification step was included in the spectrophotomic analyses to determine concentrations of chlorophyll a corrected for phaeophytin ( Method 10200 H; American Public Health Association 1989 ; Sartory and Grobbelaar 1984 ). Beginning in 1999, a separate filtered water sample was collected for subsequent determination of color (PCU) in the laboratory using spectrophotometry.

PAGE 45

45 Complete descri ptions of field and laboratory protocols are available elsewh ere (Frazer et al. 1998, 2001). Mean daily discharge (m 3 s 1 ) data were acquired online ( http://waterdata.usgs.gov / nwis/rt ) f or a United States Geological Survey gauging station located near the Gopher River and they consisted of verified values averaged over a 24 hour period. Data from a gauging station near Wilcox, FL also were acquired for use in estimating value s missing from the Gopher River dataset Wind directions were extracted from the continuous winds records of the Keaton Beach meteorological station ( http://www.ndbc.noaa.gov ) Records consisted of ten minute averages of clockwise deviations from true north in degrees. Additional color data (PCU) were acquired for two fixed stations n ear the mouth of the Suwannee River ( ht tp://waterdata.usgs.gov/nwis/dv) These surface water quality monitoring stations were sampled irregularly over the period of interest using methods comparable to those used for the primary dataset, and the resulting data were used to estimate values missing from the core water quality dataset. Finally, data were acquired for extreme weather events and two climat ic indices, i.e., the Southern Oscillation Index and the Multivariate El Nio Southern Oscillation Index. Mo nthly time series records for the climate indices were used to evaluate potential effects of global or regional climatic trends in the regression analyses. Data extracted from the Storm Event Database of the National Climatic Data Center (NCDC) for use in regression analyses include d the dates of weather events (i.e ., hurricanes, tropical storms, storm surges or flooding) that were expected to influence water quality by altering discharge, runoff, or rainfall within the study area.

PAGE 46

46 Trend Analyses The presen ce of monotonic trends was evaluated at each station for six water quality variables: total nitrogen, total phosphorus, salinity, discharge, color, and chlorophyll a corrected for phaeophytin (hereafter referred to as chlorophyll a). Objectives of the anal yses were to i) determine if trends in causal parameters, such as nutrient concentrations, were accompanied by trends in a response variable, chlorophyll a, and ii) compare any significant trends between sample sites and across the study area. Monotonic tr ends, gradual increases or decreases during the study period, were examined rather than step trends, change s in the mean value of a parameter after a pre defined date, because there was no a priori knowledge of an event that would cause a step change. Resu lts of tests for normality, the presence or absence of periodicity, and serial independence were used to select specific methods to evaluate trends for each station and parameter. Except where otherwise noted, al l analyses were conducted with the R statist ical software package (version 2.12.2; 2011 0 3 25; R Development Core Team 2011). Normality was evaluated with the Shapiro Wilk test (shapiro.test; R Development Core Team 2011) to determine if parametric or non parametric methods were appropriate for each combination of station and parameter. Where the null hypothesis of a normal distribution was rejected at a 90% significance level, non parametric tests (i.e., Mann Kendall or Seasonal Kendall) were employed due to their increased power to detect trends (B ouchard and Haemmerli 1992; Carey 2009 ). Parametric methods consisted of simple linear regressions with time as an explanatory variable. Next, the presence of periodicity, cyclic variation among months within years due to variation in factors like precipit ation, water temperature, and light availability, was

PAGE 47

47 assessed with the Kruskal Wallis Rank Sum test (kruskal.test; R Development Core Team 2011). Because periodicity can obscure the presence of a trend or suggest a trend is significant when it actually is not, methods accounting for periodicity were applied when the null hypothesis of no difference in the distributions of values between months was rejected at a 90% confidence level (Helsel and Hirsch 1991; Qian et al. 2007). For normal data analyzed with p arametric regression, periodicity was removed prior to testing for trends by subtracting the monthly median value for that station and parameter from each data point (Qian et al. 2007). Where data were non norm al, periodicity was addressed using the non pa rametric Seasonal Kendall test implemented in the program Kendall.exe ( Helsel et al. 2006 ). If periodicity was not present and data were non normal, the Mann Kendall test was used (MannKendall; McLeod 2011 ). Finally, serial independence was tested at a one month lag and a 90% confidence level using the Ljung Box Q test (Box.test; R Development Core Team 2011) because serial autocorrelation increases Type I erro rs in trend analyses (Darken et al. 2002; Hirsch and Slack 1984; Ljung and Box 1978). Independence tests on normal data were implemented after significant periodicity was removed by subtracting the monthly median value for that station and parameter from each data point. Non independence was alleviated for normally distributed time series by modeling t he serial autocorrelation with an autoregressive time series model, extracting the residuals from this model, and carrying out the trend test on the independent residuals. For Seasonal Kendall tests on non normal data, significance levels were adjusted for non independence by incorporating a covariance term into the estimation of the variance of the test statistic, S (Darken et al. 2002; Hirsch and Slack 1984). For Mann Kendall tests, a block bootstrap

PAGE 48

48 method (tsboot; Canty and Ripley 2010 ) was used to obta in improved significance tests in the presence of significant autocorrelation (see McLeod 2011). Regression Analyses Overview A combination of multivariate regression and time series modeling techniques were used to assess relationships between chlorophyll a concentrations and salinity, water temperature, light availability, concentrations of total nitrogen and total phosphorus, as well as several larger scale factors (i.e., river discharge, wind weather events, and climate indices ). To increase the power to detect relationships and to account for potential spatial variability in relationships, stations were grouped according to results of a non metric multi dimensional scaling based method on conditions observed at each site over the ten year period. For e ach group of stations a regression model was developed using pooled, log 10 transformed data from 1999 to 2008, with coefficients and intercepts estimated using ordinary least squares (OLS). After satisfying diagnostic requi rements for normality and hom o scedasticity, residuals for individual stations were extracted and evaluated for serial autocorrelation. Where autocorrelation was significant, it was modeled using a combination of first order and seasonal autoregressive or moving average terms to create residuals that satisfied assumptions associated with the OLS technique. The predicted values from each time series model were added to the corresponding predicted values for the same station from the group model to obtain an adjusted prediction that was e valuated to assess how well the revised predictions fit the real data. Finally, the predictive power of each model was evaluated for each station by generating adjusted predictions for chlorophyll a concentrations based on data in

PAGE 49

49 holdout sa mples from 1998 2009, and 2010 and comparing these results to recorded chlorophyll a concentrations. Data management Missing values. Missing values represented 17% of the whole dataset and were replaced with one of three values: an average value derived from relevant d ata a value for a similar location and time from another data sou rce, or a predicted value from regression model s based on relevant data Eleven m issing water temperatures were replaced with means calculated from data for the same station in the same mont h of all available years Data from nearby Suwannee River Water Management District sampling stations (SUW275C1 and SUW305C1) provided replacements for twenty three missing values in the color dataset for stations 1 and 2. A series of r egression an alyses provided substitutes f or 241 missing values in the corrected chlorophyll a, total nitrogen, total phosphorus, salinity color, and river discharge sub datasets. These regressions were based on data that were pooled across all stations and log 1 0 transformed after adding 0.5 to eliminate zero values. Missing values for the corrected chlorophyll a series were not replaced if they coincided with missing values for the uncorrected chlorophyll a series Missing chlorophyll a values were not estimated for fourteen sampling events to avoid biasing regression results by inserting new chlorophyll a values based on other available data Station groups. Stations were grouped according to results of an ordination implemented in PRIMER v.6 ( Clarke and Gorley 2006 ). Non me tric multi dimensional scaling (MDS; Clarke 1993) was used to ordinate a resemblance matrix of Bray Curtis similarities between stations. Similar ities were computed using range transformed data for the following parameters: temperature, salinity, color, an d concentrations of total

PAGE 50

50 nitrogen and total phosphorus. Within a given combination of parameter and month, range standardization s involved subtracting the minimum value from across all stations and then dividing by the range of values across all stations. Thus, the transformed values were scaled from 0 to 1 while retaining the relative relationships among stations and reducing the influence of differences in the absolute magnitudes of the parameters. Descriptive analyses. D ata were summarized across the sa mpling peri od (1998 to 2010) and for each year usin g descriptive statistics that were visualized with time series plots Descriptive statistics calculated in Microsoft Excel consisted of the means, medians, standard deviations, and minimum and maximum valu es for each year and across the entire sampling period for each station and each parameter. Time series plots of raw data for groups of stations were generated for each parameter across the entire study period. Additional v ariable s. A number of lagged and interaction terms were generated in Microsoft Excel for consideration in the regression analyses. Terms with a one month lag were defined for each quantitative variable as the observed value for the month before the given sampling period. Additional lag te rms were defined for discharge by extracting the mean daily discharge for one day, two days, and two months prior from the full discharge dataset available online. Second order interaction terms were generated between all quantitative variables by multiply ing the raw values of the respective component variables. A single third order interaction term was generated by multiplying values for total nitrogen, total phosphorus, and color within each time period. Several qualitative variables were defined as 1 for present and 0 for absent to account for unique conditions that could affect a sampling event. Dummy variables

PAGE 51

51 consisted of terms for wind events, storm events, years with high discharge, and sampling dates with high discharge values. Wind events were defi ned as continuous episodes of predominantly southwest winds for periods of 12, 24, 36, or 48 h before the time of sampling as derived from average, hourly wind directions. Dummy variables for storm events were defined as a flood, severe thunderstorm, tropi cal storm, or hurricane preceding sampling by one month or less, with data extracted from the Storm Event Database for Dixie County, Florida. Years with high discharge were defined as those with an average annual discharge greater than the average discharg e for 13 years (1998 2010) at the Gopher River station. Periods of increased discharge also were incorporated into the model by way of dummy variables for individual days of sampling with discharge values within the top 50% or top 5% of discharges recorded during the period of study. Dummy variables with more than two levels were designated for the appropriate month and water season for each observation to incorporate these potential sources of variation in the regressions. Dummy variables for sampling mon th were defined with January as the base level (i.e., samples collected during January were not coded in the analyses and dummy variables for the remaining months represented effects beyond those observed in the month of January ). Dummy variables represent ing the water season for a given observation (i.e., early dry, late dry, early wet, or late wet) were defined using the early dry season as the base level. Water seasons were defined after the United States Geological Survey (USGS) convention where June, J uly, and August are the early wet season; September, October, and November are the late wet season;

PAGE 52

52 December, January, and February are the early dry season; and March, April, and May are the late dry season. Finally, all quantitative variables were log 10 transformed after adding a constant value of 0.5 to eliminate undefined values due to zeros. Ordinary least squares regressions Model selection Linear regression models were developed for each group of stations using log 10 transformed data from 1999 2008 pooled across stations with in each group and an interactive stepwise procedure based on pre defined variable selection criteria. Subsets of variables considered for each model were selected from the complete set of variables defined above based on individu al relationships with the response variable, log 10 transformed chlorophyll a ( log 10 chl a ) concentrations. These significance using an asymptotic t approximation (cor. test; R Development Core Team 2011). If the linear correlation with log 10 chl a concentrations was significant at an 85% confidence level, the variable was retained for consideration. Variables were prioritized for inclusion according to the absolute value of the being considered first. Variables were evaluated one at a time, with the exception of interaction terms, which were evaluated alongside their component terms, b y adding them to the previous candidate model and estimating the new model parameters (lm; R Development Core Team 2011). The initial reference point for future candidate models was an intercept only model with no additional covariates. Each variable was e valuated using a combination of four criteria: i) the significance of the coefficient estimates for each variable based on a t test and a 95% confidence

PAGE 53

53 level; ii) the significance of an F test comparing nested models at a 90% confidence level; iii) the re lative increase or decrease in the Akaike information criterion (AIC) value after the new variable(s) were added to the model; and 4) whether or not the variance inflation factor of the new variable was greater than 10. An additional criterion for retainin g interaction terms in a given model was that the component terms, as well as the interaction term, had significant coefficient estimates to ensure that the interaction term provided information above that provided by any of the component terms alone. Comb ining these criteria created six potential scenarios that determined if variables were retained in the model or removed from consideration at each step. Variables were kept in the following cases: i) when both the coefficient t tests and model F tests were significant; ii) when the coefficient t tests were significant, the model F test was not significant, and there was a decrease in the AIC value; or iii) when the coefficient t test was not significant, the variance inflation factor for the new variable wa s greater than 10, and there was a decrease in the AIC value. Variables were not retained in the candidate model when: i) the coefficient t test was significant but the model F test was not significant and there was no decrease in AIC value; ii) the coeffi cient t test was not significant and the variance inflation factor for the new variable was less than 10; or iii) the coefficient t test was not significant, the variance inflation factor was greater than 10, and there was no decrease in AIC value. Variabl es were re evaluated and retained or eliminated at each step in model development. Terms were added or removed from candidate models until all the prioritized variables in the screened subset were evaluated. Due to the stepwise nature of the model develo pment and the potential for the order of variable evaluation to affect

PAGE 54

54 variable significance linear correlations were tested between model residuals and the at the end of the variable selecti on procedure (cor.test; R Development Core Team 2011). This final check was used to identify any variables that were initially removed from the model, but that could still have a significant relationship with chlorophyll a. Variables showing significant li near correlation at this step were added to the model and their contributions were re evaluated by applying the four variable selection criteria. This process was repeated until no additional, significant variation in the model residuals coul d be explained. Finally, the resulting model for each group of stations was subjected to a last set of diagnostics assessing normality, homogeneous variances and independence of the residuals before it was chosen as the final model. Model diagnostics Resi duals for each model were assessed for normality and homogeneity of variance at a 95% confidence level using Shapiro Wilk (shapiro.test; R Development Core Team 2011) and Breusch Pagan tests (bptest; Zeileis and Hothorn 2002), respectively. If residuals we re non normal or had heterogeneous variance, datasets were examined for influential outliers. Influential observations were identified using plots of the studentized residuals as a function of hat values which measure distance from the center of the distr ibution of the explanatory variables, for each observation. Plots were annotated with lines for the threshold values of each statistic. The threshold for potential outliers was a studentized residual for a given observation greater than 3.0. Observations w ere considered potentially influential i.e., exerted significant leverage, when the hat value exceeded a threshold, h This threshold was calculated separately for each model using Equation 3 1 based on k the number of

PAGE 55

55 parameters e stimated in the model and N the number of observations used to estimate model parameters 3 1 Influential outliers were examined further to determine if they were erroneous values or if they were due to unique characteristics of the given sampling effort (e .g., a n extended period of heavy rainfall). Values were deleted when there was a known reason for an error (e.g., equipment failure). Otherwise, a new dummy variable was created for the particular situation (e.g., for the time periods associated with a rai nfall event). New dummy variables were added to the model, evaluated according to the four criteria outlined previously, and retained in the model where they added significant information. After adding a new dummy variable, the residuals were re evaluated for normality and homoscedasticity. Where diagnostic tests failed again, identification of outliers and influential values continued until residual diagnostic tests were satisfactory. Residuals for each station were tested individually using the Ljung Box test for overall independence (Box.test; R Development Core Team 2011). If residuals for a given station were independent, no further action was taken and the model selected during OLS model selection was considered the final model. For stations with resid uals that were not independent at a 9 0 % confidence level, a t ime series model was developed. Time series models Time series models were developed for autocorrelated residuals from individual stations to create a model that yielded residuals satisfying the OLS regression assumption of independence and to improve overall model fit. Serial autocorrelation

PAGE 56

56 was modeled using a combination of autoregressive and moving average terms that were either non seasonal (i.e., with a 1 month period) or seasonal ( i.e., wit h a 12 month period). Potential parameters were identified for each station based on a visual inspection of plots of the partial autocorrelation function (pacf; R Development Core Team 2011) for each set of residuals and the results of the extended sample autocorrelation function (eacf; Chan 2010). In general, where the partial autocorrelation function plot indicated no autocorrelation beyond a lag of p an autoregressive term of lag p was evaluated and where the partial autocorrelation function plot exhib ited dampening oscillations with peaks at multiples of lag q a moving average term for lag q was evaluated. The extended autocorrelation function provided a set of potential autoregressive and moving average terms based on iteratively fitting least square s estimates to the dataset and determining which terms were significant ( Tsay and Tiao 1984 ). From these potential parameters, a set of candidate models were defined that ranged from including only one parameter to including as many as six parameters. T o m inimize model complexity n one of the candidate models included an intercept term and both non seasonal and seasonal terms were limited to a maximum lag of 2 (i.e., only up to 2 or 24 months). Each candidate model was fit to the appropriate residual s using maximum likeliho od criteria (arima; Chan 2010). The fit of each candidate model was evaluated by examining the significance of coefficient s and whether the residuals were independent. Significance of estimated t te st at a 90% confidence level, and independence of the new residuals was evaluated using the Ljung Box test at a 9 5 % confidence level. Models meeting these two criteria were ranked in increasing order of

PAGE 57

57 their respective AIC values and the model with the lo west AIC value was selected as the final model. Where there were multiple models with AIC values within a range of 2, the least complex model was selected as the final time series model. Fitted values from the final models were used as a time series adjust ment for fitted values from the OLS regression models. Final model evaluation Final models were evaluated for each station by assessing fit for data from 1999 2008 and by examining the predictive ability of each model for holdout sam ples from 1998, 200 9, and 2010. Model fit was quantified with adjusted R squared and root mean squared error values calculated when the final model was applied to data from 1999 2008 for each station. Fit statistics were calculated for both the final model without time serie s adjustment s and the final model with time series adjustments wherever applicable. Time series adjustments were incorporated into the predictions by adding the predicted value from the time series model to the predicted value from the OLS regression mode l. Furthermore, root mean squared error values were calculated for yearly subsets of the study period to examine where each model performed well and where it did not. These smaller scale statistics were supplemented by visualizations consisting of time ser ies plots of the observed and adjusted pred icted values for each station. Finally, the predictive ability of each station model was evaluated by calculating root mean squared error values when the models were fit to data from 1998, 2009, and 2010. The accu racy of hindcasts and forecasts also were evaluated qualitatively by examining plots of the observed and adjusted predicted values from each model for each year and station.

PAGE 58

58 Results Trend Analyses Test selection Tests were selected to produce consistency a nd to address violation of assumptions. Non parametric methods were used for all trend analyses to facilitate comparisons among results even though Shapiro Wilk tests indicated a violation of the normality assumption in only 1 of the 50 cases tested. Perio dicity was detected in 31 cases Of these 5 were independent and evaluated with the standard Seasonal Mann Kendall test while 26 were not independent and evaluated with an adjusted form of the Seasonal Mann Kendall test. In the remaining cases with no periodicity, 2 were independent and evaluated with the standard Mann Kendall test while 17 were not independent and the significance of the Mann Kendall test was adjusted using block bootstrap methods. Trends Trends were detected in some parameters at some stations. Increasing trends were significant at a 90% confidence level for total nitrogen at stations 1 and 5 and for total phosphorus at station 9 (Tables 3 2 and 3 3 ). No trends were significant in the corrected chlorophyll a series at any stations at a 90% confidence level (Table 3 4). However, chlorop hyll a exhibited a decreasing trend at station 3 and an increasing trend at station 6 that were significant at an 85% confidence level (Table 3 4 ). Significant i ncreasing trends were detected for total nit rogen at station 7 and total phosphorus at station 10 at an 85% confidence level (Tables 3 2 and 3 3 ). No significant trends were observed in the color or salinity time series at any station at an 85% significance level or higher (Tables 3 5 and 3 6 ). Mean daily discharge values at the Gopher River gauging

PAGE 59

59 station did not show a significant trend over the course of the study period ( S =299; P =0.68 ). Regression Analyses Data management Missing values. Regression models were used to estimat e m issing values for color, salinity, total nitrogen, total phosphorus, and discharge and were based on a variety of explanatory variables (Tables 3 7 to 3 11). Station grouping Non metric multi dimensional scaling resu lted in four groups of stations with a two dimensional stress value of 0.01 which indicated a reliable result (Figure 3 3 ). The first group consisted of stations 1 and 2. The second group contained only station 4. The third group consisted of stations 3, 5, and 6 and the fourth group contai ned stations 7, 8, 9, and 10. Descriptive statistics. T ime series plots and descriptive statistics for chlorophyll a, total nitrogen, total phosphorus, color, water temperature, and salinity showed similarities and differences in patterns among stations and parameters. Complete sets of descriptive statistics for each parameter and station are available in the Appendix. Across the study period, chlorophyll a concentrations were least variable and lowest at stations 1 and 2 and most variable and highest at stations 3, 5, and 6 (Figures 3 4 and 3 5). In fact, variability in chlorophyll a at stations 3, 5, and 6 (Tables A 3, A 5, and A 6) was at least twice as high as that observed at stations 1 and 2 (Tables A 1 and A 3). Within years, the period between 2000 and 2002 was the most variable for all stations except station 1 and the lowest variability in chlorophyll concentrations was during 1998, 2006, or 2007 for all but stations 4 and 6 (Tables A 1

PAGE 60

60 to A 10). Despite the low variability during 2006, a lar ge, positive deviation in chlorophyll a was observed at all stations during this year (Figures 3 4 to 3 7). These deviations during 2006 coincided with the end of a period of relatively high variability in concentrations of total nitrogen, total phosphorus color, salinity, and discharge, which appeared to begin around 2003. Relatively higher variability in total nitrogen concentrations between 2003 and 2006 was most apparent at stations 3 through 10 (Figures 3 8 to 3 11); although at stations 1 and 2, the apparent decrease in concentrations between 2003 and 2006 seemed to track the pattern that occurred between 1998 and 2002 (Figure 3 7). A similar, but weaker signal during these two periods also is apparent at stations 3 through 6, but not at stations 7 th rough 10 (Figures 3 7 to 3 11). Most stations did not experience any large deviations, with the exception of station 7, which had a high value during 2009 that was more than three times the mean value across the study period (Table A 17). The period of hig h variability in the middle of the study period also was noticeable in time series plots of total phosphorus concentrations. This relative increase in variability was most conspicuous at stations 7 through 10 (Figure 3 15), but it also was evident at stati ons 4, 3, 5, and 6 (Figures 3 16 and 3 17). In addition to this period of variability, large deviations in total phosphorus were observed during 2008 at stations 1, 2, and 4 (Figures 3 12 and 3 14) and during 2004 at stations 1, 2, 7, 8, 9, and 10 (Figure s 3 12 and 3 15), which was similar to patterns in total nitrogen F inally mean concentrations of total phosphorus across the study period were lowest at stations 7, 8,

PAGE 61

61 9, and 10 and almost four times as high at stations 1 and 2 ( Table s A 27 to A 28, A 21 and A 22). In addition to a period of high variability during 2003 to 2006 that also was observed in nutrient concentrations color ex hibited an additional period of relatively high variability at most stations between 2008 and 2010 (Figures 3 16 to 3 19 ). Furthermore, color was elevated at all stations during 1998 (Figures 3 16 to 3 19). Maximum color values at all stations were observed during one of these three time periods with color values being consistently highest at stations 1 and 2 (Tables A 31 and A 32). S tation 1 also exhibited the largest range in color values, while values at station 8 exhibited the smallest range of all stations (Tables A 31 and A 38). In contrast to the other parameters, patterns in water temperature did not display noticea ble periods of increased variability across the study period. At all stations, water temperatures fluctuated in a cyclical pattern that followed expected seasonal trends (i.e., colder temperatures in the winter months; Figures 3 20 to 3 23). However, stati ons 3, 4, and 7 were subject to a wider range of temperatures than other stations across the study period and station 4 exhibited the lowest temperature (Tables A 43, A 44, and A 47). Finally, timing of patterns in salinity and discharge were similar to patterns observed in the color series with the presence of three periods of increased variability (Figures 3 24 to 3 28). However, this pattern was not present at stations 1 and 4 (Figures 3 24 and 3 26). Station 1 had minimal variability in salinity acro ss the study period (Table A 51) while salinity at station 4 was the most variable across the study period with no apparent patterns (Table A 54; Figure 3 26). Where patterns were

PAGE 62

62 visible, decreases in salinity tended to occur close to increases in dischar ge (Figures 3 24 to 3 28). In fact, the minimum salinity value of 0 ppt was observed at half of the stations (1 through 5) during 1998, the same year that the maximum mean daily discharge was observed (Figures 3 24 t o 3 26, 3 28). Model descriptions Stat ions 1 and 2 The group model developed using pooled data fr om stations 1 and 2 had nineteen terms and explained 72% of the variability in log 10 chl a from 1999 to 2008 ( F 19, 217 = 33.29, P = 0.00; Table 3 12). Two of the terms were customized dummy variab les for 99 th percentile chlorophyll and November 2007 The 99 th percentile chlorophyll dummy variable represented sampling events where the observed chlorophyll a concentration was in the top 1% of observed values across all stations and the entire study p eriod (i.e., values > 30.23 1 ) while the November 2007 term was used to model influential observations during that month Other variables included an interaction term between salinity and total phosphorus and two lagged terms for discharge and total p hosphorus. Each of these terms displayed a negative relationship with log 10 chl a (Table 3 13). Residuals extracted from the group model were significantly autocorrelated for both stations 1 and 2 (Table 3 14). A utocorrelation in station 1 residuals was mo deled with one non seasonal, first order a utoregressive term (Coefficient = 0.33, S.E = 0.09, P = 0.00) while the model for station 2 residuals consisted of a non seasonal, first order a utoregressive term (Coefficient = 0.19, S .E. = 0.09, P = 0.01) and a f irst order, seasonal m oving average term (Coefficient = 0.24, S.E. = 0.11, P = 0.01).

PAGE 63

63 Station 4 The final model for station 4 explained 34 % of variation in log 10 chl a and it had the fewest terms of all the group models, i.e., four ( F 4 114 = 1 6.12 P = 0.00 ; Table 3 15). Of the four terms, log 10 temperature displayed a large, positive relationship with log 10 chl a while log 10 color was the only variable with a negative relationship to the response variable (Table 3 16). Residuals from this model were significantly autocorrelated (Table 3 14) and a time series model was developed that had one first order, non seasonal moving average term (Coefficient = 0. 34, S.E. = 0. 09 P = 0.01). Stations 3, 5, and 6 The final model explained 34 % percent of the varia tion in log 10 chl a for pooled data from 1999 2008 ( F 6, 3 51 = 31.57 P = 0.00; Table 3 17 ). The model contained two dummy variables for observations coinciding with the following conditions: g reater than 50% discharge (i.e., values greater than 5973 m 3 s 1 ) or southwest 36 h wind events. Residuals extracted from the group model were independent for station 3 while residuals for station s 5 and 6 we re significantly autocorrelated (Table 3 14). Autocorrelation was modeled for station 5 and 6 residuals with fi rst order, non seasonal autoregressive terms (Coefficient = 0.20, S.E. = 0. 09 P = 0.00 and Coefficient=0.25, S.E. = 0. 09 P = 0.00, respectively). Stations 7, 8, 9, and 10 The final model for pooled data from stations 7, 8, 9, and 10 explained 5 5 % of var iation in the observed log 10 chl a series for 1999 2008 ( F 19 45 6 = 3 0.99 P = 0.00; Table 3 19 ). Among the nineteen variables in this model were five dummy varia bles, including dummy variables for month of the year and one term generated to account for in fluential outliers during October of 2004 Log 10 salinity was

PAGE 64

64 the most significant parameter overall and statistics indicated a negative relationship with log 10 chl a (Table 3 20). Residuals from the group model were significantly autocorrelated for stati ons 7, 8, and 9, but not for station 10 (Table 3 14). Therefore, time series models were developed for only stations 7, 8, and 9. Time series m odels for stations 7 and 9 had one first order, non seasonal a utoregressive term (Coefficient = 0.17, S.E. = 0 .09 P = 0.0 5 and Coefficient = 0. 24, S.E. = 0.09, P = 0.01, respectively). The time series model f or station 8 had one first order, non seasonal m oving average term (Coefficient = 0. 27, S.E. = 0. 09 P = 0.0 3 ) and one first order, seasonal moving average term (Coefficient = 0.22, S.E. = 0. 10 P = 0.0 1). Model fit Fitted values generally followed overall trends in observed chlorophyll at individual stations from 1999 to 2008 (Figures 3 29 to 3 32). However, most unadjusted fitted values underestimated large dev iations in log 10 chl a Overall, the accuracy of unadjusted fitted values as measured by root mean squared errors (RMSE) was highest for stations 1 and 2 and lowest for stations 3, 5 and 6, with results for station 3 being least accurate (Table 3 21). When accuracy was assessed within years, it was highest across all stations during 2000, 2001, 2002, and 2007 and lowest during 1999 (Table 3 21). Throughout the most accurate years, stations 7 and 8 were among the top three most accurate stations (Table 3 21) Station 7 fitted values also were the most accurate relative to other stations during 1999, whereas accuracy for station 8 during this year was among the lowest of all stations (Table 3 21). Relative accuracy varied among stations across

PAGE 65

65 other years, alt hough some general patterns existed. First, fitted values from stations 1 and 2 were consistently among the most accurate of all s tations for all years (Table 3 21). Second, fitted values from stations 9 and 10 showed the most variability in relative accuracy across the years. In fact, the highest observed RMSE came from station 10 during 1999, and the second highest accuracy came from this station during 2004 (Table 3 21). Finally, fitted values for stations 3, 4, 5, and 6 tended to be among the least accurate across all years (Table 3 21). Accuracy of fitted values was differentially affected after time series adjustments were applied. Across the time period used to develop the models, time series adjustments had no effect on individual model accuracy at most stations, with improvements being observed only for stations 4 and 5 (Table 3 21). The improved fit for stations 4 and 5 was apparent in fitted values that more closely followed short term variation in observed log 10 chl a (Figures 3 30 and 3 31B) Time series adjustments also increased accuracy in many instances when RMSE values were calculated within years; h owever, patterns of improvement at this shorter timescale varied among years and stations. Time series adjustments resulted in improvement for all of the eight relevant stations during 1999 (Table 3 21). Other years where accuracy was improved at most stations included 2003, 2005, and 2006 (Table 3 21). On the other hand, time series adjustments were least effective in 2008 where accuracy of fitted values was improved at only stations 2 and 4 (Table 3 21). When considered across years and within stations, time series adjustments were least effective in improving accuracy at stations 5, 7, and 8 and most effective at stations 1, 2, and 4 (Table 3 21). In fact, the largest absolute improvement in accuracy was observed

PAGE 66

66 at station 4 during 2005 where the RMSE decreased from 0.30 to 0.26 after applying time series adjustments. Model evaluation Hindcasts and forecasts were generally less accurate tha n models fitted to data from 1999 2008. Averaged across the relevant years and stations, accuracy was poorer for time series adjusted predictions derived for the holdout samples (i.e., RMSE for 1998, 2009, and 2010 = 0.28 and RMSE for 1999 2008 = 0.21). Si milarly, unadjusted RMSE values for hindcasts and forecasts were larger than 73% of similar values for data from 1999 2008 (Table 3 21). Accuracy of estimates was improved for 2010 and essentially maintained for 1998 and 2009 when time series adjustments w ere applied. Time series adjustments also improved accuracy in at least one instance for stations 1, 2, 4, 7, and 8 across the three holdout years (Table 3 21). Improvements were visible in plots of observed, hindcasted, and forecasted values for stations 1 and 2. For example, unadjusted values missed some of the variation in observed values at station 1 during the fall and spring of 1998, the summer and fall of 2009, and the spring and summer of 2010, but time series adjusted values performed better (Figur es 3 33A C). However, there were periods where time series adjusted forecasts overestimated fluctuations in observed chlorophyll (e.g., for station 2 during April and May of all years; Figures 3 34A C). Hindcasts and forecasts for stations 3, 5, and 6 also missed some of the fine scale variation in observed values. Overall, the predictions were most accurate at station 3 during 2010 (Table 3 21); nevertheless, predicted values did not match the minimal fluctuation in chlorophyll observed at this station dur ing early 1998 (Figure 3 35A) or February through April of 2009 (Figure 3 35B). Variation at station 5 was not captured

PAGE 67

67 for May to September 2009 or for 2010 (Figures 3 36B and 3 36C), and unadjusted predicted values for station 5 had the worst fit across all stations for the holdout samples (Table 3 21). Lastly, time series adjustments did not improve accuracy of the model for station 6, where large changes in chlorophyll c oncentrations occurred (Figures 3 37A C) In fact, RMSE values were unchanged for 19 98 and 2010 and increased for 2009 (Table 3 21). Like stations 5 and 6, time series adjustments for station 4 had small impacts on the accuracy of predictions, and the only increase in accuracy was observed during 2010 (Table 3 21). This pattern is evident in plots of observed, hindcasted, and forecasted values for each year, with unadjusted predictions tending to miss key variation in observed values (e.g., in September of 1998; March, June, and August of 2009; and November and December of 2010) and time s eries adjusted values overestimating fluctuations during the same time periods (Figures 3 38A C). Finally, the pattern of following overall trends in observed values, but missing larger deviations also was seen in hindcasts and forecasts for stations 7, 8, 9, and 10. For 1998, unadjusted hindcasts at stations 7, 8, and 9 predicted less variation than what was observed (Figures 3 39A, 3 40A, and 3 41A), and hindcasts at station 10 underestimated observed values, particularly in the beginning of the year (Fig ure 3 42A). However, time series adjusted hindcasts more closely followed observed values as indicated by the reduction in RMSE values for this year (Table 3 21). In addition to missing some variation in observed chlorophyll values, initial values for both unadjusted and adjusted 1998 hindcasts were very different from the observed values for all stations (Figures 3 39A, 3 40A, and 3 41A). In contrast, forecasts for 2009 and 2010

PAGE 68

68 followed overall trends well at stations 7, 8, and 9 (Table 3 21), and time series adjusted forecasts did not overestimate fluctuations as much as in 1998 (Figures 3 39B, 3 40B, and 3 41B). At station 10, forecasts performed well in 2009, with the exception of November where an increase in chlorophyll was not predicted (Figu re 3 42B), and the accuracy of forecasts decreased after April 2010 (Figure 3 42C). Overall, forecasts for station 10 were the most accurate of the four stations, with an average RMSE across the three years of 0.27 compared to values of 0.28 for stations 7 and 8 and 0.29 for station 9. Discussion Trend Analyses Few significant trends were detected in water quality time series from the lower Suwannee River and estuary between 1998 and 2010. Additionally, there were no apparent spatial patterns in significant trends and trends in nutrients did not coincide with trends in chlorophyll a. The small number of significant trends may be a consequence of the absence of monotonic trends over the period of study or may be due to the insensitivity of the trend analysis methods to short term trends The lack of significant trends in discharge, salinity, and color may reflect true characteristics of the system because dynamics of these parameters are influenced by physical processes that fluctuate within or between years, with greater stability on longer timescales. For instance, flow in the Suwannee River consists primarily of input from tributaries driven by precipitation and groundwater flux (Grubbs and Crandall 2007). Patterns in precipitation and groundwater flux are influenced strongly by regional and global climatic cycles that occur across multiple years or longer periods; although, events like tropical storms can cause significant, short term increases in flow (Copeland

PAGE 69

69 et al. 2009; Carlson et al. 2010). In fact, t he study period did include El Nio events during 1998 and 200 3 2003 and La Nia events during 1999 2001 and 2007 2008 (Grubbs and Crandall 2007); however, periods of high and low discharge associated with these events and other short term incidents seemed to be sufficiently balanced across the 13 y study period so that no overall trend was detected. Riverine discharge significantly alters several other parameters. Thus, it is not surprising that no trends were detected in salinity or color because these pa rameters are related strongly to discharge in the Suwannee estuary (Bales et al. 2006; Quinlan et al. 2009 ). The lack of trend in discharge also may explain the small number of significant trends in total nitrogen and total phosphorus given that concentrat ions of these nutrients also can be linked to river discharge (Bledsoe and Phlips 2000). Finally, unquantified spatial and temporal variation in freshwater inputs or surface runoff from salt marshes and tidal creeks to the north and south of the river mout h could have masked overall trends. Although most parameters did not have significant monotonic trends across the study period, significant trends were detected at some sites for concentrations of total nitrogen, total phosphorus, and chlorophyll a (Tables 3 2 through 3 4). Significant trends in nutrients did not exhibit obvious spatial patterns as trends were detected in each region of the study area; however, all trends indicated significant increases across the study period. Furthermore, none of the sign ificant nutrient trends coincided with the significant trends in chlorophyll a concentrations detected at stations 3 and 6 The fact that nutrient trends were not manifested as clearly linked trends in chlorophyll a

PAGE 70

70 concentrations may be due to differences in other factors regulating phytoplankton biomass ( e.g., grazing or light availability ). In addition to the possible natural explanations, the small number of significant trends may be an artifact of the trend analysis. One potentially influential aspect was the timescale at which trends were assessed. The potential influence of temporal scale is apparent when trends in total nitrogen are considered. In the present study, total nitrogen trends were assessed over 13 y using methods that were insensitive to trends at shorter timescales (Hirsh and Slack 1984). This insensitivity can explain why significant trends were found at only three stations, although time series plots of total nitrogen concentrations suggested trends at shorter timescales. For example, time series plots for stations 1 and 2 suggested decreasing concentrations of total nitrogen from 1998 through 2002, a large increase the following year, and then a subsequent decrease through 2008 (Figure 3 8). Furthermore, when examined in the context of longer datasets, total nitrogen concentrations are not as stable as the present results suggest. For instance, a 2 3 y dataset showed increases in annual mean total nitrogen concentrations of almost 50% within downstream portions of the river (see Chapter 2 ). Regression Analyses Model interpretation Relationships between chlorophyll a concentrations and a range of explanatory terms varied across time and space, as indicated by differences in model parameters among groups of stations. Nevertheless, there wer e consistent ly significant relationship s between rates of change in chlorophyll a concentrations and rates of change in salinity, color, and concentrations of total phosp horus at all stations. The prevalence of these terms, which are related to water resid ence time and ligh t and

PAGE 71

71 nutrient availability, agree with other work showing these factors are important environmental control s of phytoplankton dynamics (Bledsoe et al. 2004; Lawrenz et al. 2010 ). Short water residence times negatively affect accumulatio n of phytoplankton biomass in the Suwannee River and estuary, where flushing rates range from 1 to 6 days, depending on discharge (Bledsoe et al. 2004). Discharge is closely linked to salinity (Bales et al. 2006); thus, the prevalence of a salinity term ma y represent the influence of residence time on phytoplankton biomass (i.e., residence times increase when discharge is relatively low and salinities rise). The extent of the riverine plume varies across the Suwannee estuary based on discharges, tides, curr ents, and coastal winds (Bales et al. 2006). More specifically, these factors influence how far offshore and alongshore the plume reaches, thereby affecting light and nutrient availability across the system. The water in the plume tends to have high concen trations of colored dissolved organic matter, which lowers light availability ( Bledsoe et al. 2004 ). In fact, color is generally the largest contributor to light attenuation in the estuary (Bledsoe et al. 2004). This relationship between light availability and color helps explain why changes in color concentrations explained considerable variation in chlorophyll a concentrations across all sites considered in this study. Interestingly, a positive relationship was indicated between log 10 color and log 10 chl for the group model for stations 7, 8, 9, and 10, contrary to the expected negative relationships that were observed at the other stations (Table 3 20). This result may be due to unique characteristics of the nearshore region, which are discussed below.

PAGE 72

72 S imilar to light availability, nutrient availability across the system is impacted by changes in the extent of the Suwannee River plume, with subsequent effects on phytoplankton dynamics. Rivers can contribute substantial nutrient loads to coastal systems ( Billen and Garnier 1997; Turner et al. 2007), and large, riverine nutrient loads have been observed in the Suwannee system ( Ashbury and Oaksford 1997 ). Riverine nutrient loads typically support phytoplankton growth, as observed by Bledsoe et al. (2004), w ith bioassays indicating no nutrient limitation in the lower Suwannee River and parts of the Suwannee estuary during most of that study. However, the same study documented nitrogen and phosphorus limitation in the oyster reef and nearshore areas of the est uary during some periods, and nitrogen was commonly the limiting nutrient (Bledsoe et al. 2004). In fact, nitrogen is generally considered to be the primary limiting nutrient in marine systems ( Boynton et al. 1982). Therefore, significant relationships bet ween chlorophyll a and total phosphorus concentrations, instead of total nitrogen concentrations, represented unexpected results (see Frazer et al. 1998). These results may be due to differences in sources and dynamics of nitrogen and phosphorus in the sy stem. Nitrogen in the Suwannee River originates primarily from groundwater sources (Katz et al. 1997), and these nitrogen rich groundwaters sustain the base flow of the river during low flow periods (Katz et al. 1999). On the other hand, phosphorus enters the system through runoff and point sources, which primarily are influenced by weather events and higher discharges (Asbury and Oaksford 1997). Therefore, during periods of low or below average flow, nitrogen will still be supplied to the estuary through g roundwater inputs, but phosphorus inputs will be decreased. Furthermore, the extent of the plume will decrease during low flow periods, resulting in

PAGE 73

73 decreasing phosphorus concentrations further from shore. Over the long term, these differing dynamics could result in nitrogen concentrations being more stable than phosphorus concentrations, which would explain why changes in total phosphorus explained more of the variation in chlorophyll a than changes in total nitrogen concentrations. Nevertheless, further r esearch will be required to fully elucidate the dynamics of nitrogen and phosphorus limitation in the Suwannee estuary, and the present findings highlight the need for continued investigation into impacts from both nitrogen and phosphorus. Other terms exp laining significant variation in chlorophyll a also were related to residence time and the availability of light and nutrients, although these relationships varied among stations. Such differences across the study area are not surprising given results of p revious studies showing differences in nutrient limitation of phytoplankton growth, primary drivers of light availability, and influence of the river plume among the river, oyster reef, and nearshore areas of the Suwannee Estuary. Within the river, short residence times and low light availability (relative to other parts of the system) tend to limit accumulation of phytoplankton biomass despite the presence of sufficient nutrients to support growth (Bledsoe and Phlips 2000; Bledsoe et al. 2004). These char acteristics were observed in the present study for stations 1 and 2, which had the highest average concentrations of total nitrogen, total phosphorus, and color and the lowest average concentrations of chlorophyll a across the study period (Appendix A). Th e importance of residence time also was indicated by the significance of terms for discharge in the previous month and an interaction term between salinity and total phosphorus. The lagged term shows the potential for delayed responses to

PAGE 74

74 changes in discha rge, and the negative coefficient for the interaction term can be interpreted as an antagonistic interaction between salinity and total phosphorus. More specifically, this interaction indicates the potential for the importance of one explanatory variable t o be decreased by the presence of another variable ( Cohen et al. 2003 ). In this case, an expected increase in phytoplankton biomass due to increased total phosphorus concentrations can be reduced if residence time decreases as indicated by a decrease in s alinity. Residence time and availability of light also impact phytoplankton dynamics in the oyster reef area of the Suwannee estuary. In the oyster reef region, increased discharge shortens residence times and decreases light availability (Bledsoe et al. 2004). However, the reef structure can slow flushing to some extent, which can allow phytoplankton biomass to accumulate (Bledsoe et al. 2004). Residence times in the reef region also can be increased by onshore winds, as indicated by the significance of a term for a 36 h southwest wind event prior to sampling in the model for stations 3, 5, and 6 (Table 3 18 ) In terms of light availability, models for station 4 and stations 3, 5, and 6 indicated similar, negative relationships between log 10 color and log 1 0 chl a (Tables 3 16 and 3 18). Estimated coefficients for each model were nearly equal but smaller than the coefficient in the model for stations 1 and 2 (Table 3 13), which suggests an increased contribution to light attenuation from tripton in the shall ower, reef area as previously demonstrated by Bledsoe et al., (2004). Generally shallower depths at the reef stations (Frazer et al. 1998) also help to explain the significant effect of temperature in each of the relevant models, because shallow water woul d experience a

PAGE 75

75 wider range of temperatures, including periods of cold that could reduce phytoplankton growth rates ( Irwin and Finkel 2008 ). In contrast to the river and reef, the nearshore region of the estuary is frequently nutrient limited, with suffici ent light to support phytoplankton growth (Bledsoe et al. 2004). These characteristics can be linked to the decreased influence of the river ine plume further from shore (Bledsoe et al. 2004; Quinlan and Phlips 2007). However, the plume periodically extends across the nearshore region during periods of high flow, and the significance of these events for phytoplankton dynamics in the region was highlighted in the significant parameters contained in models for stations 7, 8, 9, and 10. For instance, five of th e nineteen terms were related to discharge, and three of these terms specifically represented periods of high discharge when the plume is expected to impact the nearshore sites (i.e., dummy variables for Greater than 50% discharge Storm event and October 2004 ). When the plume extends offshore, light attenuation and nutrient availability increase. The influences of these changes are represented in the regression model by significant parameters for total phosphorus concentrations and color (Table 3 20). Coe fficient estimates for both terms indicated a positive effect on chlorophyll a concentrations, which was unexpected for the color term. The positive relationship with color may reflect the dominant effect of increased nutrient concentrations. Evaluation o f modeling methods The combination of multiple regression and time series techniques provided insight into relationships between chlorophyll a and water quality parameters in the lower Suwannee River and estuary despite some known limitations.

PAGE 76

76 A primary a dvantage of the regression approach was the ability to highlight parameters explaining variation in chlorophyll a dynamics, which can provide more information than basic correlative analyses. For example, Bledsoe et al (2004) hypothesized that sustained s outhwesterly winds could increase residence time within the estuary resulting in an accumulation of phytoplankton biomass. This hypothesis was based on observations of elevated chlorophyll concentrations after periods of onshore winds and it is supported by results for two groups of stations in the present study. Models for reef stations (3, 5, and 6) and the nearshore stations (7, 8, 9, and 10) indicated a positive relationship between log 10 chl a concentrations and a 36 h period of southwesterly winds pr ior to the sampling event (Tables 3 18 and 3 20). Despite these results, a direct link cannot be confirmed between these or any other explanatory variables until cause and effect relationships have been explored through experimental studies. In addition t o suggesting future experimental studies results highlighted the need to explore other factors that influence chlorophyll a dynamics. One area highlighted for further consideration was the influence of salt marsh es and tidal creek s around the estuary. Sam pling has shown higher nutrient concentrations in these areas relative to the oyster reef and nearshore parts of the estuary with no straightforward relationships between these sources and discharge from the Suwannee River (FDEP 2010). The unknown influen ce of these inputs may explain why models for stations 3, 4, 5, and 6 in the reef region had the poorest fit. Future modeling should attempt to quantify freshwater input and nutrient loads from these areas in addition to inputs originating from the main ch annel of the Suwannee River, which was the focus of this study.

PAGE 77

77 Another advantage of the modeling arose from the time series component s which were used to improve fit and satisfy assumptions regarding residual independence. Although time series adjustmen ts did not improve fit to the extent expected, adjustments did bring predicted values closer to observed values where large deviations occurred. This improvement is likely due to the presence of inherent serial autocorrelation in biomass accumulation i.e. the amount of biomass at a given time is dependent on the amount of biomass present at one or more earlier time steps. In fact, in a chlorophyll time series model developed by Lehman (1992), exogenous variables explained very little of the observed varia nce in comparison to temporal correlation among chlorophyll concentrations. On the other hand, decreased fit after time series adjustments to models of the Suwannee data most often was due to exaggeration of large deviations. For example, the time series c omponent in the station 4 model improved predictions of the timing of larger deviations in chlorophyll a values, but the magnitudes of predicted increases were too high (e.g., in September of 2000; Figure 3 38). The prevalence of overestimates points to t he value of incorporating other controls on phytoplankton biomass. For example the model s did not incorporate competition for resource s grazing pressure, losses due to sinking or senescence, or other factors that can regulate large increases phytoplankto n biomass; therefore predictions exceeded values that would occur naturally. A failure to incorporate biological controls and other factors also may help explain differences in the performance of models among stations and time periods. For example, grazi ng by zooplankton, bacteria, and benthic filter feeders has been documented as an important top down control of phytoplankton biomass in many

PAGE 78

78 coastal systems ( Strom and Strom 1996; Li and Smayda 1998; Calbet and Landry 2004 ) with the potential for grazing by microzooplankton, in particular, reported to yield large decreases in phytoplankton standing crop in the Suwannee estuary ( Robinson 2007 ; Quinlan et al. 2009). In the present study, grazing pressure may have varied across sites as previously demonstra ted across zones indentified by differ ing salinit ies (Robinson 2007). Temporal variation in grazing pressure also may have been generated by changes in temperature. For instance, Li and Smayda (1998) suggested that warmer temperatures during the winter and spring enhanced zooplankton growth and grazing rates enough to impact phytoplankton blooms in Narragansett Bay, Rhode Island. In addition to unquantified factors, the use of ordinary least squares (OLS) required transformation s to satisfy assumptions for h om os c edasticity and normality of residuals, and the method only assessed linear, unidirectional relationships. Water quality data tend to be non normal and have non constant variance s (Carey 2009 ); therefore, a log transformation was needed to attain h om os c edasticity and normality of residuals. However, this transformation alters the interpretation of relationships identified in the model which may complicate interpretations for managers attempting to use raw data ( Osborne 2002 ). Furthermore, the inabili ty to model complex non linear relationships obviates evaluation of feedback loops, such as increasing competition for nutrients as phytoplanktonic biomass accumulates Other approaches (e.g., multivariate or hierarchical models) can accommodate such inte ractions, and they should be considered in future modeling studies.

PAGE 79

79 Summary The trend and regression analyses presented here provided insight into the dynamic relationships between chlorophyll a and other water quality parameters in the Suwannee River and its estuary. This information highlights the importance of considering spatial and temporal variability and the results will be valuable for managers working to address water quality concerns in the lower Suwannee River, its estuary, and other affected ar eas. Spatial and temporal variation in relationships between chlorophyll a and other water quality parameters were highlighted using time series and regression modeling. Spatial variability was linked to river plume dynamics, which generates differences i n top down and bottom up controls on phytoplankton biomass across the study area. The b ottom up influence of nutrient s and light were consistently important as indicated by presence of significant terms for color and concentrations of total phosphorus in a ll models. The significance of these influences, and other terms related to water residence time, corroborated results from previous studies that noted the influence of the river plume. However, the importance of total phosphorus noted here contradicted findings by Bledsoe et al. (2004) which identified nitrogen as the primary limiting nutrient in the estuary. This inconsistency may be due to differences in relative variability in nitrogen and phosphorus loads but further research will be required to fu lly understand the relevant dynamics Other studies also have reported the potential for significant top down control of phytoplankton biomass that can vary with plume dynamics (Robinson 2007; Quinlan et al. 2009) and previously reported variation in grazing may explain the variable fit of regression models among groups of stations. Overall, the importance of the river ine plume as a regulator of phytoplankton biomass across the

PAGE 80

80 study area suggests that managers should contemplate impacts from altere d flow along with nutrient loads and other sources of pollution. Variations in fit also highlighted the importance of considering climatic cycles. For instance, conclusions based on data collected under only one set of hydrologic conditions may not be ap plicable in different circumstances, which could lead to misguided management decisions. Furthermore, the importance of timescales was highlighted in the trend analyses where trends at shorter timescale s averaged out over the full study Such variation be comes important when choosing a baseline for reference or target values for criteria as well as when assessing the success of management actions (e.g., l ower nutrient concentrations). With these considerations in mind, future work should includ e a balanc e between fine scale studies and large scale modeling. Fine scale experimental or descriptive studies could help clarify the causes and magnitude of temporal and spatial variability in processes across the study area. For example, if patterns in phosphorus limitation coincide with agricultural planting seasons, efforts to reduce fertilizer runoff can be stressed. On a larger scale, the modeling approach demonstrated in this study can be applied across the Big Bend region to identify which driving factors re main important at a regional scale. Such information could identify similarities or differences among systems and support customized management initiatives.

PAGE 81

81 Figure 3 1. Suwannee River basin with the A) Withlacoochee sub basin, B) Alapaha River sub basin C) Upper Suwannee River sub basin, D) Suwannee River sub basin, and E) Santa Fe River sub basin illustrated. Data source: Suwannee River Water Management District ( http://www.srwmd.state.fl.us/ index.aspx?NID=319 ). Table 3 1. Supplementary data sources. ( NOAA: National Oceanographic and Atmospheric Association ) Parameter Data source Site ID Site name Mean daily discharge United States Geological Survey 2323592 Above Gopher River 2323500 Wilcox Wind Direction NOAA National Data Buoy Center KTNF1 K eaton Beach Water Color Suwannee River Water Management District SUW275C1 At Gopher River SUW305C1 West Pass

PAGE 82

82 Figure 3 2. Water quality sampling sites with the A) river, B) reef, and C) nearshore regions illustrated (Quinlan 2003; Raabe et al. 2007) Oyster reefs are outlined in gray.

PAGE 83

83 Table 3 2. Details and results of trend analyses for total nitrogen by station. Station Trend Test S P 1 Adjusted Seasonal Kendall 254 0.271 0.038 2 Mann Kendall with bootstrap 36 0.003 0.956 3 Adjusted Seasonal Kendall 44 0.047 0.698 4 Mann Kendall with bootstrap 39 0.003 0.953 5 Seasonal Kendall 151 0.161 0.008 6 Mann Kendall with bootstrap 6 0.000 0.993 7 Adjusted Seasonal Kendall 171 0.183 0.118 8 Mann Kendall with bootstrap 82 0.007 0. 900 9 Mann Kendall with bootstrap 96 0.008 0.882 10 Mann Kendall with bootstrap 20 0.002 0.975 Table 3 3. Details and results of trend analyses for total phosphorus by station. Station Trend Test S P 1 Mann Kendall with bootstra p 41 0.003 0.950 2 Mann Kendall with bootstrap 19 0.002 0.976 3 Adjusted Seasonal Kendall 2 0.002 0.993 4 Adjusted Seasonal Kendall 88 0.094 0.354 5 Adjusted Seasonal Kendall 74 0.079 0.451 6 Adjusted Seasonal Kendall 91 0.097 0.370 7 Adjusted Seas onal Kendall 57 0.061 0.626 8 Adjusted Seasonal Kendall 92 0.098 0.447 9 Adjusted Seasonal Kendall 196 0.209 0.075 10 Adjusted Seasonal Kendall 134 0.143 0.135

PAGE 84

84 Table 3 4. Details and results of trend analyses for chlorophyll a by station. Station Tren d Test S P 1 Seasonal Kendall 38 0.042 0.504 2 Seasonal Kendall 81 0.089 0.151 3 Seasonal Kendall 92 0.100 0.106 4 Seasonal Kendall 14 0.015 0.817 5 Adjusted Seasonal Kendall 69 0.076 0.204 6 Adjusted Seasonal Kendall 107 0 .117 0.148 7 Adjusted Seasonal Kendall 59 0.064 0.581 8 Mann Kendall with bootstrap 66 0.006 0.918 9 Mann Kendall 660 0.055 0.308 10 Mann Kendall 637 0.054 0.320 Table 3 5. Details and results of trend analyses for color by station. Station Trend T est S P 1 Adjusted Seasonal Kendall 100 0.107 0.441 2 Adjusted Seasonal Kendall 31 0.033 0.821 3 Adjusted Seasonal Kendall 12 0.013 0.934 4 Adjusted Seasonal Kendall 38 0.041 0.786 5 Adjusted Seasonal Kendall 13 0.014 0.929 6 Man n Kendall with bootstrap 13 0.001 0.985 7 Mann Kendall with bootstrap 85 0.007 0.897 8 Mann Kendall with bootstrap 78 0.007 0.905 9 Mann Kendall with bootstrap 58 0.005 0.929 10 Mann Kendall with bootstrap 115 0.010 0.860 Table 3 6. Details and results of trend analyses for salinity by station. Station Trend Test S P 1 Adjusted Seasonal Kendall 38 0.041 0.735 2 Adjusted Seasonal Kendall 37 0.040 0.758 3 Adjusted Seasonal Kendall 59 0.063 0.628 4 Adjusted Seasonal Ke ndall 20 0.021 0.863 5 Adjusted Seasonal Kendall 31 0.033 0.737 6 Adjusted Seasonal Kendall 67 0.072 0.537 7 Adjusted Seasonal Kendall 111 0.119 0.343 8 Mann Kendall with bootstrap 3 0.000 0.996 9 Mann Kendall with bootstrap 10 0.001 0.988 1 0 Mann Kendall with bootstrap 9 0.001 0.989

PAGE 85

85 Table 3 7. Analysis of variance for model used to estimate missing color values with log 10 color as the response variable. Asterisks indicate a dummy variable. SSE d f F P Intercept 15.46 1 284.86 0.000 Sta tion* 78.19 9 160.03 0.000 Hi flow year* 0.53 1 9.75 0.002 50 th percentile discharge* 0.89 1 16.37 0.000 One day lag: l og 10 discharge 4.73 1 87.17 0.000 Two day lag: l og 10 discharge 4.06 1 74.77 0.000 Residuals 68.35 1259 Table 3 8. Analysis of va riance for model used to estimate missing salinity values with log 10 salinity as the response variable. Asterisks indicate a dummy variable. SSE d f F P Intercept 4.35 1 88.29 0.000 Station* 20.64 9 46.56 0.000 Log discharge 0.39 1 7.85 0.005 One month lag: l og 10 salinity 3.53 1 71.68 0.000 Two month lag: l og 10 salinity 1.36 1 27.54 0.000 Two day lag: l og 10 discharge 0.74 1 14.99 0.000 One month lag: l og 10 discharge 0.37 1 7.41 0.007 95 th percentile discharge* 0.90 1 18.29 0.000 Residuals 68. 25 1386

PAGE 86

86 Table 3 9. Analysis of variance for model used to estimate missing total nitrogen values with log 10 total nitrogen as the response variable. Asterisks indicate a dummy variable. SSE d f F P Intercept 4.48 1 272.84 0.000 Station* 9.89 9 66.89 0.000 Month* 2.10 11 11.62 0.000 High discharge year* 0.21 1 12.57 0.000 One month lag: l og 10 total nitrogen 1.19 1 72.26 0.000 Two month lag: l og 10 discharge 0.11 1 6.95 0.008 One day lag: l og 10 discharge 0.70 1 42.85 0.000 95 th percentile dis charge* 0.11 1 6.74 0.010 90 th percentile discharge* 0.18 1 10.81 0.001 Residuals 22.14 1348 Table 3 10. Analysis of variance for model used to estimate missing total phosphorus values with log 10 total phosphorus as the response variable. Asterisks i ndicate a dummy variable. SSE d f F P Intercept 0.70 1 22.01 0.000 High discharge year* 0.17 1 5.27 0.022 One month lag: l og 10 total phosphorus 58.52 1 1838.71 0.000 One month lag: l og 10 discharge 3.29 1 103.42 0.000 One day lag: l og 10 dischar ge 1.82 1 57.31 0.000 90 th percentile discharge* 0.44 1 13.92 0.000 Station* 0.14 2 2.18 0.114 Month* 3.02 11 8.61 0.000 Residuals 43.48 1366 Table 3 11. Analysis of variance for model used to estimate miss ing mean daily discharge values for the Go pher River dataset. Asterisks indicate a dummy variable. SSE d f F P Intercept 2.28E+08 1 142.31 0.000 Wilcox discharge 1.52E+10 1 9447.39 0.000 Squared Wilcox discharge 2.32E+07 1 14.48 0.000 Residuals 6.17E+09 3841

PAGE 87

87 Figure 3 3. Tw o dimensional ordination of sampling stations (stress value = 0.01) based on range transformed values of temperature, salinity, color, and concentrations of total nitrogen and total phosphorus Short distances between points indicate similarity, with point s farther apart being less similar. Table 3 12. Analysis of variance for group model for stations 1 and 2 with log 10 chlorophyll a as the respons e variable. Asterisks indicate dummy variable s SSE d f F P Intercept 1.31 1 36.70 0.000 Month* 4.94 11 12.59 0.000 Log 10 color 1.62 1 45.40 0.000 One month lag: l og 10 total p hosphorus 0.29 1 8.07 0.005 Log 10 (salinity X total phosphorus) 0.29 1 8.00 0.005 Log 10 salinity 0.60 1 16.75 0.000 Log 10 total phosphorus 0.30 1 8.52 0.004 One month lag: l og 10 discha rge 0.86 1 23.96 0.000 99 th percentile chlorophyll* 1.38 1 38.63 0.000 November 2007* 0.22 1 6.05 0.015 Residuals 7.75 217

PAGE 88

88 Table 3 13. Coefficient estimates for group model for stations 1 and 2. Asterisks indicate dummy variable s Coefficient S.E. t P Intercept 2.06 0.38 5.40 0.000 February* 0.06 0.06 1.06 0.290 March* 0.12 0.06 1.92 0.057 April* 0.31 0.06 5.11 0.000 May* 0.51 0.06 7.99 0.000 June* 0.30 0.06 4.82 0.000 July* 0.26 0.06 4.28 0.000 August* 0.25 0.06 4.04 0.000 September* 0.24 0.06 3.79 0.000 October* 0.26 0.06 4.20 0.000 November* 0.04 0.06 0.67 0.502 Dece mb er* 0.04 0.06 0.58 0.561 Log 10 color 0.37 0.05 6.74 0.000 One month lag: l og 10 total phosphorus 0.36 0.13 2.84 0.005 Log 10 (salinity X total phosphorus) 0.26 0. 09 2.83 0.005 Log 10 salinity 0.60 0.15 4.09 0.000 Log 10 total phosphorus 0.47 0.16 2.92 0.004 One month lag: l og 10 discharge 0.30 0.06 4.90 0.000 99 th percentile chlorophyll* 0.72 0.12 6.22 0.000 November 2007 0.37 0.15 2.46 0.015 Table 3 14. R esults of Box tests for autocorrelation for station residuals extracted from group models. Station 2 d f P 1 12.59 1 0.000 2 3.97 1 0.046 3 2.63 1 0.103 4 8.70 1 0.003 5 4.32 1 0.038 6 7.56 1 0.006 7 5.49 1 0.019 8 6.76 1 0.009 9 6.52 1 0.011 10 0.23 1 0.632

PAGE 89

89 Table 3 15. Analysis of variance for station 4 model. SSE d f F P Intercept 0.56 1 8.53 0.004 Log 10 total phosphorus 0.38 1 5.83 0.017 Log 10 color 0.53 1 8.08 0.005 Log 10 salinity 0.83 1 12.63 0.001 Log 10 temperature 0.68 1 10.31 0.002 Residuals 7.49 111 Table 3 16. Coefficient estimates for station 4 model. Coefficient S.E. t P Intercept 1.09 0.37 2.92 0.004 Log 10 total phosphorus 0.51 0.21 2.42 0.017 Log 10 color 0.26 0.09 2.84 0.005 Log 10 salinity 0.24 0.07 3.55 0.001 Log 10 temperature 0.72 0.22 3.21 0.002 Table 3 17. Analysis of variance for group model for stations 3, 5, and 6. Asterisks indicate dummy variable s SSE d f F P Intercept 4.10 1 48.83 0.000 Log 10 salinity 1.44 1 17.16 0.000 Log 10 total phosphorus 3.14 1 37.33 0.000 Log 10 color 0.83 1 9.88 0.002 Log 10 temperature 4.24 1 50.43 0.000 S outhwest 36 h wind event* 0.38 1 4.55 0.034 Highest 50% discharge* 1.38 1 16.39 0.000 Residuals 29.49 351

PAGE 90

90 Table 3 18. Coefficient estimates for group model for stations 3, 5, and 6. Asterisks indicate dummy variable s Coefficient S.E. t P Intercep t 1.70 0.24 6.99 0.000 Log 10 salinity 0.35 0.08 4.14 0.000 Log 10 total phosphorus 0.64 0.10 6.11 0.000 Log 10 color 0.24 0.08 3.14 0.002 Log 10 temperature 0.99 0.14 7.10 0.000 Southwe st 36 h wind event* 0.17 0.08 2.13 0.034 Highest 50% discharge* 0.17 0.04 4.05 0.000 Table 3 19. Analysis of variance for group model for stations 7, 8, 9, and 10. Asterisks indicate dummy variable s SSE d f F P Intercept 0.85 1 18.86 0.000 Log 10 total phosphorus 3.89 1 86.34 0.000 Highest 50% discharge* 1.58 1 35.05 0.000 Southwe st 36 h wind event* 0.40 1 8.85 0.003 Log 10 discharge 0.88 1 0.88 0.000 Log 10 color 0.23 1 5.21 0.023 October 2004* 2.42 1 53.65 0.000 Log 10 salinity 0.59 1 13.21 0.000 Month* 2.18 11 4.39 0.000 Storm event* 0.60 1 13.44 0.000 R esiduals 20.53 456

PAGE 91

91 Table 3 20. Coefficient estimates for group model for stations 7, 8, 9, and 10. Asterisks indicate dummy variable s Coefficient S.E. t P Intercept 1.86 0.43 4.34 0.000 Log 10 total phosphorus 0.71 0.08 9.29 0.000 Highest 50% disc harge* 0.19 0.03 5.92 0.000 Southwe st 36 h wind event* 0.15 0.05 2.98 0.003 Log 10 discharge 0.34 0.08 4.43 0.000 Log 10 color 0.14 0.06 2.28 0.023 October 2004* 1.02 0.14 7.33 0.000 Log 10 salinity 0.79 0.22 3.63 0.000 February 0.18 0.05 3.73 0.000 March 0.06 0.05 1.28 0.203 April 0.08 0.05 1.57 0.117 May 0.13 0.05 2.44 0.015 June 0.04 0.05 0.80 0.427 July 0.00 0.05 0.09 0.930 August 0.05 0.05 1.03 0.306 September 0.02 0.05 0.39 0.698 October 0.04 0.05 0.69 0.489 Nove mber 0.05 0.05 0.94 0.350 December 0.03 0.05 0.57 0.569 Storm event 0.22 0.06 3.67 0.000

PAGE 92

92 Table 3 21. Root mean squared error (RMSE) values for individual station models calculated for each year. Values in bold are RMSE calculated after the appropr iate time series adjustment was applied. Dashes indicate no ti me series adjustments were applied. Station 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 1998 2010 1 0.26 0.23 0.17 0.15 0.16 0.16 0.15 0.12 0.18 0.18 0.20 0.20 0.25 0.17 0.26 0.21 0.19 0.15 0.14 0.13 0.15 0.11 0.18 0.20 0.20 0.17 0.24 0.17 2 0.32 0.23 0.20 0.14 0.18 0.16 0.12 0.12 0.17 0.19 0.20 0.20 0.24 0.17 0.31 0.22 0.19 0.18 0.16 0.15 0.11 0.12 0.17 0.16 0.20 0.20 0.24 0.17 3 0.34 0.21 0.26 0.22 0.23 0.26 0.32 0.2 9 0.32 0.42 0.36 0.39 0.17 0.29 4 0.28 0.23 0.29 0.29 0.22 0.24 0.25 0.30 0.26 0.24 0.21 0.17 0.26 0.25 0.30 0.21 0.27 0.28 0.20 0.23 0.28 0.26 0.27 0.23 0.20 0.19 0.24 0.24 5 0.33 0.32 0.23 0.23 0.22 0.30 0.32 0.33 0.30 0 .16 0.35 0.33 0.34 0.28 0.35 0.31 0.24 0.22 0.23 0.31 0.31 0.31 0.29 0.17 0.35 0.35 0.37 0.27 6 0.32 0.32 0.24 0.24 0.27 0.29 0.35 0.35 0.30 0.18 0.19 0.31 0.26 0.27 0.32 0.30 0.26 0.24 0.28 0.29 0.33 0.32 0.30 0.21 0.19 0.36 0.26 0.27 7 0.39 0.18 0. 13 0.10 0.14 0.20 0.15 0.32 0.27 0.14 0.19 0.18 0.23 0.18 0.39 0.16 0.13 0.10 0.15 0.21 0.16 0.33 0.25 0.14 0.19 0.19 0.21 0.18 8 0.35 0.29 0.13 0.13 0.14 0.33 0.18 0.23 0.26 0.16 0.14 0.28 0.28 0.20 0.34 0.24 0.13 0.16 0.14 0.33 0.21 0.22 0.22 0.17 0 .16 0.30 0.25 0.20 9 0.38 0.23 0.14 0.15 0.13 0.19 0.24 0.18 0.28 0.18 0.13 0.22 0.23 0.18 0.38 0.20 0.14 0.15 0.14 0.18 0.25 0.18 0.27 0.19 0.13 0.23 0.23 0.18 10 0.27 0.49 0.22 0.33 0.32 0.22 0.13 0.20 0.23 0.16 0.21 0.22 0.26 0.25

PAGE 93

93 Figure 3 4 Time series plot of chlorophyll a concentrations ( 1 ) for stations 1 and 2 Figure 3 5 Time series plot of chlorophyll a concentrations ( 1 ) for stat ions 3, 5 and 6

PAGE 94

9 4 Figure 3 6 Time series plot of chlorophyll a concentrations ( 1 ) for station 4. Figure 3 7 Time series plot of chl orophyll a concentrations ( 1 ) for station s 7, 8, 9, and 10.

PAGE 95

95 Figure 3 8 Time series plot of total nitrogen concentrations ( g L 1 ) for stations 1 and 2 Figure 3 9 Time series plot of total nitrogen concentrations ( 1 ) for stat ions 3, 5 a nd 6

PAGE 96

96 Figure 3 10 Time series plot of total nitrogen concentrations ( 1 ) for station 4. Figure 3 11 Time series plot of total nitrogen concentrations ( 1 ) for station s 7, 8, 9, and 10.

PAGE 97

97 Figure 3 12 Time series plot of total phosphorus co ncentrations ( g L 1 ) for stations 1 and 2 Figure 3 13 Time series plot of total phosphorus concentrations ( 1 ) for stat ions 3, 5 and 6

PAGE 98

98 Figure 3 14 Time series plot of total phosphorus concentrations ( 1 ) for station 4. Figure 3 15 Time series plot of total phosphorus concentrations ( 1 ) for station s 7, 8, 9, and 10.

PAGE 99

99 Figure 3 16 Time series plot of color (PCU) for stations 1 and 2 Figure 3 17 Time series plot of color (PCU) for stat ions 3, 5 and 6

PAGE 100

100 Figure 3 18 Time series plot of color (PCU) for station 4. Figure 3 19 Time series plot of color (PCU) for station s 7, 8, 9, and 10.

PAGE 101

101 Figure 3 20 Time series plot of water temperature (C) for stations 1 and 2 Figure 3 21 Time series plot of water temperature ( C) for stations 3, 5, and 6.

PAGE 102

102 Figure 3 22 Time series plot of water temperature (C) for station 4. Figure 3 23 Time series plot of water temperature (C) for stations 7, 8, 9, and 10.

PAGE 103

103 Figure 3 24 Time series plot of salinity (ppt) for stations 1 and 2 Figure 3 25 Time series plot of salinity (ppt) for stat ions 3, 5 and 6

PAGE 104

104 Figure 3 26 Time series plot of salinity (ppt) for station 4. Figure 3 27 Time series plot of salinity (ppt) for station s 7, 8, 9, and 10.

PAGE 105

105 Figure 3 28 Time series plot of mean daily discharge (m 3 s 1 ) at the Gopher River gauge station.

PAGE 106

106 Figure 3 29. Time series plots of monthly o bserved and fitted values of log 10 transformed chlorophyll a concentrations from models for A) station 1 a nd B) station 2 Observed values are represented with a solid grey line. Time series adjusted fitted values are represented with a solid black line and unadjusted fitted values are represented with a dashed black line. A) B)

PAGE 107

107 Figure 3 30. Time series plots of monthly o bserved and fitted values of log 10 transformed chlorophyll a concentrations from models for station 4 Observed values are represented with a solid grey line. Time series adjusted fitted values are represented with a solid black line and unadjusted fitted values are represented with a dashed black line.

PAGE 108

108 Figure 3 31. Time series plots of monthly o bserved and fitted values of log 10 transformed chlorophyll a concentrations from models for A) station 3, B) s tation 5, and C) station 6 Observed values are represented with a solid grey line. Time series adjusted fitted values are represented with a solid black line and unadjusted fitted values are represented with a dashed black line. A) B)

PAGE 109

109 Figure 3 31. Continued C)

PAGE 110

110 Figure 3 32. Time series plots of monthly o bserved and fitted values of log 10 transformed chlorophyll a concentrations from models for A) station 7, B) station 8, C) station 9, and D) station 10 Observed values are represented w ith a solid grey line. Adjusted fitted values are represented with a solid black line. Unadjusted fitted values are represented with a dashed black line. A) B)

PAGE 111

111 Figure 3 32. Continued C) D)

PAGE 112

112 Figure 3 33. Time series plots of hindcasts and forecasts o f log 10 transformed chlorophyll a concentrations for station 1 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. Time series adjusted values are represent ed with a solid black line and unadj usted values are represented with a dashed black line. A) B) C)

PAGE 113

113 Figure 3 34. Time series plots of hindcasts and forecasts of log 10 transformed chlorophyl l a concentrations for station 2 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. Time series adjusted values are represent ed with a solid black line and unadjusted values are represented with a dashed black line. A) B) C)

PAGE 114

114 Figure 3 35. Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 3 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. U nadjusted values are represented with a dashed black line. A) B) C)

PAGE 115

115 Figure 3 36. T ime series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 5 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. Time series adjusted values are represent ed with a solid black line and unadjusted values are represented with a dashed black line. A) B) C)

PAGE 116

116 Figure 3 37. Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 6 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. Time series adjusted values are represent ed with a solid black line and unadjusted values are represented with a dashed black line. A) B) C)

PAGE 117

117 Figure 3 38. Time seri es plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 4 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. Time series adjusted values are represen t ed with a solid black line and unadjusted values are represented with a dashed black line. A) C) B)

PAGE 118

118 Figure 3 39. Time series plots of hindcasts and forecasts of log 10 transformed chlorophyl l a concentrations for station 7 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. Time series adjusted values are represent ed with a solid black line and unadjusted values are represented with a dashed black line. A) B) C)

PAGE 119

119 Figure 3 40. Time series plots of hindcasts and forecasts of log 10 transformed chlorophyll a concentrations for station 8 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. Time series adjusted values are represent ed with a solid black line and unadjusted values are represented with a dashed black line. A) B) C)

PAGE 120

120 Figure 3 41. Time series plots of hindcasts and forecasts of log 10 transformed chlorophyl l a concentrations for station 9 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. Time series adjusted values are represent ed with a solid black line and unadjusted values are represented with a dashed black line. A) B) C)

PAGE 121

121 Figure 3 42. Time series plots of hindc asts and forecasts of log 10 transformed chlorophyll a concentrations for station 10 and f or holdout samples from A) 1998, B) 2009 and C) 2010. Observed values are represented with a solid grey line. Unadjusted values are represented with a dashed black li ne. A) B) C)

PAGE 122

122 C HAPTER 4 CONCLUSIONS This study characterized trends and relationships for water quality parameters measured over decades at multiple stations within the Suwannee River and its estuary. Results of these analyses provided an improved understanding of w ater quality dynamics in the Suwannee system and highlighted several important considerations for managers. In spite of an improve d understanding of some aspects of nitrogen and chlorophyll dynamics, the overall complexity of the Suwannee system remains a challenge which has implications for the management of this system and others. This complexity was most apparent in links between results from this study and previous work where diverse influential factors included short term weather events long term c limatic cycles, coastal hydrodynamics, groundwater transport biogeochemistry, social dynamics (e.g., fertilizer use and land use decisions), and biological processes. Beyond identifying these factors as important components of chlorophyll and nutrient dyn amics worthy of further research, the links also highlighted challenges facing managers as they address water quality concerns such as nutrient impairment. More specifically, it is unrealistic to expect a complete and detailed understand ing of the intric acies of nutrient dynamics or biological responses such as phytoplankton abundance and distribution. Attaining an adequate understanding of these interactions and influences is especially challenging given the limited amount of time and resources typicall y available. With these challenges in mind, this study highlighted some key considerations for managers seeking to define attainable magnitude, duration, and frequency components of water quality criteria.

PAGE 123

123 One consideration is the influence of events and c limatic cycles, which cannot be managed, but potential ly affect nutrient and phytoplankton dynamics. The influence of events and cycles was evident in nutrient trends in the river and estuary where increases in total nitrogen concentrations around 2003 mos t likely were related to increased rainfall associated with an El Nio event (e.g., Figures 3 8 and 3 1 ). However, a closer examination of the time series indicated that climatic conditions prior to this event probably affect ed the system response For i nstance, larger increases in total nitrogen were observed after the 2003 El Ni o event than after the relatively stronger 1997 1998 El Nio because the former event was preceded by a sustained drought and the latter was preceded by relatively normal rainfa ll ( Crandall et a l. 1999; Copeland et al. 2009). Changes in precipitation due to climatic cycles also can affect phytoplankton abundance and distribution indirectly by altering nutrient loads, residence time s or light availability (Justic et al. 1997). Be cause impacts from unmanageable factors like climatic cycles can be significant, managers should strive to collect data across a range of conditions and time periods in an effort to document the full range of variability This information will assist manag ers as they define the frequency and duration components of water quality criteria. Another consideration for managers is the spatial scale of sampling and analysis. The scale should be chosen carefully and it should match the scale at which the process or organism of interest opera tes (Levin 1992 ). When scales are mismatched or not considered, results may lead to false conclusions (Warner et al. 1995). Although determining appropriate scales is challenging, modeling like that implemented in this study can help identify drivers at a given scale and multiple models targeting different

PAGE 124

124 scales can be used to evaluate how patterns change across scales (Levin 1992 ). Such information will help researchers and managers interpret and apply results more appropri ately, even if the ideal scale of sampling or analysis cannot be attained due to logistical constraints Finally, it will be challenging to address the issues above without broad scale, long term datasets. Such datasets place short term and localized event s in the context of long term broad scale patterns, which can help identify potential drivers (e.g., climatic cycles) and distinguish unusual event s that require further evaluation (Magnuson 1995; Burt et al. 2010). Long term datasets are also the only wa y to assess processes that occur over a long timescale. For example, groundwater residence times in the Upper Floridan aquifer are expected to delay responses to nutrient management (Katz et al. 1999; Upchurch et al. 2007). Without the information provided by long term datasets, the effectiveness of a particular management s trategy could be underestimated or overvalued which will lead to poor outcomes (Burt et al. 2010).

PAGE 125

125 APPENDIX DESCRIPTIVE STATISTI CS Chlorophyll a Table A 1. Descriptive statistics for c hlorophyll a ( g L 1 ) at station 1. Time Period Mean Median Standard deviation Minimum Maximum 1998 1.01 0.44 1.22 0.17 4.30 1999 3.49 1.71 3.75 0.37 12.43 2000 1.65 1.21 1.34 0.23 4.91 2001 1.23 0.66 1.45 0.23 5.27 2002 3.14 2.75 2.12 0.37 6.52 2003 0.53 0.44 0.3 2 0.23 1.24 2004 1.51 0.55 1.78 0.21 5.16 2005 0.53 0.49 0.29 0.23 1.27 2006 5.85 4.08 8.51 0.21 30.51 2007 2.23 1.23 2.01 0.34 6.14 2008 1.90 0.95 2.03 0.19 5.81 2009 1.35 0.78 1.09 0.34 3.24 2010 1.84 1.28 1.53 0.00 4.92 1998 2010 2.00 0.94 3.07 0.00 30.51 Table A 2. Descriptive statistics for chlorophyll a ( g L 1 ) at station 2. Time Period Mean Median Standard deviation Minimum Maximum 1998 1.82 0.91 2.24 0.11 7.07 1999 3.49 3.49 2.56 0.37 7.62 2000 2.20 1.92 1.49 0.37 5.09 2001 1.80 1.21 1.62 0.30 6.25 2002 3.96 2.52 2.62 1.09 9.76 2003 0.81 0.44 1.1 8 0.23 4.48 2004 1.75 0.64 2.12 0.21 6.56 2005 0.60 0.56 0.29 0.18 1.17 2006 5.86 2.51 6.99 0.21 21.72 2007 3.72 3.32 2.66 0.67 9.87 2008 3.49 3.07 3.08 0.22 8.94 2009 1.58 1.56 1.09 0.34 3.24 2010 3.90 1.75 4.13 0.45 11.59 1998 2010 2.69 1.48 3.19 0.11 21.72

PAGE 126

126 Table A 3. Descriptive statistics for chlorophyll a ( g L 1 ) at station 3. Time period Mean Median Standard deviation Minimum Maximum 1998 22.99 18.10 18.07 3.78 63.98 1999 11.75 10.62 7.68 0.30 29.57 2000 13.65 8.78 16.58 1.71 63.53 2001 7.48 5.89 7.25 1.55 27.37 2002 10.02 7.82 7.26 1.71 21.28 2003 14.07 12.48 9.49 1.79 34.78 2004 10.66 4.84 12.17 1.74 41.47 2005 10.39 8.93 5.46 3.49 20.81 2006 15.02 13.08 13.61 1.24 52.82 2007 8.91 5.81 6.81 1.01 22.01 2008 11.33 8.16 8.88 1.34 29.72 2009 9.43 7.37 6.33 2.23 21.67 2010 12.08 7.82 9.35 1.45 2 6.81 1998 2010 12.15 9.94 10.91 0.30 63.98 Table A 4. Descriptive statistics for chlorophyll 1 ) at station 4. Time period Mean Median Standard deviation Minimum Maximum 1998 1.82 0.91 2.24 0.11 7.07 1999 3.49 3.49 2.56 0.37 7.62 2000 2.20 1.92 1.49 0.37 5.09 2001 1.80 1.21 1.62 0.30 6 .25 2002 3.96 2.52 2.62 1.09 9.76 2003 0.81 0.44 1.18 0.23 4.48 2004 1.75 0.64 2.12 0.21 6.56 2005 0.60 0.56 0.29 0.18 1.17 2006 5.86 2.51 6.99 0.21 21.72 2007 3.72 3.32 2.66 0.67 9.87 2008 3.49 3.07 3.08 0.22 8.94 2009 1.58 1.56 1.09 0.34 3.24 20 10 3.90 1.75 4.13 0.45 11.59 1998 2010 2.69 1.48 3.19 0.11 21.72

PAGE 127

127 T able A 5. Descriptive statistics for chlorophyll 1 ) at station 5. Time period Mean Median Standard deviation Minimum Maximum 1998 11.47 4.96 18.34 0.72 67.40 1999 9.39 9.69 6.95 0.51 22.81 2000 8.19 5.90 7.52 2.69 29.26 2001 8.54 7.59 5.70 2. 36 21.48 2002 10.67 11.04 7.06 2.44 23.84 2003 11.72 7.99 11.23 1.55 41.03 2004 6.93 5.36 6.14 0.52 21.82 2005 10.08 6.23 10.83 1.49 33.15 2006 13.58 11.51 10.62 1.06 38.97 2007 11.89 9.18 9.80 3.13 34.45 2008 12.36 8.72 11.77 2.46 43.13 2009 9.04 7.49 6.43 0.67 21.90 2010 10.44 7.37 10.09 0.89 31.28 1998 2010 10.33 8.02 9.76 0.51 67.40 Table A 6. Descriptive statistics for chlorophyll a 1 ) at station 6. Time period Mean Median Standard deviation Minimum Maximum 1998 8.85 8.23 6.61 3.27 26.85 1999 9.13 8.78 6.15 0.51 21.48 2000 11.31 7.44 12.57 2.11 47.28 2001 8.56 8.36 4.83 2.11 18.25 2002 6.04 6.48 3.90 0.30 15.26 2003 13.5 8 8.32 12.74 2.28 46.07 2004 8.06 5.21 8.20 0.46 30.20 2005 13.54 9.41 10.98 1.74 35.81 2006 12.49 7.41 12.67 2.60 47.11 2007 11.66 8.40 7.62 4.13 27.48 2008 13.03 13.52 4.64 5.03 20.45 2009 10.53 8.35 7.91 2.23 23.69 2010 9.92 8.04 7.32 3.24 25.47 1998 2010 10.52 8.29 8.67 0.30 47.28

PAGE 128

128 Table A 7. Descriptive statistics for chlorophyll 1 ) at station 7. Time period Mean Median Standard deviation Minimum Maximum 1998 10.59 6.48 9.73 2.03 37.14 1999 5.25 5.36 3.23 0.37 12.82 2000 3.11 2.57 2.01 1.02 8.17 2001 4.53 2.49 5.14 1.5 5 17.85 2002 3.71 3.36 2.55 0.94 9.38 2003 8.98 5.32 7.29 1.24 24.26 2004 3.41 2.64 2.64 0.93 9.61 2005 9.82 9.34 8.12 1.93 32.32 2006 7.72 6.15 7.69 1.24 29.06 2007 3.40 2.63 2.13 1.12 7.49 2008 8.33 4.47 7.56 1.56 23.91 2009 7.18 5.48 5.24 2.23 2 1.45 2010 7.22 5.31 4.93 1.68 16.87 1998 2010 6.39 4.12 6.13 0.37 37.14 Table A 8. Descriptive statistics for chlorophyll 1 ) at station 8. Time period Mean Median Standard deviation Minimum Maximum 1998 9.18 6.71 5.92 2.52 21.69 1999 5.22 4.93 3.76 0.37 13.40 2000 3.89 2.61 2.86 0.87 10.04 2001 3.96 2.90 3.08 0.8 7 11.38 2002 3.36 2.94 1.91 0.79 6.52 2003 8.06 8.09 5.27 0.94 16.65 2004 3.91 3.04 4.26 0.76 16.40 2005 9.05 8.17 5.50 2.07 19.36 2006 7.28 6.51 4.72 1.45 18.14 2007 4.81 3.07 4.64 1.56 18.16 2008 7.14 4.80 4.95 1.90 14.52 2009 7.33 4.81 5.04 2.57 16.42 2010 6.93 4.97 5.36 0.67 17.54 1998 2010 6.16 4.68 4.80 0.37 21.69

PAGE 129

129 Table A 9. Descriptive statistics for chlorophyll 1 ) at station 9. Time period Mean Median Standard deviation Minimum Maximum 1998 10.19 7.39 7.27 2.60 22.30 1999 3.45 3.15 2.02 0.44 7.71 2000 4.01 2.77 3.22 1.24 13.11 2001 3.90 3.61 2.75 0.94 10.52 2002 1.95 1.55 1.30 0.44 4.83 2003 8.36 6.44 6.79 1.02 20.06 2004 4.54 3.16 4.82 0.72 18.23 2005 5.70 4.92 2.40 1.96 10.54 2006 7.55 4.85 9.33 1.52 36.49 2007 4.10 3.74 3.53 1.01 14.41 2008 6.43 4.13 5.20 2.12 20.22 2009 6.19 4.25 4.77 1.56 15.42 2010 6.18 5.92 4.59 1.34 16.01 1998 2010 5.58 4.13 5.25 0.44 36.49 Table A 10. Descriptive statistics for chlorophyll 1 ) at station 10. Time period Mean Median Standard deviation Minimum Maximum 1998 7.94 5.36 6.85 2.52 23.53 1999 3.61 2.32 3.16 1.17 12.63 2000 3.49 2.44 2.63 1.32 10.90 2001 3.23 3.28 1.99 0 .79 6.98 2002 2.12 1.40 1.89 0.37 5.98 2003 4.96 4.18 3.91 0.87 12.63 2004 5.71 2.76 8.79 0.46 32.71 2005 6.11 5.47 4.50 1.13 13.91 2006 4.76 4.12 3.67 1.49 14.76 2007 6.31 3.08 11.34 1.01 42.01 2008 5.79 4.47 5.55 1.12 21.00 2009 4.62 3.13 3.20 1. 45 11.17 2010 4.33 3.63 2.40 1.34 8.49 1998 2010 4.84 3.24 5.37 0.37 42.01

PAGE 130

130 Total Nitrogen Table A 1) at station 1. Time period Mean Median Standard deviation Minimum Maximum 1998 1090.83 1155 250.03 670 1400 1999 998.33 1045 135.84 690 1170 2000 986.67 935 187.73 740 1340 2001 917.50 870 185.63 690 1260 2002 804.17 765 271.58 490 1490 2003 1248.33 1310 152.90 1010 1460 2004 1102.50 1125 195.13 840 1380 2005 1167.50 1145 201.18 880 1500 2006 1020.42 985 160.77 835 1340 2007 991.67 1030 262.15 580 1420 2008 1122.50 1120 171.10 860 14 50 2009 1291.67 1280 160.39 1040 1540 2010 1253.33 1275 141.00 980 1440 1998 2010 1076.57 1075 232.29 490 1540 Table A 1 ) at station 2. Time period Mean Median Standard deviation Minimum Maximum 1998 991.92 947 211.12 700 1320 1999 850.00 860 188.05 460 1070 2000 862.50 855 205.08 570 1270 2001 841.67 895 26 9.51 480 1310 2002 582.00 530 208.84 280 1050 2003 1226.75 1225 133.95 1040 1510 2004 1101.67 1065 176.63 860 1390 2005 1145.00 1085 196.49 840 1430 2006 916.67 905 204.15 590 1280 2007 835.83 865 257.03 490 1210 2008 1004.17 1020 290.31 430 1570 2 009 1221.67 1190 148.13 960 1470 2010 1166.67 1230 191.42 870 1440 1998 2010 980.50 1000 272.40 280 1570

PAGE 131

131 Table A 1 ) at station 3. Time period Mean Median Standard deviation Minimum Maximum 1998 798.50 795 281.34 400 400 1999 511.67 425 227.51 250 250 2000 537.17 520 181.48 260 260 2001 533.33 390 282.6 9 240 240 2002 521.67 565 167.49 230 230 2003 790.00 730 285.21 460 460 2004 639.17 590 250.43 320 320 2005 711.67 650 247.71 400 400 2006 580.00 560 169.65 340 340 2007 459.17 435 120.56 280 280 2008 574.17 575 253.64 270 270 2009 710.58 695 172.0 7 507 507 2010 693.33 665 180.72 400 400 1998 2010 620.03 580 239.63 230 1480 Table A 14. Descriptive statistics for 1 ) at station 4. Time period Mean Median Standard deviation Minimum Maximum 1998 877.25 810 155.81 730 1180 1999 728.33 685 206.83 440 1050 2000 667.50 650 226.12 370 1030 2001 691.67 640 257.57 340 1130 2002 496.67 510 148.16 2 70 710 2003 1023.33 1050 254.78 530 1390 2004 807.50 770 259.97 490 1360 2005 989.17 1000 101.93 810 1170 2006 817.50 870 301.85 310 1300 2007 627.50 665 144.61 370 860 2008 726.67 580 338.94 250 1490 2009 1060.00 1110 234.25 590 1300 2010 987.50 1 065 265.44 440 1260 1998 2010 807.74 775 278.06 250 1490

PAGE 132

132 Table A 1 ) at station 5. Time period Mean Median Standard deviation Minimum Maximum 1998 598.75 558 150.24 400 830 1999 425.83 410 162.45 170 660 2000 586.67 525 245.30 340 1240 2001 542.08 470 222. 73 300 1160 2002 564.17 540 187.44 270 970 2003 894.17 815 325.28 370 1350 2004 660.00 555 327.39 240 1360 2005 795.83 770 266.00 430 1220 2006 638.33 570 206.35 410 1040 2007 521.67 475 215.65 260 1030 2008 555.83 585 167.63 260 790 2009 737.00 72 5 214.07 400 1110 2010 721.67 735 185.02 460 980 1998 2010 659.38 570 391.60 170 4400 Table A 1 ) at station 6. Time period Mean Median Standard deviation Minimum Maximum 1998 520.50 483 209.13 190 960 1999 388.33 390 123.86 210 560 2000 494.17 490 140.03 240 720 2001 544.17 545 271.7 4 240 1130 2002 441.67 355 306.62 230 1380 2003 765.83 760 240.51 410 1140 2004 590.83 460 331.26 270 1270 2005 648.33 740 208.49 220 860 2006 484.17 485 188.27 250 770 2007 400.00 395 103.92 250 630 2008 476.67 495 162.22 240 700 2009 718.67 705 1 70.92 444 1060 2010 673.33 695 220.67 370 1090 1998 2010 549.74 510 238.23 190 1380

PAGE 133

133 Table A 1 ) at station 7. Time period Mean Median Standard deviation Minimum Maximum 1998 385.25 377 140.56 170 730 1999 327.50 335 76.29 170 430 2000 308.33 305 66.99 220 430 2001 326.67 280 136.80 160 580 2002 351.67 330 93.21 220 510 2003 464.17 440 149.88 260 750 2004 420.83 335 203.71 230 900 2005 422.50 385 141.81 260 690 2006 349.17 340 142.60 210 740 2007 293.75 270 74.93 205 440 2008 375.83 330 167.57 180 630 2009 592.50 435 414.23 35 0 1860 2010 464.17 465 141.39 260 810 1998 2010 390.95 360 183.75 160 1860 Table A 1 ) at station 8. Time period Mean Median Standard deviation Minimum Maximum 1998 349.58 343 96.64 210 590 1999 294.17 295 70.64 170 410 2000 315.83 285 90.50 240 520 2001 296.92 290 60.18 19 3 440 2002 316.67 310 80.26 220 470 2003 496.67 460 137.20 340 770 2004 423.33 345 226.97 250 1030 2005 431.67 430 120.29 240 650 2006 366.67 350 142.53 220 650 2007 269.00 240 110.13 50 420 2008 350.83 315 126.02 170 580 2009 533.33 480 164.83 330 880 2010 492.50 435 163.66 250 760 1998 2010 379.78 350 150.45 50 1030

PAGE 134

134 Table A 1 ) at station 9. Time period Mean Median Standard deviation Minimum Maximum 1998 387.08 338 110.05 260 640 1999 279.17 265 86.18 80 410 2000 301.67 275 78.14 220 460 2001 310.83 320 75.97 17 0 470 2002 310.83 305 90.90 200 530 2003 419.17 465 125.44 210 560 2004 401.67 340 212.81 160 960 2005 453.33 470 141.38 230 710 2006 325.00 270 116.27 220 560 2007 286.67 270 80.04 190 430 2008 363.33 355 120.70 170 620 2009 495.83 425 179.52 280 880 2010 463.33 455 167.08 260 760 1998 2010 369.07 325 142.29 80 960 Table A 20. 1 ) at station 10. Time period Mean Median Standard deviation Minimum Maximum 1998 357.25 319 121.58 210 560 1999 275.83 285 88.67 140 410 2000 288.33 285 39.04 220 360 2001 300.00 335 70.45 150 370 200 2 296.67 280 65.55 240 430 2003 374.17 400 110.24 200 570 2004 420.83 345 251.95 140 1110 2005 410.83 390 129.79 250 640 2006 302.50 265 109.55 160 470 2007 287.50 265 65.10 200 390 2008 331.17 310 150.11 160 730 2009 461.67 445 102.76 290 720 2010 390.00 315 171.25 230 790 1998 2010 345.90 320 133.99 140 1110

PAGE 135

135 Total Phosphorus Table A 1 ) at station 1. Time period Mean Median Standard deviation Minimum Maximum 1998 91.83 91.0 27.27 51 141 1999 81.08 77.5 22.83 60 148 2000 85.42 70.5 37.56 61 190 2001 91.75 87.0 24.07 66 133 2002 102.42 97.0 26.17 70 140 2003 110.42 101.5 31.09 78 172 2004 118.00 102.0 47.88 64 219 2005 117.33 116.5 22.99 87 174 2006 83.58 80.0 15.15 70 126 2007 95.00 101.5 19.80 65 117 2008 113.58 102.0 47.26 68 251 2009 118.33 114.5 26.26 82 175 2010 100.33 101.0 14.00 74 122 1998 2010 100.70 95.0 31.44 51 251 Table A 1 ) at station 2. Time period Mean Median Standard deviation Minimum Maximum 1998 93.08 90.5 28.97 55 147 1999 76.50 74.5 8.16 61 89 2000 81.83 72.5 19.79 64 119 2001 89.42 87.5 22.03 61 13 7 2002 90.58 86.0 29.12 46 135 2003 109.17 106.0 25.93 75 165 2004 109.92 100.0 42.26 58 211 2005 119.75 115.0 22.59 97 178 2006 83.08 82.0 26.50 23 142 2007 87.75 79.0 23.12 54 128 2008 103.17 97.0 40.04 46 202 2009 115.08 107.0 22.17 91 154 2010 93.17 98.5 13.26 65 110 1998 2010 96.35 93.0 28.64 23 211

PAGE 136

136 Table A 23. Descriptive 1 ) at station 3. Time period Mean Median Standard deviation Minimum Maximum 1998 84.58 75.5 30.14 47 155 1999 55.67 48.0 30.94 9 119 2000 43.50 42.0 16.46 21 79 2001 47.83 38.0 29.38 19 116 2002 53.75 53.5 18.73 28 79 2003 83.50 81.5 39.91 22 148 2004 69.83 67.5 35.23 23 138 2005 81.50 72.5 43.21 36 178 2006 61.83 55.0 19.60 34 96 2007 44.67 41.0 20.21 14 73 2008 48.25 51.0 22.40 17 84 2009 66.58 68.0 33.33 25 142 2010 66.42 65.5 29.10 32 132 1998 2010 6 2.15 55.5 31.67 9 178 Table A 1 ) at station 4. Time period Mean Median Standard deviation Minimum Maximum 1998 91.67 86.5 26.64 64 148 1999 70.17 68.5 16.35 28 96 2000 65.42 65.0 23.40 31 116 2001 70.17 72.5 28.22 31 1 33 2002 68.08 60.5 26.96 28 112 2003 99.25 100.5 33.92 43 159 2004 92.25 79.0 35.14 54 163 2005 111.67 105.0 27.87 90 189 2006 80.08 80.5 25.42 38 135 2007 66.00 66.0 19.01 26 95 2008 78.58 69.5 56.07 19 225 2009 96.50 104.0 24.63 44 128 2010 86.6 7 91.0 23.90 37 122 1998 2010 82.81 79.0 31.97 19 225

PAGE 137

137 Table A 25 1 ) at station 5 Time period Mean Median Standard deviation Minimum Maximum 1998 74.17 67.0 34.05 38 146 1999 45.08 44.0 14.00 28 68 2000 55.17 53.5 24.47 25 117 2001 56.00 53.0 23.24 29 116 2002 7 2.58 74.0 17.32 39 97 2003 94.83 73.0 52.04 23 207 2004 75.58 65.5 45.84 20 165 2005 91.58 83.5 41.85 38 199 2006 68.50 65.5 24.56 19 107 2007 55.17 48.5 26.79 20 92 2008 57.75 61.0 22.63 21 89 2009 70.17 65.0 33.17 20 139 2010 66.50 72.0 26.61 18 114 1998 2010 67.93 63.0 33.41 18 207 Table A 26 1 ) at station 6 Time period Mean Median Standard deviation Minimum Maximum 1998 64.83 54.5 40.22 19 144 1999 36.50 37.5 6.71 26 51 2000 43.58 39.0 14.22 22 66 2001 49.83 43.0 30.04 23 116 2002 42. 50 40.0 20.89 18 102 2003 74.25 72.5 30.64 24 120 2004 67.08 48.0 47.10 24 156 2005 70.08 70.5 31.82 22 129 2006 51.67 57.0 20.38 14 76 2007 44.67 45.0 19.35 19 75 2008 46.58 43.0 21.30 20 88 2009 62.58 60.5 28.06 23 113 2010 57.00 44.5 27.44 27 10 1 1998 2010 54.71 46.0 29.26 14 156

PAGE 138

138 Table A 27 1 ) at station 7 Time period Mean Median Standard deviation Minimum Maximum 1998 37.25 37.0 14.86 11 62 1999 25.50 22.5 11.49 9 45 2000 18.83 17.0 6.60 10 30 2001 20.83 19.0 8.36 13 37 2002 28.75 2 9.5 11.05 16 51 2003 37.25 41.0 15.76 14 60 2004 42.25 34.5 29.73 16 120 2005 39.50 34.0 19.46 13 74 2006 33.17 26.5 18.96 15 76 2007 19.08 16.0 9.14 10 43 2008 26.00 21.5 14.04 11 48 2009 43.83 29.0 48.72 19 194 2010 31.67 25.0 16.62 13 71 1998 2 010 31.07 26.0 21.26 9 194 Table A 28 1 ) at station 8 Time period Mean Median Standard deviation Minimum Maximum 1998 33.42 33.5 10.94 14 52 1999 22.42 20.0 8.62 13 44 2000 21.75 18.5 8.71 14 40 2001 19.33 19.5 5.21 10 30 2002 25.58 24. 5 8.71 14 39 2003 43.92 37.0 22.90 13 97 2004 42.25 32.5 30.91 16 124 2005 40.33 34.5 20.15 15 70 2006 34.08 32.0 15.20 16 63 2007 22.25 21.0 9.85 11 39 2008 22.58 17.5 11.75 9 47 2009 39.58 31.5 22.52 22 95 2010 35.75 34.5 15.04 18 68 1998 2010 3 1.02 26.0 17.89 9 124

PAGE 139

139 Table A 2 9 1 ) at station 9 Time period Mean Median Standard deviation Minimum Maximum 1998 37.00 35.0 12.81 23 70 1999 19.92 21.5 7.29 6 30 2000 20.92 19.5 7.49 12 36 2001 20.58 20.0 3.06 16 26 2002 19.08 17 .5 7.59 10 36 2003 34.00 36.5 14.24 12 62 2004 38.25 31.5 27.54 14 113 2005 42.25 36.5 20.71 15 80 2006 30.33 27.0 14.98 13 62 2007 20.42 17.5 9.49 10 38 2008 23.33 19.5 11.87 9 47 2009 34.83 29.0 22.74 13 97 2010 31.50 29.0 14.72 13 63 1998 2010 28.65 24.5 16.44 6 113 Table A 1 ) at station 10. Time period Mean Median Standard deviation Minimum Maximum 1998 35.33 33.0 15.93 16 71 1999 18.58 18.0 6.56 8 32 2000 19.00 18.5 4.79 12 26 2001 19.00 19.0 6.32 10 32 20 02 18.58 15.5 7.24 10 34 2003 25.50 29.5 9.93 9 37 2004 38.67 25.5 34.22 9 136 2005 39.17 34.0 19.41 15 78 2006 23.75 20.0 9.46 14 45 2007 18.67 16.5 7.27 11 33 2008 20.92 17.5 12.44 9 56 2009 30.17 28.0 11.00 16 52 2010 25.92 22.5 18.11 13 79 199 8 2010 25.63 22.0 15.99 8 136

PAGE 140

140 Water Color Table A 31. Descriptive statistics for color (PCU) at station 1. Time period Mean Median Standard deviation Minimum Maximum 1998 127.67 82.5 110.00 15 340 1999 36.58 31.0 27.13 12 110 2000 53.17 35.0 48.00 1 4 189 2001 92.67 73.5 77.68 11 243 2002 48.17 24.5 52.98 13 195 2003 163.75 130.5 93.65 60 307 2004 175.84 121.0 177.76 17 540 2005 158.07 157.5 89.62 42 312 2006 57.25 20.0 69.98 10 228 2007 57.33 56.5 49.77 7 167 2008 145.50 122.0 104.14 19 306 2009 141.67 115.5 109.53 30 357 2010 100.92 105.5 72.00 11 246 1998 2010 104.51 73.5 99.92 7 540 Table A 32. Descriptive statistics for color (PCU) at station 2. Time period Mean Median Standard deviation Minimum Maximum 1998 217.42 167.5 149.13 79 522 1999 32.50 30.5 18.42 18 86 2000 49.92 36.0 39.94 16 153 2001 85.17 58.0 77.30 13 231 2002 42.83 23.5 45.77 12 173 2003 164.00 131.0 96.58 60 321 2004 171.85 117.5 177.41 18 532 2005 155.66 150.5 88.27 42 318 2006 66.08 23.5 85.92 11 242 2007 51.92 48.5 40.31 13 133 2008 125.50 96.0 94.76 22 312 2009 127.33 102.0 96.54 31 339 2010 107.75 102.0 95.63 10 354 1998 2010 107.53 72.0 107.18 10 532

PAGE 141

141 Table A 33. Descriptive statistics for color (PCU) at station 3. Time period Mean Median Standard deviation Minimum Maximum 1998 102.17 61.0 77.30 40 267 1999 29.25 27.5 13.48 10 54 2000 20.25 15.0 11.98 11 54 2001 34.17 16.5 38.33 9 140 2002 27.83 15.0 42.65 7 162 2003 75.92 52.0 62.98 18 195 2004 82.01 35.0 116.41 17 432 2005 87.85 52.5 76. 38 16 234 2006 42.33 29.0 40.37 10 140 2007 15.42 14.5 6.73 8 31 2008 47.83 26.5 47.53 13 160 2009 67.33 39.0 91.27 23 351 2010 53.42 39.0 47.40 9 166 1998 2010 52.75 26.0 63.98 7 432 Table A 34. Descriptive statistics for color (PCU) at station 4. Time period Mean Median Standard deviation Minimum Maximum 1998 161.58 97.5 122.06 63 422 1999 38.33 31.5 21.97 12 86 2000 33.75 28.5 19.85 13 77 2001 60.67 29.5 64.03 11 216 2002 25.08 18.5 23.72 7 98 2003 144.58 114.5 106.30 27 318 2004 134.95 78.6 148.02 20 512 2005 144.89 118.5 92.63 37 314 2006 64.42 26.0 80.65 10 242 2007 31.08 17.0 25.97 14 96 2008 88.67 41.0 94.12 19 312 2009 109.50 100.5 89.28 30 348 2010 90.00 87.0 69.52 12 240 1998 2010 86.73 46.5 93.07 7 512

PAGE 142

142 Table A 35. Descri ptive statistics for color (PCU) at station 5. Time period Mean Median Standard deviation Minimum Maximum 1998 113.50 68.0 86.05 45 297 1999 26.67 21.0 14.76 10 60 2000 25.33 20.0 13.12 11 54 2001 38.67 21.0 36.06 8 113 2002 24.25 16.0 25.52 7 95 2 003 104.08 72.0 99.78 16 276 2004 121.41 36.0 167.52 15 504 2005 111.33 84.0 91.38 20 308 2006 47.17 23.0 58.23 10 206 2007 17.33 17.0 7.44 7 31 2008 46.92 32.0 36.89 15 139 2009 81.25 44.5 110.44 18 378 2010 72.25 44.0 69.72 10 236 1998 2010 63.86 31.0 83.15 7 504 Table A 36. Descriptive statistics for color (PCU) at station 6. Time period Mean Median Standard deviation Minimum Maximum 1998 94.33 56.5 71.74 37 247 1999 23.75 20.0 11.45 12 50 2000 20.50 19.0 7.69 7 34 2001 27.50 18.0 27.72 6 99 2002 20.25 11.5 22.21 5 81 2003 89.75 60.0 87.08 18 277 2004 101.85 26.9 155.65 14 504 2005 66.37 56.5 44.52 14 162 2006 36.50 24.0 42.88 8 158 2007 16.67 15.0 8.22 6 33 2008 39.08 29.5 25.95 14 100 2009 69.08 32.5 89.52 15 282 2010 53.33 35.0 53.83 8 178 1998 2010 50.69 26.0 68.51 5 504

PAGE 143

143 Table A 37. Descriptive statistics for color (PCU) at station 7. Time period Mean Median Standard deviation Minimum Maximum 1998 54.33 32.5 41.50 21 143 1999 15.75 17.0 6.68 5 29 2000 9.58 9.0 3.42 5 15 2001 12.50 8.5 10.08 4 32 2002 10.17 8.5 6.51 4 29 2003 42.33 32.5 35.09 9 109 2004 64.45 17.0 109.42 10 392 2005 42.38 29.5 36.69 6 115 2006 26.17 14.0 38.46 5 144 2007 9.67 9.0 4.81 4 19 2008 23.92 18.0 18.43 4 63 2009 33.58 18.5 49.26 13 189 2010 28.25 25.0 27.99 4 107 1998 2010 28.70 15.5 42.74 4 392 Table A 38. Descriptive statistics for color (PCU) at station 8. Time period Mean Median Standard deviation Minimum Maximum 1998 52.08 31.0 39.88 20 137 1999 15.50 15.0 7.28 6 28 2000 9.8 3 9.0 4.13 5 18 2001 11.58 10.5 8.82 3 35 2002 8.83 8.0 5.17 3 23 2003 49.08 32.0 50.26 7 186 2004 61.77 17.7 107.50 7 380 2005 35.83 21.5 34.12 9 119 2006 26.67 13.0 37.53 4 139 2007 8.33 8.0 3.50 2 14 2008 20.92 15.0 15.47 7 63 2009 41.00 18.5 6 4.75 10 243 2010 31.08 22.0 29.38 5 105 1998 2010 28.66 15.0 44.61 2 380

PAGE 144

144 Table A 39. Descriptive statistics for color (PCU) at station 9. Time period Mean Median Standard deviation Minimum Maximum 1998 49.00 29.5 37.11 19 128 1999 13.67 13.5 5.87 7 26 2000 9.83 9.5 3.64 5 16 2001 11.83 11.0 7.99 4 33 2002 7.67 6.0 6.11 2 24 2003 29.25 29.5 21.68 7 85 2004 52.94 21.0 80.19 9 276 2005 39.15 19.5 43.33 8 156 2006 25.83 12.0 35.89 5 134 2007 7.33 6.5 3.20 3 14 2008 20.17 18.0 12.88 9 57 2009 3 9.25 18.0 65.70 11 243 2010 28.25 17.0 27.86 4 95 1998 2010 25.71 15.0 37.75 2 276 Table A 40. Descriptive statistics for color (PCU) at station 10. Time period Mean Median Standard deviation Minimum Maximum 1998 42.25 25.5 32.13 16 111 1999 13.75 13.5 5.05 7 22 2000 8.50 7.0 4.10 4 19 2001 8.17 7.0 3.21 4 14 2002 6.17 5.0 4.88 3 19 2003 22.67 18.5 21.77 7 88 2004 64.15 19.1 122.19 6 436 2005 33.08 17.0 37.33 7 140 2006 16.75 12.0 15.57 4 60 2007 7.00 6.5 3.30 3 15 2008 18.58 14.0 17.04 8 7 1 2009 29.00 17.5 35.17 4 129 2010 19.08 14.0 15.05 2 47 1998 2010 22.24 13.0 40.87 2 436

PAGE 145

145 Water Temperature Table A 41. Descriptive statistics for water temperature (C) at station 1. Time period Mean Median Standard deviation Minimum Maximum 1998 21.95 21.90 4.95 15.00 28.30 1999 23.31 23.55 5.13 15.70 30.50 2000 23.43 22.95 4.78 15.80 29.20 2001 22.89 23.40 5.26 14.12 29.26 2002 23.57 25.22 4.68 16.21 28.85 2003 21.51 22.13 5.08 11.17 27.45 2004 22.67 23.81 6.01 14.09 30.16 2005 21.59 20.97 3.76 16.58 27.92 2006 23.40 24.65 4.85 15.06 30.15 2007 23.36 23.44 4.60 16.05 29.16 2008 22.14 20.63 5.23 15.25 29.51 2009 22.24 23.04 4.87 13.87 28.07 2010 22.14 22.05 5.98 11.92 30.00 1998 2010 22.63 22.70 4.90 11.17 30.50 Table A 42. Descripti ve statistics for water temperature (C) at station 2. Time period Mean Median Standard deviation Minimum Maximum 1998 22.40 22.75 5.52 15.00 30.53 1999 22.44 20.85 5.42 15.10 29.80 2000 23.01 22.15 5.11 15.90 29.50 2001 22.37 23.07 5.71 12.52 29.10 2002 23.26 24.80 4.72 15.57 29.05 2003 21.39 21.86 4.89 11.39 27.40 2004 22.71 23.88 5.57 14.39 29.14 2005 21.43 20.66 4.12 15.75 27.78 2006 23.17 24.65 5.74 12.86 31.53 2007 23.85 23.99 5.28 15.96 32.31 2008 22.43 21.01 5.65 13.79 30.49 2009 22.40 22.72 5.18 13.86 29.74 2010 22.04 22.37 6.10 12.08 29.94 1998 2010 22.53 22.20 5.17 11.39 32.31

PAGE 146

146 Table A 43. Descriptive statistics for water temperature (C) at station 3. Time period Mean Median Standard deviation Minimum Maximum 1998 23.25 23.35 6.61 15.20 31.40 1999 22.08 21.35 7.02 12.30 31.80 2000 22.68 22.45 6.51 11.90 30.50 2001 22.50 22.97 6.45 11.63 29.82 2002 23.01 25.22 6.03 13.85 30.43 2003 21.81 22.48 7.08 9.85 29.95 2004 23.01 25.58 6.92 11.82 30.20 2005 21.85 21.08 5.98 12.86 2 9.87 2006 23.27 24.89 5.79 15.07 31.71 2007 23.24 23.08 5.74 14.27 31.85 2008 21.56 20.46 6.84 11.80 30.56 2009 23.02 24.94 6.76 10.25 30.85 2010 22.09 21.48 6.79 13.51 31.20 1998 2010 22.57 23.44 6.29 9.85 31.85 Table A 44. Descriptive statistics for water temperature (C) at station 4. Time period Mean Median Standard deviation Minimum Maximum 1998 22.32 21.85 6.15 15.10 30.50 1999 21.63 20.70 6.64 10.60 30.20 2000 22.32 21.60 6.72 9.35 31.60 2001 21.97 23.40 5.90 12.38 28.57 2002 22.68 24. 87 5.03 14.60 29.23 2003 20.83 22.08 6.11 8.72 27.89 2004 23.06 23.94 5.50 14.62 29.82 2005 21.34 21.00 4.85 13.90 28.20 2006 22.50 24.09 5.63 12.92 31.09 2007 22.81 22.96 5.66 14.32 31.54 2008 21.83 21.79 6.61 11.79 30.07 2009 22.45 22.49 6.14 10.7 5 30.20 2010 21.77 21.77 6.71 12.02 31.10 1998 2010 22.12 22.35 5.80 8.72 31.60

PAGE 147

147 Table A 45. Descriptive statistics for water temperature (C) at station 5. Time period Mean Median Standard deviation Minimum Maximum 1998 23.42 24.20 6.26 15.00 30.84 1999 22.96 22.40 6.90 13.10 32.40 2000 23.34 23.90 6.21 12.87 31.00 2001 22.73 24.16 6.33 11.55 29.99 2002 23.73 25.51 5.93 14.81 30.87 2003 21.91 22.74 6.97 10.67 30.00 2004 23.09 24.61 6.56 12.91 30.61 2005 22.11 21.14 5.97 13.20 30.97 2006 23.46 24.39 5.83 14.20 32.07 2007 23.36 23.40 5.64 14.66 31.50 2008 21.89 20.52 6.48 12.12 29.72 2009 22.96 23.62 6.28 11.29 30.79 2010 22.14 22.20 6.64 13.09 31.17 1998 2010 22.85 23.89 6.10 10.67 32.40 Table A 46. Descriptive statistics for water tempe rature (C) at station 6. Time period Mean Median Standard deviation Minimum Maximum 1998 23.27 24.10 6.58 14.50 30.65 1999 23.15 22.70 6.57 13.30 31.80 2000 23.13 23.40 6.64 11.11 30.60 2001 22.81 23.62 6.29 12.21 30.50 2002 23.45 25.14 5.72 14.40 30.72 2003 22.35 22.85 7.19 11.26 30.56 2004 23.06 24.83 7.06 11.10 30.80 2005 22.57 21.96 6.36 13.18 31.20 2006 23.37 24.01 6.04 14.17 32.49 2007 23.48 23.31 5.58 14.46 31.66 2008 21.59 20.46 6.57 11.94 30.23 2009 23.12 25.64 6.65 10.61 30.38 2010 22.23 22.14 6.81 12.32 31.29 1998 2010 22.89 23.77 6.25 10.61 32.49

PAGE 148

148 Table A 47. Descriptive statistics for water temperature (C) at station 7. Time period Mean Median Standard deviation Minimum Maximum 1998 23.41 23.30 6.96 13.70 32.50 1999 22.18 20.90 6.87 12.00 30.80 2000 22.66 22.30 6.21 13.80 30.80 2001 22.29 22.19 6.32 11.94 30.19 2002 22.52 24.66 5.84 14.30 29.26 2003 21.85 23.24 6.87 10.33 29.79 2004 22.76 24.76 7.03 11.57 30.53 2005 21.95 20.97 6.03 12.00 31.20 2006 23.05 24.95 6.09 13.91 31.51 2007 22.99 22.95 5.84 13.86 31.03 2008 22.07 20.41 6.54 12.34 30.41 2009 22.58 24.57 7.08 10.29 30.39 2010 22.42 22.21 6.99 11.87 32.48 1998 2010 22.52 23.11 6.29 10.29 32.50 Table A 48. Descriptive statistics for water temperature (C) at station 8. Time period Mean Median Standard deviation Minimum Maximum 1998 23.64 23.75 6.67 14.50 31.57 1999 22.38 21.50 6.92 12.10 31.30 2000 22.79 22.65 6.07 14.10 30.10 2001 22.50 22.25 6.50 12.75 31.41 2002 22.55 24.48 5.67 14.27 29.62 2003 21.96 23.43 6.73 10.72 29.95 2004 23.00 25.27 7.19 11.68 31.18 2005 21.91 21.01 6.13 12.50 30.98 2006 23.23 24.70 5.94 14.10 32.07 2007 23.17 23.29 5.77 13.91 30.96 2008 22.02 20.50 6.58 11.62 30.50 2009 22.84 25.15 6.96 10.51 30.48 2010 22.36 22.21 7.08 11.47 32.25 1998 2010 22.64 23.43 6.26 10.51 32.25

PAGE 149

149 Table A 49. Descriptive statistics for water temperature (C) at station 9. Time period Mean Median Standard deviation Minimum Maximum 1998 23.71 23.70 6.62 14.50 31.54 1999 22.68 21.70 6.71 1 2.50 31.00 2000 22.84 22.53 6.11 14.30 30.50 2001 22.96 23.13 6.61 12.15 31.47 2002 23.23 25.49 5.94 14.38 30.76 2003 22.92 23.76 7.22 11.19 31.92 2004 23.20 25.07 7.09 11.99 31.41 2005 22.17 21.25 6.31 12.79 31.28 2006 23.31 24.77 6.00 13.71 32.11 2007 23.22 23.40 5.62 13.94 30.87 2008 22.18 20.56 6.53 12.58 31.09 2009 22.88 25.45 6.95 10.96 30.53 2010 22.36 22.12 7.05 11.17 31.78 1998 2010 22.90 23.75 6.30 10.96 32.11 Table A 50. Descriptive statistics for water temperature (C) at station 1 0. Time period Mean Median Standard deviation Minimum Maximum 1998 23.48 23.35 6.60 14.50 31.06 1999 22.69 22.00 6.60 13.10 31.10 2000 22.94 22.73 6.09 14.40 30.50 2001 22.88 23.25 6.58 11.40 30.75 2002 24.13 25.66 6.97 14.49 30.47 2003 22.86 23.99 7.10 11.17 30.92 2004 23.19 24.97 7.05 12.62 31.13 2005 22.41 21.44 6.41 12.58 31.09 2006 23.38 23.75 5.70 16.11 31.79 2007 23.24 22.95 5.82 14.06 31.11 2008 22.14 20.30 6.52 12.14 30.65 2009 22.58 23.83 6.75 10.66 30.44 2010 22.43 22.29 7.02 11.73 31.84 1998 2010 22.95 23.61 6.33 10.66 32.50

PAGE 150

150 Salinity Table A 51. Descriptive statistics for salinity (ppt) at station 1. Time period Mean Median Standard deviation Minimum Maximum 1998 0.05 0.00 0.09 0.00 0.20 1999 0.18 0.20 0.06 0.10 0.30 2000 0.18 0.20 0.04 0.10 0.20 2001 0.20 0.12 0.23 0.01 0.89 2002 0.22 0.20 0.11 0.09 0.42 2003 0.17 0.14 0.17 0.03 0.70 2004 0.12 0.12 0.05 0.04 0.17 2005 0.17 0.13 0.17 0.03 0.70 2006 0.23 0.18 0.21 0.00 0.86 2007 0.19 0.18 0.06 0.10 0.36 2008 0.14 0.1 3 0.08 0.03 0.35 2009 0.12 0.13 0.04 0.03 0.18 2010 0.13 0.13 0.05 0.03 0.19 1998 2010 0.16 0.14 0.13 0.00 0.89 Table A 52. Descriptive statistics for salinity (ppt) at station 2. Time period Mean Median Standard deviation Minimum Maximum 1998 1.03 0.05 3.05 0.00 10.70 1999 2.00 0.50 2.79 0.10 8.40 2000 2.45 1.43 2.33 0.30 6.40 2001 3.28 0.37 4.83 0.11 14.57 2002 2.22 2.65 1.78 0.10 4.87 2003 0.13 0.14 0.06 0.03 0.20 2004 0.57 0.16 1.00 0.04 3.22 2005 0.33 0.15 0.36 0.03 1.01 2006 4.58 0.75 7.07 0.05 20.64 2007 5.04 2.98 5.57 0.13 17.14 2008 2.77 0.79 4.23 0.04 14.83 2009 1.57 0.37 2.21 0.03 6.98 2010 2.43 0.16 3.65 0.04 10.14 1998 2010 2.18 0.30 3.74 0.00 20.64

PAGE 151

151 Table A 53. Descriptive statistics for salinity (ppt) at station 3. Time period Mean Median Standard deviation Minimum Maximum 1998 14.22 14.95 8.79 2.00 26.30 1999 22.47 23.35 5.25 14.40 34.30 2000 24.28 24.95 3.96 16.20 28.60 2001 23.46 25.61 6.12 9.46 29.79 2002 23.30 24.36 3.41 13.33 25.50 2003 18.86 19.44 6.26 10.41 28.70 2004 18.51 18.70 5.24 5.50 24.31 2005 16.08 15.88 9.78 0.09 30.14 2006 18.98 17.63 7.70 3.06 29.26 2007 26.16 26.59 2.46 21.66 29.77 2008 21.93 22.73 4.33 14.87 28.10 2009 17.47 19.09 6.23 2.48 25.19 2010 17.08 18.85 8.11 5.42 26.31 1998 201 0 20.21 22.25 7.01 0.09 34.30 Table A 54. Descriptive statistics for salinity (ppt) at station 4. Time period Mean Median Standard deviation Minimum Maximum 1998 4.68 1.25 7.44 0.00 22.90 1999 6.73 5.30 6.81 0.60 26.90 2000 14.72 15.00 6.01 3.40 22. 76 2001 12.85 15.21 7.64 0.15 23.29 2002 17.92 20.56 6.99 2.26 25.24 2003 7.43 3.69 8.22 0.04 20.93 2004 7.55 5.05 6.50 0.05 19.25 2005 5.92 4.75 6.02 0.09 15.70 2006 8.79 4.71 9.00 0.06 21.18 2007 17.38 17.12 8.56 6.58 28.03 2008 13.88 13.28 9.20 0.95 27.87 2009 6.94 5.39 6.27 0.04 18.18 2010 7.09 4.17 7.80 0.10 22.28 1998 2010 10.14 7.68 8.44 0.00 28.03

PAGE 152

152 Table A 55. Descriptive statistics for salinity (ppt) at station 5. Time period Mean Median Standard deviation Minimum Maximum 1998 13.70 16.85 9.55 0.00 25.70 1999 21.77 23.85 6.46 10.30 30.60 2000 20.54 19.10 3.76 13.80 26.80 2001 20.96 22.59 4.67 14.02 28.17 2002 22.75 23.67 3.71 13.52 26.46 2003 14.69 17.07 10.95 0.30 29.61 2004 15.72 17.82 7.60 0.52 25.27 2005 14.48 18.15 11.09 0 .04 26.85 2006 17.42 18.76 5.43 9.02 27.80 2007 23.81 24.56 5.20 15.49 29.77 2008 20.84 20.63 4.11 14.20 28.36 2009 16.89 17.83 5.92 0.21 24.67 2010 16.61 18.14 6.67 2.24 25.19 1998 2010 18.48 19.30 7.49 0.00 30.60 Table A 56. Descriptive statistic s for salinity (ppt) at station 6. Time period Mean Median Standard deviation Minimum Maximum 1998 16.28 19.95 9.03 2.00 26.00 1999 23.89 25.10 5.82 11.70 29.70 2000 23.41 22.45 2.62 20.77 28.40 2001 23.79 25.34 5.36 14.56 30.55 2002 25.57 26.22 4.1 7 16.77 32.24 2003 17.52 18.51 7.67 3.64 27.73 2004 19.26 22.32 7.25 1.21 26.16 2005 19.06 19.93 5.87 4.83 28.45 2006 20.85 21.34 7.00 10.76 29.15 2007 25.35 26.03 3.83 17.91 29.63 2008 20.43 20.42 4.69 11.95 26.96 2009 19.96 20.63 5.22 5.87 25.40 2010 18.59 20.81 7.62 3.35 28.01 1998 2010 21.07 22.40 6.55 1.21 32.24

PAGE 153

153 Table A 57. Descriptive statistics for salinity (ppt) at station 7. Time period Mean Median Standard deviation Minimum Maximum 1998 23.56 24.8 4.55 15.00 32.20 1999 28.05 28.8 5. 87 15.50 34.60 2000 30.43 31.0 1.94 27.40 33.00 2001 29.43 29.9 3.00 23.01 33.60 2002 28.26 29.3 2.71 23.26 31.23 2003 24.12 23.7 4.73 17.30 30.82 2004 24.70 25.2 3.44 17.53 29.64 2005 24.35 24.8 5.09 17.11 32.70 2006 26.74 28.5 4.75 16.43 31.79 20 07 29.87 30.8 3.00 24.86 33.24 2008 26.07 26.2 4.07 17.12 31.81 2009 24.67 25.1 4.29 13.16 30.22 2010 23.63 24.1 6.06 9.12 29.94 1998 2010 26.45 27.16 4.77 9.12 34.60 Table A 58. Descriptive statistics for salinity (ppt) at station 8. Time period Me an Median Standard deviation Minimum Maximum 1998 23.67 24.50 6.85 8.00 31.40 1999 28.18 29.15 5.31 15.40 34.00 2000 30.14 30.44 1.68 26.80 32.40 2001 28.85 29.38 3.49 22.26 33.10 2002 28.51 30.07 3.18 21.10 31.65 2003 23.20 23.05 5.54 14.72 31.41 2004 24.96 26.18 5.37 10.00 30.48 2005 25.50 26.67 4.68 16.74 30.90 2006 26.20 27.01 4.02 19.32 31.05 2007 29.98 30.57 2.56 25.89 32.93 2008 25.85 26.51 3.83 17.33 31.41 2009 24.72 26.61 5.23 10.00 29.41 2010 24.33 25.01 4.27 12.96 29.28 1998 2010 2 6.47 27.64 4.91 8.00 34.00

PAGE 154

154 Table A 59. Descriptive statistics for salinity (ppt) at station 9. Time period Mean Median Standard deviation Minimum Maximum 1998 24.03 24.50 5.66 12.00 30.80 1999 29.84 30.55 4.18 20.70 34.10 2000 30.39 30.04 2.21 25.60 33.58 2001 28.81 28.99 3.06 22.06 32.18 2002 29.95 30.62 2.80 22.96 32.84 2003 25.10 24.55 4.05 19.68 31.45 2004 25.62 26.84 4.22 14.40 29.03 2005 25.76 27.09 5.15 13.03 30.77 2006 26.34 27.34 4.13 18.87 31.81 2007 30.19 30.61 2.37 26.50 33.82 200 8 26.33 26.57 3.07 19.37 31.18 2009 25.01 25.57 5.42 9.01 31.61 2010 24.78 26.20 4.43 14.50 29.48 1998 2010 27.09 27.96 4.51 9.01 34.10 Table A 60. Descriptive statistics for salinity (ppt) at station 10. Time period Mean Median Standard deviation M inimum Maximum 1998 25.07 25.40 6.21 10.00 32.00 1999 30.38 31.25 4.08 20.00 34.50 2000 31.00 30.70 2.19 27.03 34.30 2001 29.61 30.24 2.05 26.02 32.36 2002 30.18 31.22 2.72 24.54 33.13 2003 26.91 26.62 2.82 21.99 31.51 2004 25.49 27.45 6.14 8.49 30. 26 2005 26.50 28.32 5.93 9.29 31.41 2006 27.31 28.21 3.89 20.27 33.03 2007 30.94 32.13 2.33 27.12 33.46 2008 26.42 27.23 3.26 20.62 30.27 2009 26.26 25.81 2.89 21.04 32.06 2010 25.07 26.16 5.10 11.47 30.16 1998 2010 27.78 28.94 4.51 8.49 34.50

PAGE 155

155 LIS T OF REFERENCES Alam, J.B., A. Hossain, S.K Khan, B.K. Banik, M.R. Islam, Z. Muyen, and M. Habibur Rahman. 2007 Deteriora tion of water quality of Surma R iver. Environmental Monitoring and Assessment 134 : 233 42. American Public Health Association. 1989. S tandard methods for the examination of water and wastewater 17 th edition. New York: American Public Health Association, Inc. Asbury, C.A., and E.T. Oaksford. 1997. A comparison of drainage basin nutrient inputs with instream nutrient loads for seven rive rs in Georgia and Florida, 1986 90 US Geological Survey Water Resources Investigations Report 97 4006. Reston: US Geological Survey. Bales, J.D., S.A. Tomlinson, G. Tillis, and G. Norton. 2006 Flow and salt transport in the Suwannee River estuary, Florid a, 1999 2000: Analysis of data and three dimensional simulations US Geological Survey Professional Paper 1656 B Reston: US Geological Survey. Billen, G., and J. Garnier. 1997. The Phison River plume: Coastal eutrophication in response to changes in lan d use and water management in the watershed. Aquatic Microbial Ecology 13: 3 17. Bledsoe, E. L., and E.J. Phlips. 2000. Relationships between phytoplankton standing crop and physical, chemical, and biological gradients in the Suwannee River and plume region USA. Estuaries 23 : 458 473. Bledsoe, E.L., E.J. Phlips, C.E. Jett, and K.A. Donnelly. 2004. The relationships among phytoplankton biomass, nutrient loading and hydrodynamics in an inner shelf estuary. Ophelia 58: 29 47. Bouchard, A., and J. Haemmerli. 19 92. Trend detection in water quality time series of LRTAP Quebec network lakes. Water, Air, and Soil Pollution 62: 89 110. Boynton, W.R., W.M. Kemp, and C.W. Keefe. 1982. A comparative analysis of nutrients and other factors influencing estuarine phytopla nkton production. In Estuarine Comparisons ed. Victor S. Kennedy, 69 90. New York: Academic Press. Briceo, H. O., and J.N. Boyer. 2010. Climatic controls on phytoplankton biomass in a sub tropical estuary, Florida Bay, USA. Estuaries and Coasts 33: 541 55 3. Bricker, S.B., J.G. Ferreira, and T. Simas. 2003 An integrated methodology for assessment of estuarine trophic status Ecological Modelling 169 : 39 60

PAGE 156

156 Brown, M.T., K.C. Reiss, M.J. Cohen, J.M. Evans, P.W. Inglett, K.S. Inglett, K.R. Reddy, T.K. Fraze r, C.A. Jacoby, E.J. Phlips, R.L. Knight, S.K. Notestein, R.G. Hamann, and K.A. McKee. 2008. Summary and synthesis of the available literature on the effects of nutrients on spring organisms and systems Gainesville: University of Florida Burt, T.P., N.J. K. Howden, F. Worrall, and M.J. Whelan. 2010. Long term monitoring of river water nitrate: how much data do we need?. Journal of Environmental Monitoring 12: 71 79. Cabrera, V.E. 2004. Modeling north Florida dairy farm strategies to alleviate ecological impacts under varying climatic conditions: An interdisciplinary approach Gainesville: University of Florida. Calbet, A., and M.R. Landry. 2004. Phytoplankton g rowth, microzooplankton grazing, and carbon cycling in marine systems. Limnolo gy and Oceanography 49: 51 57. Camargo, J.A. and A. Alonso. 2006. Ecological and toxicological effects of inorganic nitro gen pollution in aquatic ecosystems: A global assessment. Environment International 32: 831 849. Canty, A., and B. Ripley. 2010. boot: Bootstrap R S Plus Functions. R package version 1.3 43. Carey, R.O. 2009. Evaluating temporal and spatial land use influences affecting nutrient water quality in the Biscayne Bay watershed, Florida Gainesville: Uni versity of Florida. Carlson P.R., L.A. Yarbro, K.A. Kaufman,and R.A. Mattson. 2010. Vulnerability and resilience of seagrasses to h coast. Hydrobiologia 649: 39 45. Chan, K.S. 2008 TSA. R package version 0.98. Cla rke, K.R., and R.N. Gorley. 2006. PRIMER v6: User Manual/Tutorial Plymouth: PRIMER E. Clarke, K.R. 1993. Non parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18: 117 143. Cloern, J.E. 2001. Our evolving mod el of the coastal eutrophication problem. Marine Ecology Progress Series 210: 223 253. Cohen, J., P. Cohen, S.G. West, and L.S. Aiken. 2003. Applied multiple regression/correlation analysis for the behavioral sciences 3 rd ed. Mahwah: Lawrence Erlbaum Asso ciates, Inc.

PAGE 157

157 Copeland, R., N.A. Doran, A.J. White and S.B. Upchurch. 2009. Regional and 2003) Florida Geologic al Survey Bulletin No. 69 Tallahassee: Florida Geological Survey. Crand all, C.A., B.G. Katz, and J.J. Hirten. 1999. Hydrochemical evidence for mixing of river water and groundwater during high flow conditions, lower Suwannee River basin, Florida, USA. Hydrogeology Journal 7: 454 467. Darken, P.F., C.E. Zipper, G.I. Holtzman, and E.P. Smith. 2002. Serial correlation in water quality variables: Estimation and implications for trend analysis. Water Resources Researc h 38: 23 29. de Jonge, V.E., J.F. Bakker and M.R. van Stralen. 1996. Possible change in the contribution of the ri ver Rhine and the North Sea to the eutrophic status of the western Dutch Wadden Sea. Netherlands Journal of Aquatic Ecology 30: 27 39. Duarte, C.M. 1995. Submerged aquatic vegetation in relation to different nutrient regimes. Opheli a 41: 87 112. Florida Department of Environmental Protection. 2002. Verified list of impaired waters for the Suwannee River basin (Group 1) Tallahassee: Florida Department of Environment Protection. Florida Department of Environmental Protection. 2010. Site specific informatio n in support of establishing numeric nutrient criteria in Suwannee Estuary/Suwannee Sound/Cedar Keys, Waccasassa Bay, and Withlacoochee Bay Tallahassee: Florida Department of Environment Protection. Fortin, M.J., and M.R.T. Dale. 2005. Spatial analysis: A guide for ecologists Cambridge: Cambridge University Press. Frazer, T .K., M.V Hoyer, S.K. Notestein, J.A. Hale, D.E. Canfield, and S.B. Blitch. 2001. Water quality characteristics of the nearshore Gulf coast waters adjacent to Citrus, Hernando and Levy counties Final Report. Tampa: Southwest Florida Water Management Distri ct. Frazer, T.K., M.V. Hoyer, S.K. Notestein, J.A. Hale, D.E. Canfield, F.E. Vose, W.R. Leavens, S.B. Blitch, and J. Conti. 1998. Nitrogen, phosphorus and chlorophyll relations in selected rivers an d nearshore coastal waters along the Big Bend region of Florida Final Report. Live Oak: Suwannee River Water Management District (SRWMD Contract No. 96/97 156) and the Southwest Florida Water Management District (SWFWMD Contract No. 96/97/157R). Galloway, J.N., W.H. Schlesinger, H. Levy II, A. Michaels, and J.L. Schnoor. 1995. Nitrogen fixation: Anthropogenic enhancement environmental response. Global Biogeochemical Cycles 9: 235 252.

PAGE 158

158 Gibbons, R. D. 1999. Use of combined Shewhart CUSUM control charts for ground water monitoring applications. Ground Water 37: 68 3 691. Giese G.L., and M.A. Franklin. 1996a Magnitude and frequency of floods in the Suwannee River Water Management District US Geological Survey Water Resources Inv estigations Report 96 4176 Reston: US Geological Survey. Giese G.L., and M.A. Franklin. 1996b Magnitude and frequency of low flows in the Suwannee River Water Management District, Florida US Geological Survey Water Resources Inv estigations Report 96 4308 Reston: US Ge ological Survey. Grubbs, J.W., and C.A Crandall. 2007. Exchanges of water between the Upper Floridan aquifer and the lower Suwannee and lower Santa Fe Rivers, Florida US Geological Survey Professional Paper 1656 C Reston: US Geological Survey. Helsel, B .D.R., and R.M. Hirsch. 1991. Statistical methods in water resources. Hydrologic analysis and interpretation Reston: US Geological Survey. Helsel, D.R., D.K. Mueller, and J.R. Slack. 2006. Computer Program for the Kendall family of trend tests US Geologi cal Survey Scientific Investigations Report 2005 5275. Reston: US Geological Survey. Hirsch, R.M., and J.R. Slack. 1984. A nonparametric trend test for seasonal data with serial dependence. Water Resources Research 20: 727 732. Howarth, R. W. 2008. Coasta l nitrogen pollution: A review of sources and trends globally and regionally. Harmful Algae 8: 14 20. Humborg, C., V. Ittekkot, A. Cociasu, and B.V. Bodungen. 1997. Effect of Danube River dam on Black Sea biogeochemistry and ecosystem structure. Nature 386 : 385 388. Irwin, A.J., and Z.V. Finkel. 2008. Mining a sea of data: Deducing the environmental controls of ocean chlorophyll. Plos One 3: e3836. Justic, D., N.N. Rabalais, and R.E. Turner. 1997. Impacts of climate change on net productivity of coastal wat ers: Implications for carbon budgets and hypoxia. Climate Research 8: 225 237. Katz, B. G., and J.K. Bhlke. 2000. Monthly variability and possible sources of nitrate in ground water beneath mixed agricultural land use, Suwannee and Lafayette Counties, Flor ida U S Geological Survey Water Resources Inv estigations Report 00 4219 Reston: US Geological Survey. Katz, B.G., R.S. DeHan, J.J. Hirten and J.S. Catches. 1997. Interactions between groundwater and surface water in the Suwannee River basin. Journal of t he American Water Resources Association 33: 1237 1254.

PAGE 159

159 Katz, B.G., D.H. Hornsby, J.F. Bohlke, and M.F. Mokray. 1999. Sources and chronology of nitrate contamination in spring waters, Suwannee River Basin, Florida US Geological Survey Water Resources Inv es tigations Report 99 4252 Reston: US Geological Survey. Lawrenz, E., J.L. Pinckney, M.L. Ranhofer, H.L. MacIntyre, and T.L. Richardson. 2010. Spectral irradiance and phytoplankton community composition in a blackwater dominated estuary, Winyah Bay, South C arolina, USA. Estuaries and Coasts 33: 1186 1201. Lehman, P.W. 1992. Environmental factors associated with long term changes in chlorophyll concentration in the Sacramento San Joaquin delta and Suisun Bay, California. Estuarie s 15: 335 348. Levin, S.A. 199 2. The problem of pattern and scale in ecology. Ecology 73: 1943 1967. Li, Y., and T.J. Smayda. 1998. Temporal variability of chlorophyll a in Narragansett Bay, 1973 1990. ICES Journal of Marine Science 55: 661 667. Ljung, G.M., and G.E.P. Box. 1978. On a measure of lack of fit in time series models. Biometrika 66: 66 72. Macfarlane, S.I. 1996. Shellfish as the impetus for embayment management. Estuaries 19: 311 319. Mallin, M.A., K.E. Williams, E.C. Esham, and R.P. Lowe. 2000. Effect of human development on bacteriological water quality in coastal watersheds. Ecological Applications 10: 1047 1056. Manly, B.F. J., and D.I. Mackenzie. 2003. CUSUM environmental monitoring in time and space. Environmental and Ecological Statistics 10: 231 247. McLeod, A.I., 20 11. Kendall: Kendall rank correlation and Mann Kendall trend test. R package version 2.2. Mertens, K., I. Vaesen, J. Lffel, B. Ostyn, B. Kemps, B. Kamers, F. Bamelis, J. Zoons, P. Darius, E. Decuypere, J. De Baerdemaeker, and B. De Ketelaere. 2008. Data based design of an intelligent control chart for the daily monitoring of the average egg weight. Computers and Electronics in Agriculture 61: 22 3 232. Mesnil, B., and P. Petitgas. 2009. Detection of changes in time series of indicators using CUSUM contro l charts. Aquatic Living Resources 22:187 192. National Research Council. 2000. Clean Coastal Waters: Understanding and reducing the effects of nutrient pollution Washington, DC : National Academy Press Nixon, S.W. 1995. Coastal marine eutrophication: a definition, social causes, and future concerns. Ophelia 41: 199 219.

PAGE 160

160 Norgart, C. Journal of Land Use and Environmental Law 19: 347 376. Orlando, S.P., L.P. Rozas, G.H. Ward, and C.J. Klein. 1993. Salinity characteristics of Gulf of Mexico estuaries Silver Spring: National Oceanic and Atmospheric Administration, Office of Coastal Resources, Conservation and Assessment. Osborne, J. 2002 Notes on the use of data transformations. Pr actical Assessment, Research & Evaluation 8 http://P AREonline.net/getvn.asp?v=8&n=6 Accessed 20 October 2011 Paerl, H.W. 1999. Cultural eutrophication of shallow coastal waters: coupling changing anthropogenic nutrient inputs to regional management app roaches. Limnologica 29: 249 254. Page, E.S. 1954. Continuous inspection schemes. Biometrika 41: 100 115. Painting, S.J., M.J. Devlin, S.J. Malcolm, E.R. Parker, D.K. Mills, C. Mills, P. Tett, A. Wither, J. Burt, R. Jones, and K. Winpenny 2007 Assessing the impact of nutrient enrichment to estuaries: Susceptibility to eutrophication Marine Pollution Bulletin 55 : 7 4 9 0 Petitgas, P. 2009. The CUSUM out of control table to monitor changes in fish stock status using many indicators. Aquatic Living Resources 22: 201 206. Pittman, J.R., H.H. Hatzell and E.T. Oaksford. 1997. Spring contributions to water quantity and nitrate loads in the Suwannee River during base flow in July 1995 U S Geological Survey Water Resources Inv estigations Report 97 4152 Reston: US Geological Survey. Qian, Y., K.W. Migliaccio, Y. Wan, Y.C. Li, and D. Chin. 2007. Seasonality of selected surface water constituents in the Indian River Lagoon, Florida. Journal of Environmental Q ualit y 36: 416 25. Quinlan, E. 2003. Consequences of nutri ent loading in the Suwannee River and Estuary, Florida, USA. Gainesville: University of Florida. Quinlan E L C.H. Jett, and E.J. Phlips 2009 Microzooplankton grazing and the control of phytoplankton biomass in the Suwannee River estuary, USA. Hydrobiologia 632:127 137 Quinlan, E.L., and E.J. Phlips. 2007. Phytoplankton assemblages across the marine to low salinity transition zone in a blackwater dominated estuary. Journal of Plankton Research 29: 401 416.

PAGE 161

161 R Development Core Team. 2006. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Web: http://www.R project.org. Raabe, B.E.A., R.E. Edwards, C.C. Mcivor, J.W. G rubbs, and G.D. Dennis. 2007. Habitat and hydrology: Assessing biological resources of the Suwannee River estuarine system US Geological Survey Open File Report 2007 1382 Reston: US Geological Survey. Redden, A.M., and N. Rukminasari. 2008. Effects of in creases in salinity on phytoplankton in the Broadwater of the Myall Lakes, NSW, Australia. Hydrobiologia 608: 87 97. Robinson, K.L. 2007. Interactions among phytoplankton, microzooplankton, and mesozooplankton in riverine coastal systems along the west coa st of peninsular Florida. Gainesville: University of Florida. Rogerson, P. 2006. Formulas for the design of CUSUM quality control charts. Communications in Statistics: Theory and Methods 35: 373 383. Rosenberg, R. 1985. Eutrophication The future marine coa stal nuisance? Marine Pollution Bulletin 16: 227 231. Sartory, D.P., and J.U. Grobbelaar. 1984. Extraction of chlorophyll a from freshwater phytoplankton for spectrophotometric analysis. Hydrobiologia 114: 177 187. Seitz, R.D., R.N. Lipcius, N.H. Olmstead, M.S. Seebo, and D.M. Lambert. 2006 Influence of shallow water habitats and shoreline development on abundance, biomass, and diversity of benthic prey and predators in Chesapeake Bay Marine Ecology Progress Series 326 : 11 27 Smith, S.V. 1984 Phosphorus versus nitrogen limitation in the marine environment. Limnology and Oceanograph y 29 : 1149 1160. Strom, S.L., and M.W. Strom. 1996. Microzooplankton growth, grazing, and community structure in the northern Gulf of Mexico. Marine Ecology Progress Series 130 : 229 240. Tam, D. 2009. A theoretical analysis of cumulative sum slope (CUSUM Slope) statistic for detecting signal onset (begin) and offset (end) trends from background noise level. The Open Statistics and Probability Journal 1: 43 51. Tsay, R.S. and G. C. Tiao 1984. Consistent estimates of autoregressive parameters and extended sample autocorrelation function for stationary and nonstationary ARMA models. Journal of the American Statist ical Association 79: 84 96.

PAGE 162

162 Turner, R.E., N.N. Rabalais, R.B. Alexand er, G. McIsaac, and R.W. Howarth. 2007. Characterization of nutrient and organic carbon and sediment loads and concentrations from the Mississippi River into the northern Gulf of Mexico. Estuaries and Coasts 30: 773 790. Upchurch, S.B., C.R. Cain, and J. C hen. 2007. Trends of nitrate concentrations in waters of the Suwannee River Water Management District. Live Oak, Florida: Suwannee River Water Management District. U.S. Environmental Protection Agency. 2006. Final Peace River Basin, Florida Dissolved Oxyge n, Nutrient, Turbidity and TSS Total Maximum Daily Loads Washington, DC,: US Environmental Protection Agency Vitousek, P.M., J.D. Aber, R.W. Howarth, G.E. Likens, P.A. Matson, D.W. Schindler, W.H. Schlesinger and D.J. Tilman. 1997. Human alteration of th e global nitrogen cycle: Sources and consequences. Ecological Applications 7: 737 750. Warner, S.C., K.E. Limburg, A.H. Ari o, M. Dodd, J. Dushoff, K.I. Stergiou, and J. Potts. 1995. Time series compared across the land sea gradient. In Ecological Time Ser ies T.M. Powell and J.H. Steele, 243 273. New York: Chapman and Hall. Wazniak, C.E., M.R. Hall, T.J.B. Carruthers B. Sturgis W.C. Denni son, and J.R. Orth. 2007 Linking water quality to living resources in a mid Atlantic lagoon system, USA. Ecological A pplications 17: S64 S78. Wolfe, L.E. and S.H. Wolfe. 1985. The ecology of the Suwannee River Estuary: An analysis of ecological data from the Suwannee River Water Management District study of the Suwannee River Estuary, 198 3 1983. Florida Department of En vironmental P rotection, Tallahassee, FL Zeileis, A., and T. Hothorn 2002. Diagnostic checking in regression r elat ionships. R News 2: 7 10. Web: http ://CRAN.R project.org/doc/Rnews.

PAGE 163

163 BIOGRAPHICAL SKETCH Andrea Krzystan was born in Panama City, Florida bu t grew up in Severna Park, Maryland. The proximity to the Chesapeake Bay and the Smithsonian museums in Washington D.C. inspired a passion for science and nature from a young age. She pursued these interests by returning to Florida and studying Marine Scie nce at Eckerd College in St. Petersburg. The unique education provided by Eckerd College included extensive field and laboratory experiences, a variety of research internships, and other opportunities including playing varsity volleyball and participating in many service learning experiences. Each of these experiences prepared her for life after graduation. After graduation in 2006, Andrea worked as a Water Quality Specialist for the Florida Department of Environmental Protection at Tampa Bay Aquatic Prese rves. During her time there, Andrea applied her experience to develop a water quality monitoring program and also developed new interests in resource management and geographic information systems. After three years, she decided to further her education in marine ecology by returning to graduate school and joined the lab of Dr. Tom Frazer at the University of Florida in 2009. resource management have been strengthened. After earning he r m Andrea p lans to pursue a career that combine s these interests.