Analyses of Temporal Changes in Trophic State Variables in Florida Lakes

MISSING IMAGE

Material Information

Title:
Analyses of Temporal Changes in Trophic State Variables in Florida Lakes
Physical Description:
1 online resource (151 p.)
Language:
english
Creator:
Bigham, Dana L
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Fisheries and Aquatic Sciences, Forest Resources and Conservation
Committee Chair:
Canfield, Daniel E
Committee Co-Chair:
Duarte, Carlos M
Committee Members:
Cichra, Charles E
Bachmann, Roger Werner
Delfino, Joseph J

Subjects

Subjects / Keywords:
changes -- florida -- lakes -- trophic
Forest Resources and Conservation -- Dissertations, Academic -- UF
Genre:
Fisheries and Aquatic Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Appropriate assessments of lake change and trends are necessary to advance limnological studies to best estimate factors driving lake change, such as climate. The citizen monitoring Florida LAKEWATCH database was used to evaluate decadal-scale trends in the trophic state variables; total phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements. Two subsets of the LAKEWATCH database were used: monthly samples of the trophic state variables collected for at least 20 years for 27 Florida lakes, 193 lakes with data collected for at least 15 years. Linear regression, Kendall-Tau, and ARMA/ARIMA time series models were evaluated to detect trends in the trophic state variables for the 27 Florida lakes. Different statistical results were found among the evaluated methods. An alternative approach was developed, separating data into six categories prior to linear regression analysis, which provided similar detection of trends as ARMA/ARIMA time series models. For the 193 Florida lakes, the alternative approach detected increasing trends in total phosphorus (21%), total nitrogen (26%), chlorophyll concentrations (12%), and decreasing trends in water clarity measurements (18%). Less than 5% of the lakes experienced trends in all trophic state variables. Three clusters of lakes with similar trends in the trophic state variables were identified across the State of Florida. Patterns of phytoplankton growth and senescence (seasonal changes) were recurrent and synchronous for the examined Florida lakes. Annual elevated chlorophyll concentrations occurred June through October following annual climate cycles of air temperature and rainfall. The occurrence of extreme chlorophyll events increased in three of the 27 lakes that had the longest (= 20 years) record. Seasonal patterns in waters classified as hypereutrophic differed from other trophic categories. The resulting assessment of lake change over multiple scales of time and space focuses future research and management efforts in the State of Florida and at a global level.
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 Dana L Bigham.
Thesis:
Thesis (Ph.D.)--University of Florida, 2012.
Local:
Adviser: Canfield, Daniel E.
Local:
Co-adviser: Duarte, Carlos M.
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 2012
System ID:
UFE0044896:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 ANALYSES OF TEMPORAL CHANGES IN TROPHIC STATE VARIABLES IN FLORIDA LAKES By DANA LYNNE BIGHAM A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Dana Lynne Bigham

PAGE 3

3 To my loving family and friends

PAGE 4

4 ACKNOWLEDGMENTS I th ank Jesse Stephens for his encouragement, light hearted attitude, and love; without his support my time as a graduate student would not have been as enjoyable. The serenity and reassurance of my family (Charles, Dona, Kent, Grant, Sheila, Bill, Alfred, Theresa, Bonnie, Marie Joel, and Leah) and close friends ( Ann, And rew, Dave, Andy, Paula, Loren, Colleen, and Felicia) provided the zeal I needed to complete my degree at the University of Florida. I am extremely grateful to my chair and cochair, Daniel E. Canfield Jr. and Carlos M. Duarte, and my supervisory committee, Roger W. Bachmann, Charles E. Cichra, and Joseph J. Delfino. Dan Canfield taught me to say no, to be confident, and to always ask questions. He also provided me with invaluable opportunities to better myself both professionally and personally. Carlos Du arte taught me how to ask research questions and the importance to relate to the big picture. Roger Bachmann taught me to how to be an objective scientist and provided irreplaceable lessons that illustrated the need to take a moment to think. Chuck Cichra taught me to follow my dreams and to take advantage of opportunities, like the countless opportunities he offered to improve m y teaching and extension efforts Joe Delfino taught me the importance of practical research, to have achievable objectives that w ill advance science and will always be a fellow Badger. The guidance, patience, and support of Mark Hoyer showed me the power of perseverance and motivation, of which I am very thankful I thank Felipe Carvahlo, Drs. Robert Carlson, Mike Allen Rob Ahrens, and Daryl Parken for their guidance in the use of various statistical methods. I greatly appreciate the encouragement and confidence of Dr. Marilyn Bachmann, Christine Horsburg h and Harry Nelson. Finally, I am indebted to

PAGE 5

5 the LAKEWATCH staff a nd ci tizen scientist s for developing, sampling, and sustaining an excellent database. I gr atefully acknowledge research funding and financial support from the Florida LAKEWATCH program, the University of Florida, College of Agriculture and Life Sciences Jack L. Fry Award for Excellence in Graduate Student Teaching, University of Florida Agricultural Womens Club Vam C. Yor k Award, and the University of Florida Womens Club Award. Additionally support for conference travel was provided by the University of Florida Graduate Student Council, University of Florida College of Agriculture and Life Sciences, Association for the S ciences of Limnology and Oceanography, North American Lake Management Society, and the Florida Lake Management Society.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 8 LIST OF FIGURE S ........................................................................................................ 11 ABSTRACT ................................................................................................................... 14 CHAPTER 1 INTRODUCTION .................................................................................................... 16 2 STAT ISTICAL METHODS AND AN ALTERNATIVE APPROACH TO DETECT TRENDS IN TROPHIC STATE VARIABLE TIME SERIES DATA .......................... 21 Background ............................................................................................................. 21 Materials and Procedures ....................................................................................... 23 Assessment ............................................................................................................ 24 Components of Vari ance .................................................................................. 25 Linear Regression ............................................................................................ 26 Kendall Tau ...................................................................................................... 28 Time Series Modeling ....................................................................................... 30 Alternative Approach ........................................................................................ 35 Discussi on .............................................................................................................. 36 Comments and Recommendations ......................................................................... 39 3 DECADAL SCALE TRENDS IN TROPHIC STATE VARIABLES WITHIN A LARGE POPULATION OF FLORIDA LAKES ......................................................... 64 Background ............................................................................................................. 64 Methods .................................................................................................................. 66 Results .................................................................................................................... 70 Discussion .............................................................................................................. 73 4 SEASONAL PATTERN S OF PHYTOPLANKTON BIOMASS AND RESPONSES TO CLIMATE IN SUBTROPICAL, FLORIDA LAKES ...................... 94 Background ............................................................................................................. 94 Methods .................................................................................................................. 97 Datasets ........................................................................................................... 97 Approach to Identify Seasonal Patterns in Phytoplankton Biomass ................. 98 Statistical Analyses .......................................................................................... 99 Climate Relationships ..................................................................................... 102 Results .................................................................................................................. 103

PAGE 7

7 Magnitude of Phytopl ankton Biomass ............................................................ 103 Variability of Seasonal Patterns across Florida Lakes .................................... 104 Variability of Seasonal Patterns by Lakeyear Trophic State Category .......... 105 Temporal Shifts in the Occurrence of Extreme Chlorophyll Events ................ 107 Climate Relationships ..................................................................................... 108 Discussion ............................................................................................................ 109 5 SUMMARY, RECOMMENDATIONS, AND MAJOR CONCLUSIONS .................. 134 Summary and Recommendations ......................................................................... 134 Major Conclusions ................................................................................................ 139 LIST OF REFERENCES ............................................................................................. 141 BIOGRAPHICAL SKETCH .......................................................................................... 151

PAGE 8

8 LIST OF TABLES Table page 2 1 Summary statistics (i.e., mean, median, minimum, maximum, and coefficient of variation) for untransformed total phosphorus (g/L), total nitrogen (g/L), chlorophyll concentrations (g/L) and water clarity measurements (m) among annual mean data for the examined population of Florida lakes. ........................ 41 2 2 Results of variance component analysis and the percent of variance attributed to laketo lake differences, year to year differences, month to month differences, and residual error (includes stationto station and laboratory differences) using monthly data for the population of 27 Florida lakes. Within the individual 27 Florida lakes, the mean percent of variance attributed to year to year differences, monthto month differences, stationto station differences, and residual error (laboratory differences) using monthly data are presented. ............................................................................................ 42 2 3 Percentage of the population of lakes with monotonic increasing trends, monotonic decreasing trends, and no trends over 20plus years detected by the use of linear regression models, Kendall Tau analysis, ARMA/ARIMA time series models, and the alternative approach following Prairie (1996) and Byrhn and Dimberg (2011). Monthly data and annual data were evaluated for total phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements for the examined population of Florida lakes. ............................ 43 2 4 Linear regression analysis detection of a significant monotonic trend ( slope value, and coefficient of determination ( R) of the monthly time series logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes. Linear regression analysis completed using the residuals of the best polynomial fit, to remove variance due to season, is denoted after the .................................................................................................... 44 2 5 Linear regression analysis detection of a significant monotonic trend ( slope value, and coefficient of determination ( R) of the annual mean time series logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes. ...................................................................................................... 46 2 6 Kendall Tau analysis detection of significant monotonic trends ( with the tau value for annual mean total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity (SD) data for the 27 Florida lakes. ............................... 48 2 7 T ime ) denotes significant change and denotes significant change and monotonic trend), the Akaikes Information Criterion (AIC) value, the time lag corresponding to the significant autocorrelation (AC), and the

PAGE 9

9 coefficient of determination (R2) using monthly logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes. ............................................. 49 2 8 T denotes significant change and denotes significant change and monotonic trend), the Akaikes Information Criterion (AIC) value, the time lag corresponding to the significant autocorrel ation (AC), and the coefficient of determination (R2) using annual mean logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes. ................................... 53 2 9 Modified l inear regression analysis, with six categories of annual mean data, to detect significant monotonic trends ( and coefficient of determination ( R) for logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes. ...................................................................................... 58 3 1 Summary statistics (i.e., mean, median, minimum, maximum, and coefficient of variation (%)) for annual mean total phosphorus ( g/L), total nitrogen ( g/L), chlorophyll concentrations ( g/L), and water clarity measurements (m) for the population of 193 Florida lakes. ............................................................... 78 3 2 Summary statistics (i.e., mean, median, minimum, m aximum, and coefficient of variation (%)) for supplemental data including mean depth (m), surface area (ha), specific conductance ( S), and true color (Pt Co Units) for the population of 193 Florida lakes. .......................................................................... 79 3 3 Percentage of the population of 193 Florida lakes for total phosphorus, total nitrogen, chlorophyll concentration, and water clarity measurements within each trophic state classification (i.e., oligotrophic, mesotrophic, eutrophic, and hypereutrophic) following Forsburg and Ryding (1980). .............................. 80 3 4 Perc entage of the population of Florida lakes with increasing trends, decreasing trends, and no trends detected in annual mean total phosphorus, total nitrogen, chlorophyll concentrations and water clarity measurements over a period or record of at least 15 years. ....................................................... 81 3 5 Linear regression analysis using six data points where the point reflects the mean among the six annual mean data for the logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD). A s ignif icant monotonic trend ( corresponding slope value, and coefficient of determination ( R) for each trophic state variable for the 193 Florida lakes. .................................................. 82 3 6 Number of lakes (no. of lakes) out of the examined annual mean data for 193 Florida lakes with: 1) decadal scale increasing trends in total phosphorus (TP), total nitrogen (TN), and chlorophyll concentrations (CHL) and decreasing trends in water clarity measurements (SD) and 2) decadal scale

PAGE 10

10 decreasing trends in total phosphorus (TP), total nitrogen (TN),and chlorophyll concentrations (CHL), and increasing trends in water clarity measurements (SD), and 3) no t rend in any of the four trophic state variables. ............................................................................................................ 90 4 1 Summary statistics of monthly chlorophyll samples (g/L) collected over a 20 plus year period for 27 Florida lakes. ............................................................... 116 4 2 Linear regression analysis of slope values by year. The annual slope values were derived from determination of the percent number of records greater than a given chlorophyll concentration against the corresponding logarithmic (base 10) transformed chlorophyll concentration. Significant liner relationships indicate a change in the frequency of occurrence of extreme chlorophyll concentrations over the examined period of record for the individual 27 Florida lakes. ............................................................................... 117 4 3 S tatistical identification of periodic component of variability in chlorophyll concentrations by spectral density analysis, generated by the time series model analysis, by the Fishers Kappa Test (p values listed). Visual identification (Y= yes and N= no of a peak at 12 months in the chlorophyll variance indicated the individual Florida lake exhibited a seasonal periodic component across the examined period of record. ........................................... 118

PAGE 11

11 LIST OF FIGURES Figure page 1 1 Annual mean total phosphorus concentrations (mg/L) from 1970 to 2005 for Florida water bodies from the Florida Department of Environmental Protection 303d and 305b, 2008 report. ............................................................. 20 2 1 Linear regression model analysis and the associated 95% confidence intervals for annual mean total phosphorus concentrations ( g/L). A) Total phosphorus concentrations in Lake Lorraine located in Lake County, Florida (p < 0.0001, R2 = 0.77). B) Total phosphorus concentrations (g/L) in Little Orange Lake located in Alachua County, Florida (p = 0.03, R2 = 0.23). The Kendall Tau, A RMA/ARIMA time series models, and the proposed alternative methods detected a significant trend in total phosphorus concentrations (g/L) in Lorraine Lake, but not in Little Orange Lake. ........................................ 59 2 2 Example of a time series plot of annual mean total phosphorus concentrations (g/L) for Lake Como (located in Putnam County, Florida) the corresponding autocorrelation function (ACF) plot, and the corresponding partial autocorrelation function (Partial ACF) plot against time with successively time units (years) lagged by one. The dotted lines on the ACF and PACF plot represent the upper and lower 95% confidence intervals. The statistically sign ificant ACF and Partial ACF values along with the pattern of the lag terms were used to estimate the autocorrelation term (AR) and moving average (MA) terms of the time series model. The selected time series model for these data was ARIMA (1,1,1). ................................................ 60 2 3 Annual mean total phosphorus concentrations (g/L) in Little Lake Santa Fe located in Alachua County, Florida. Linear regression and Kendall Tau analyses detected significant increasing monotonic trends in total phosphorus (g/L), while the time series model detected a significant change in total phosphorus concentrations (g/L), but no significant trend over the examined 24year record (19862009). .............................................................. 62 2 4 Number of classes determined (three intervals as pictured) using the bivariate linear regression model and the associated 95% conf idence intervals (top figure). As the number of classes increases the coefficient of determination (R2) values increase with a value of 0.65 noted as a predictively powerful linear regression model (bottom figure). Data and figures are from Prairie (1996). ........................................................................... 63 3 1 Distribution of the examined population of 193 Florida lakes. ............................ 91 3 2 Number of years each lake was sampled (N=193 Florida lakes). The numbers above the bars denote the number of lakes that were sampled for the respective number of years. ......................................................................... 92

PAGE 12

12 3 3 Florida lakes (N=105 lakes) with detected decadal scale trends of degradation (i.e., increases in total phosphorus (TP), t otal nitrogen (TN), and chlorophyll concentrations (CHL) and decreases in water clarity (SD) measurements, and trends of improvement (i.e., decreases in TP, TN, and CHL concentrations and increases in SD). Spatial clusters of individual lakes exhibiting si milar trends of degradation in one or more of the examined variables were identified (A and B) and a spatial cluster of lakes exhibiting similar trends of improvement in one or more of the variables were identified (C). ..................................................................................................................... 93 4 1 Spectral density plot generated from time series model analysis for monthly chlorophyll concentrations over a 21year period in Lake Alto located in Alachua C ounty, Florida. The peak, at a period of 12, indicates there is a seasonal component in the chlorophyll concentrations in Lake Alto. ................ 119 4 2 Mean percent (%) difference of monthly chlorophyll concentrations over an annual cycle for 27 subtropical, Florida lakes. The bars represent the 95% confidence intervals around the mean of the monthly mean % difference in chlorophyll concentrations from the annual mean. Positive differences indicate concentrations greater than the mean and negative differences indicate concentrations less than the mean. ..................................................... 120 4 3 Frequency of occurrence of extreme chlorophyll events represented as the maximum chlorophyll concentrations (light grey bars) and the chlorophyll concentrations exceeding two times the grand mean (dark gr ey bars) summarized for each month among the years sampled for the 27 Florida lakes (N = 611 total maxima values and N = 373 values exceeding the grand mean by double). .............................................................................................. 121 4 4 Mean percent (%) difference of monthly chlorophyll concentrations calculated over an annual cycle by classification into lakeyear trophic state categories A) oligotrophic, B) mesotrophic, C) eutrophic, and D ) hypereutrophic classification. The bars represent the 95% confidence intervals associated with the mean for the monthly mean % difference in chlorophyll concentrations. ................................................................................................. 122 4 5 Frequency of occurrence of extreme chlorophyll events represented as the maximum chlorophyll concentrations (light grey bars) and the chlorophyll concentrations exceeding the grand mean by double (dark grey bars) summarized for each month among the years sampled by classification into lake year trophic categories A) oligotrophic (N= 948 total lakeyears), B) mesotrophic (N= 2051 total lakeyears), C) eutrophic (N=2365 total lakeyears), and D) hypereutrophic (N=1248 total lakeyears. ................................. 124 4 6 Monthly air temperature (C) data averaged over a 24year period for the five nearest collection sites (solid line) and the corresponding mean % difference

PAGE 13

13 of monthly chlorophyll concentrations over an annual cycle (dotted line) for the examined 27 Florida lakes. ......................................................................... 126 4 7 Average monthly rainfall sum (cm) (dotted line) and the average monthly chlorophyll concentrations (g/L) (solid line) calculated among the annual data for the individual Florida lake, which are represented as Lake Name (County of location). ......................................................................................... 1 27 4 8 Mean percent (%) difference of monthly chlorophyll concentrations over an annual cycle for a population of 193 Florida lakes (open circles connected by dotted line) and the population of 27 Florida lakes (closed circles connected by solid line). The bars represent the 95% confidence intervals around the mean of the monthly mean % diff erence in chlorophyll concentrations from the annual mean. .............................................................................................. 133

PAGE 14

14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ANALYSES OF TEMPORAL CHANGES IN TROPHIC STATE VARIABLES IN FLORIDA LAKES By Dana Lynne Bigham December 2012 Chair: Daniel E. Canfield Jr. Cochair: Carlos M. Duarte Major: Fisheries and Aquatic Sciences Appropri ate assessments of lake change and trends are necessary to advance limnological studies to best estimate factors driving lake change, such as climate. The citizen monitoring Florida LAKEWATCH database was used to evaluate decadal scale trends in the trophic state variables ; total phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements. Two subsets of the LAKEWATCH database were used : monthly samples of the trophic state v ariables collected for at least 20 years for 27 Florida lakes 193 lakes with data collected for at least 15 years Linear regression, Kendall Tau, and ARMA/ARIMA time series models were evaluated to detect trends in the trophic state variables for the 27 Florida lakes. Different statistical results were found among the evaluated methods. An alternative approach was developed separating data into six categories prior to linear regression analysis, which provided similar detection of trends as ARMA/ARIMA time series models F or the 193 Florida lakes, the alternative approach detected increasing trends in total phosphorus (21%), total nitrogen (26%), chloro phyll concentrations (12%), and decreasing trends in water clarity measurements (18%). Less than 5% of the lakes experienced trends in all

PAGE 15

15 trophic state variables. T hree clusters of lakes with similar trends in the trophic state variables were identified across the State of Florida. P atterns of phytoplankton growth and senescence (seasonal changes) were re current and synchronous for the examined Florida lakes. Annual elevated chlorophyll concentrations occurred June through October following annual climate cycles of air temperature and rainfall. The occurrence of extreme chlorophyll events increased in thre e of the 27 lakes that 20 years) record S easonal patterns in waters classified as hypereutrophic differed from other trophic categories. The resulting assessment o f lake change over multiple scales of time and space focuses future research and management efforts in the State of Florida and at a global level.

PAGE 16

16 CHAPTER 1 INTRODUCTION Concern about the functioning of the worlds aquatic ecosystems has long been of interest yet research efforts have commonly focused on aquatic ecosystems as individual, unique systems. Although lakes are of global importance (Downing et al. 2006), limnological studies have historically aspired to understand functional processes within the individual lake or at a local sca le (Thienemann 1925; Naum ann 191 9). Individual lake and local scale limnological studies of the past century have greatly advanced the aquatic sciences; however, as the needs of society change, there has been encouragement for limnologists to upscale to a global lev el (Jumars 199 0; Downing 2009) by concentrating on tractable, soluble problems to answer big environmental questions (Rigler and Peters 1995). One of the biggest environmental issues scientists, managers, and policy makers currently face is how to assess lake changes and trends over multiple scales of time and space (Williamson et al. 2009). Appropriate assessments of lake changes and trends are necessary to provide the required support for limno logical studies to move forward and investigate relationships of lake change s and global factors, such as anthropogenic or climate drivers. Lakes are constantly changing (Knowlton and Jones 2006) and the lake variables collected to measure lake change and estimate trends (e.g., trophic state variables like total phosphorus, tot al nitrogen, chlorophyll concentrations, and water clarity as measured by the use of a Secchi disk) are of a random nature, highly variable temporally and spatially ( Hkanson and Duarte 2008). Temporal and spatial variability are not frequently considered in lake assessments (Knowlton and Jones 2006), but consideration of such variability is important because identified directional changes and

PAGE 17

17 trends in lake trophic state variables may be attributed to global factors, like climate, e xacerbated by global changes (see Kernan 2010). Different statistical methods exist to determine changes and trends in lake trophic state time series data, many of which account for variability (Kendall 1938; Esterby 1997; Stow et al. 1998; Burkholder et al. 2006). However, regar dless of the statistical method used, the sampling frequency and duration suggested to best represent a lakes behavior range from 6 years of consecutive data (Molot and Dill on 1991) to 12 years (Howden et al. 2011) to at least 20 years (Knowlton and Jones 2006). There are limited long term data sets that meet these suggested requirements for individual lakes, but especially for populations of lakes. To understand how to assess lake changes and trends over time, that also account for variability, alternativ e methods that provide analogous statistical meani ngful (Bryhn and Dimberg 2011) results are needed. The need to understand how to assess lake changes and trends in lake trophic state variables is of prodigious importance in the State of Florida. For example, the proposed numeric nutrient criteria by the United States Environmental Protection Agency (USEPA) for Floridas lakes (USEPA 2010), the concern of the potential effects of Floridas increasing human population (e.g., from 1900 to 2000 Floridas population has grown by almost 3000% ), or the recent attribution of rising phosphorus levels to the cumulative effects of non point sources of pollution (Figure 11) are current issues of concern and importance for Florida waters. The above issues, however, would be better evaluated with su pportive evidence of whether Floridas lakes have experienced long term (i.e., decadal scale) trends. Fortunately, the Florida LAKEWATCH program has sampled over 1, 500 Florida lakes since 1986 and compiled an extensive dataset

PAGE 18

18 (Canf ield et al. 2002) that includes a subset of 27 Florida lakes with consecutive monthly samples c ollected for at least 20 years and a subset of 193 Florida lakes with monthly samples collected for at least 15 years. Using these LAKEWATCH data to evaluate trophic state variables in Floridas lakes not only contribute s to the understanding of how to bes t assess lake changes and trends, but also provides the background needed to advance scientific research, lake management efforts, and political agendas. Limnology is exclusive from many other sciences in that major research conclusions are written into public law globally (Downing 2009) (e.g., phosphorus chlorophyll transparency relationships ( Dillon and Reigler 1975; Jones and Bachmann 1976. Therefore, the results and conclusions gained from examination of the robust, LAKEWATCH dataset are not limited to the State of Florida, but applicable to help solve global issues as well. The examination of a long term dataset available for a population of lakes makes it possible to examine temporal and spatial change and, thereby, relate measurable response variables such as temperature or rainfall (Williamson et al. 2009; Gaiser et al. 2009). Limnologists have recognized that lakes in their natural state are influenced by edaphic, morphometric, and climatic factors (Naumann 1919; Chandler 1944; Moyle 1956) and demonstrated strong relationships among lake trophic state variables total phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements (Sakamoto 1966; Jones and Bachmann 1976; Carlson 1977; Canfield and Bachmann 1981; Bachmann et al. 2012a). The relationship between climate factors (e.g., temperature and rainfall) and phytoplankton biomass (estimated by chlorophyll concentrations) has become of particular interest due to the projection of change in the

PAGE 19

19 global climate (Mann et al. 1998; Magnuson et al. 2000; IPCC 2007) and consequent changes in the patter ns of phytoplankton biomass and the limnological mechanisms of lakes (Kernan et al. 2012; Jeppensen et al. 2007 a 2010). As seasonal patterns of phytoplankton biomass contribute to the intra annual and inter annual variability, it will be important to document and incorporate these patterns, particularly when investigating global climate change effects on lake systems. Establish ing appropriate methods to access change and trends in lake trophic state variables to understand lake variation helps overcome some of the limitations when exploring factors contributing to limnological change. In Chapter 2, various statistical methods were used to detect patterns of long term change and trends in trophic state variables among a population of Florida lakes. The results of the evaluated methods were compared and an alternat ive method was proposed to detect trends One of the proposed methods was used to evaluate the alternative hypothesis that Florida lakes, as a population, have experienced trends in total phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements over a decadal period of record, described in Chapter 3. In Chapter 4, annual seas onal patterns were identified among the population of subtropical, Florida lakes and the influence of climatic factors (i.e., temperature and rainfall ) and lake trophic status on the seasonal patterns were examined. Thereafter, a final discussion of how these results advance the understanding of change and trends in lake trophic state variables and also future aquatic research are provided.

PAGE 20

20 Figure 11. Annual mean total phosphorus concentrations (mg/L) from 1970 to 2005 for Florida water bodies from the Florida Department of Environmental Protection 303d and 305b, 2008 report.

PAGE 21

21 CHAPTER 2 STATISTICAL METHODS AND AN ALTERNATIVE APPR OACH TO DETECT TRENDS IN TROPHIC STATE VARIABLE TIME SERIES DATA Background The plight of limnologists and oceanographers is how best to detect trends in aquatic data and explain these trends to fellow scientists and nonscientists. There are many statistical methods from the basic t test to more complicated statistical models like time series models that can be used to detect patterns of change and trends in environmental data. The problem is, however, each statistical method may provide different results (Esterby 1997; Stow et al. 1998; Kundezewicz and Robson 2004). Different statistical results are of concern because there is a potential for scientists to make erroneous conclusions that may not only hinder the advancement of science, but also corroborate unsuitable development and evaluation of public policy. I t is important as the number and power of statist ical tools increase, to assess and evaluate various statistical methods to determine which are appropriate to answer ecological questions. Environmental data, particularly aquatic data (Hkanson and Duarte 2008), are variable both temporally and spatially; t he data are of a random nature (i.e., change randomly). Due to the high variability of aquatic data, statistical determination of a significant trend over a given period of record can be difficult at any reasonable confidence level and further influenced by the number of samples analyzed (Prairie 1996; Hkanson and Duarte 2008; Bryhn and Dimberg 2011). Linear regression analysis is commonly used by managers of environmental resources to detect significant trends in time series data. The assumptions of classical parametric models (i.e., normality, linearity, and independence) are not usually met by environmental data (Esterby 1997). The parametric least squares l inear regression analysis, for example,

PAGE 22

22 commonly violates underlying statistical assumptions (Lo ftis et al. 1996; Prairie 1996). Thus, nonparametric tests like the Kendall Tau (Kendall 1938) are commonly used as these methods violate fewer statistical assumptions and supposedly better account for idiosyncrasies in environmental data. T ime series model s (i.e., ARMA and ARIMA models) have been recommended as a powerful statistical method to detect trends (Burkholder et al. 2006; Bendat and Piersol 2010), especially as the model account s for extreme values, which are ubiquitous in aquatic time series dat a. T im e series models have historically been used in economics and the social sciences (McCleary 1980), but more recently in the aquatic sciences possibly due to the increased focus to a nal yze time series data. The objective of this chapter was to evaluate the statistical methods of linear regression, Kendall Tau, and ARMA/ARIMA time series models to detect decadal scale trends in the examined trophic state variables and to provide statistical meaningful (Bryhn and Dimberg 2011), alternative approach that offered a simplistic method with results comparable to a more complex statistical method. In this chapter the term change was used to describe the variability of the time series data The term trend implies the overall unidirectional movement (i .e., monotonic), either increasing or decreasing, of the time series data. The statistical evaluation of long term (i.e., monthly data collected at least 20 years) trophic state variables (i.e., total phosphorus, total nitrogen, chlorophyll concentrations and water clarity measurements) provided a direct comparison of results among different statistical methods. Furthermore, the trophic state data were available for a number of Florida lakes, enhancing comparison of the statistical

PAGE 23

23 evaluation of patterns of change and tr ends. The results from this chapter refine the understanding of how to detect patterns of change and trends in time series data, which are applicable not only to the aquatic sciences, but other scientific disciplines as well. In addition, t he examined dataset and similar datasets are policy relevant (Urquhart et al. 1998) and many times these data are central to establishing standards to protect aquatic ecosystems (e.g., Bachmann et al. 2012b ). Materials and Procedures The aquatic time seri es data examined included trophic state variables obtained from the Florida LAKEWATCH database. The Florida LAKEWATCH monitoring program began in 1986 with the goal to collect trophic state information for individual lakes across the State of Florida and t o build a long term data base (Canfield et al. 2002). The data collected by the LAKEWATCH citizen scientists for each trophic state variable are not significantly different from those collected by professionals ( Hoyer et al. 2012). Therefore, the Florida L AKEWATCH program includes reputable data that have been collected from over 1,500 Florida lakes since 1986. A subset of the Florida LAKEWATCH database was used in this chapter The subset included lakes with monthly samples (each month of the year was sampled) for total phosphorus, total nitrogen, chlorophyl l, and water clarity (as measured by the use of a Secchi disk) collected for at least 20 years and up to 24 years. For nutrients and chlorophyll, monthly data were available for 27 Flori da lakes and water clarity measurements were available for 19 Florida lakes. The number of lakes with water clarity measurements was smaller because some water clarity measurements exceeded the water depth or measurements could not be made due to the presence of aquatic ma crophytes. T he frequency and duration of the examined trophic state data exceeded most of the suggested data

PAGE 24

24 requirements to appropriately account for variance and detect trends in lake systems (Molot and Dillon 1991 Knowlton and Jones 2006, Howden et al. 2011). Assessment The examined population of Florida lakes ranged in trophic status from oligotrophic to hypereutrophic and encompassed the trophic states found across the State of Florida (Table 2 1; Canfield and Hoyer 1988). The analytical methods for t otal phosphorus, total nitrogen, and chlorophyll concentrations were consistent through time and followed the analytical procedures outlined in Canfield et al. (2002). For each trophic state variable, a mean value for the three stations sampled on each sampling date at each lake was calculated to get a monthly mean value. An annual mean was calculated from the monthly means Some months, however, were not sampled because stochastic events inhibited sample collection (e.g., hurricanes or droughts). Because time series model analysis requires continuous data (Box and Jenkins 1976), a missing monthly datum was replaced by the mean value calculated from the previous and following months. Data sets with less than 15% missing data can be repaired (i.e., the mean value replaced the missing datum) and time series analysis completed without introducing any substantial error (Kriendler and Lumsden 2006). All lakes examined had less than 15% missing data per each trophic state variable. Data were analyzed with statist ical packages JMP version 8.0 (SAS Institute 2007) and R, PC version 2.11.1 (R Development Core Team 2008). Coefficients of variation were calculated by dividing the standard deviation by the mean. When parametric statistics were used (i.e., linear regress ion and time series analyses) data were logarithmically (base 10) transformed to meet the requirements of normality

PAGE 25

25 (Snedecor and Cochran 1979). All statements of statistical significance were at a probability of < 0.05. Components of V ariance Variance c omponent analysis was completed using JMP version 8.0 (SAS Institute 2007) on the logarithmic (base 10) transformed data to better understand the factors contributing to the observed variance at the population and individual lake levels T he amount of vari ance attributed to lake, year, month, station, and residual error (including laboratory error) was examined using the monthly data for the population of 27 Florida lakes. T he amount of variance attributed to year, month, stations, and residual error was ex amined using the monthly data for the individual 27 Florida lakes. The variance component analysis for the examined population of Florida lakes demonstrated that the majority of the variance for total phosphorus (82%), total nitrogen (86%), chlorophyll (82%), and water clarity (83%) was due to laketo lake differences. The majority of the remainder of the variance was either due to year to year differences (total phosphorus (TP) = 9%, total nitrogen (TN) = 7%, chlorophyll (CHL) = 6%, water clarity (SD) = 8% ) or month to month differences (TP = 6%, TN = 5%, CHL = 10%, SD = 8%). Residual error, which includes stationto station and laboratory variance, accounted for less than 5% of the total variance (TP = 3%, TN = 2%, CHL = 2%, SD = 1 %) (Table 2 2). Variance component analysis of individual sampling units (i.e., lakes) and the primary plant nutrients, phosphorus and nitrogen, demonstrated that year to year differences accounted for on average, 45% (range 8% to 74%) of the variance in T P and 51% (range 17% to 72%) of the variance in TN. Monthto month differences accounted for, on average, 39% of the variance in TP (range 21% to 61%) and 37%

PAGE 26

26 (range 16% to 78%) of the variance in TN. For the biological variable chlorophyll and the physical variable water clarity which is most closely correlated with chlorophyll, year to year differences accounted for, on average, 55% of the variance in CHL and 48% of the variance in SD. Monthto month differences in CHL ranged from 22% to 85% and in SD ranged from 12% to 80% (Ta ble 22) Linear R egression Simple least squa res linear regression analysis Equation 21 (Snedecor and Cochran 1980) was used to examine the relationship between the trophic state variable and time for each individual lake: Yi= 0 1Xi + i (2 1) With: Yi= dependent variable Xi= independent variable 0 = intercept 1 = slope i = error term 1) was equal to zero. When the slope was not equal to zero ( p val u e < 0.05), a significant increasing trend (positive slope value) or decreasing trend (negative slope value) was determined for the time series data. Because seasonal variation has been shown to mask long term temporal trends in water quality variables (Hutchi nson 1957) and also in Florida lakes (Brown et al. 1999), polynomial fits were used (Burns et al. 1999) to determine whether season significantly attributed to the variance in each trophic state variable when the monthly

PAGE 27

27 data were examined. The residuals ( i.e., the difference between the actual value and the value as estimated by the polynomial fit ) were used as the dependent variable and plotted over the period of record, thereby removing the variance due to seasonality. Linear regression analysis was comp leted using the residuals from the polynomial fit over the period of record. If there was no significant polynomial fit for the monthly data or the variance in the data were not attributed to season, then linear regression analysis was completed using monthly means as the dependent variable. Most computer software packages have the ability to conduct linear regression analysis and most ecologists are familiar with the statistical procedures and interpretation of the results. Simple least squares linear regression analysis of the trans formed monthly data detected significant increasing monotonic trends in 63% (17 of the 27 Florida lakes) for total phosphorus, 55% (15 lakes) for total nitrogen, and 44% (12 lakes) for chlorophyll concentrations Decreasing monotonic trends were shown for 48% (9 lakes) for water clarity measurements (Table 2 3 ). The coefficients of determination ( R2) from the linear regression models of the monthly data were 0.65 or less indicating the models were not predictively powerful ( Prairie 1996). Although linear regression analysis detected significantly monotonic trends among the monthly times series data for many of the examined lak es (Table 2 4) the trend relationships were weak for TP (mean R2 value = 0. 22 range 0.01 to 0.62), TN (mean R2 = 0.19, range 0.02 to 0.57), CHL (mean R2 = 0.13, range 0.02 to 0.50), and SD (mean R2 = 0.12, range 0.01 to 0.55). On a percentage basis, examination of monthly data among the population of lakes showed 89%, 92%, 89%, and 96% of the linear models regress ing TP, TN, CHL, and SD over the examined period of record, had R2 values 0.65 or less suggesting a small

PAGE 28

28 percentage of the population, per each trophic state variable, experienced decadal scale trends. Simple least squa res linear regression analysis of the transformed annual data detected significant increasing monotonic trends for total phosphorus for 43% (12 lakes) of the Florida lakes, 40% (11 lakes) for total nitrogen, and 26% (7 lakes) for chlorophyll concentrations Significant monotonic decreasing trends were observed for 37% (7 lakes) for water clarity measurements (Table 2 3 ). The significant monotonic trends detected by linear regression analysis using the annual mean data (Table 25) were weak yet examination of annual data explained more of the variance in each trophic state variable compared to examination of monthly data (i.e., TP mean R2 value = 0. 34, range 0.0 3 to 0. 86; TN mean R2 = 0.31, range 0.04 to 0. 81; CHL mean R2 = 0.28, range 0.0 4 to 0. 78; and SD mean R2 = 0.2 0 range 0.0 1 to 0. 65) A small percentage of the population of lakes, however, experienced decadal scale trends as 89%, 92%, 89%, and 92% of the linear models regressing annual mean TP, TN, CHL, and SD data over the examined t ime record had R2 values 0.65 (Table 25) Interestingly, the lakes with 65% or more of the variance in the annual mean trophic state variable being explained by year had visually identifiable linear trends as illustrated by total phosphorus concentratio ns in Lake Lorraine (Fig ure 21 A ). The lakes wit h a R2 of 0.65 or less, generally had visually identifiable trends or no visually evident trend as illustrated by Little Orange Lake (Fig ure 2 1 B). Kendall Tau The Kendall given trophic state variable and time (i.e., year). The nonparametric Kendall Tau is a

PAGE 29

29 tistic al dependence Equation 22 as outlined by Kendall (1938) = ( ) ( ) ( ) ( 2 2) With: = Kendall Tau correlation coefficient (range concordant pairs = (x,y) pairs from examined X and Y variables where the ranks for both elements agree discordant pairs = (x, y) pairs from examined X and Y variables where the ranks for the elements disagree n = the number of observations n ( n 1) = the total number of pairs The Kendall Tau examined the number of pairs in different orders in the two rankings (Kendall 1938). The tau coefficient equaled 1 if the two rankings were in the same order and tau equaled 1 if the two rankings were inverted. Theref ore, a positive tau value indicated a significant increase and a negative tau value indicated a significant decrease over the available record of data. The Kendall Tau analysis offers a straightforward method yielding easy interpretation of results. Most c omputer programs provide the abil ity to complete this analysis. The Kendall Tau evaluation of annual mean data (Table 26) detected significant increasing monotonic trends for 37% (10 lakes) of the lakes for total phosphorus, 44% (12 lakes) for total nitro gen, and 22% (6 lakes) for chlorophyll concentrations Significant decreasing monotonic trends were detected in 31% (6 lakes) of the lakes for water clarity measurements The percentages of the population of 27 Florida lakes with

PAGE 30

30 significant trends detected by Kendall Tau analysis were similar to the results of linear regression (Table 23 ). Time Series M odeling Time series models have histor ically been used in economics, stock market analysis, in the social sciences, and more recently in the aquatic sciences. Time series model analysis can be divided into two categories; harmonic analysis and regression analysis. Regression time series model analysis, specifically the class of stochastic process models provided by the autoregression integrated moving aver age model (Box and Jenkins 1976) was used to detect long term trends in the trophic state variables. The time series models are useful to detect trends because a guiding principal of the Box and Jenkins (1976) approach is parsimony and consequently environmental variables can be modeled as a probabilistic function of past inputs (i.e., random variability) and outputs (i.e., time series observation) ( see McCleary and Hay 1980). The major components of the time series models used in analysis account for variance due to trend, season, and residual error, Equation 23 (Worrall and Burt 1998). Ytime = trend + seasonal variation + residual (2 3) These components of variance are analyzed by the use of the ARMA/ ARIMA time series model to understand the correlation between successive observations to help describe the evolution of the process through time ( see Chatfield 2004). Times series models were completed using the time series package in JMP, version 8.0. There were two important parameters, autocorrelation function (ACF) and the partial autocorrelation function (Partial ACF), used in time series model analysis. The ACF described the correlation between all pairs of observations in the time series with respect to time. The Partial ACF described the extent of the correlations between

PAGE 31

31 all successive pairs of observations. Both correlations, the ACF and the Partial ACF, are lagged correlations meaning the correlations between time series observations were shifted in time relative to one another. Lagged correlation was used in this paper because the examined trophic state variables may have delayed responses over time. The concentration of nutrients, for example, may depend on amounts of precipitation that have occurred over preceding years. All examined data were lagged by 1 unit (lag 1) across each unit of time (i.e., month or year), Equation 24. = ( ) ( ) ( ) / ( 2 4) With: = autocorrelation at lag 1 (y1, y2), (y2, y3), (y3, y4)( yn 1, yn) = pairs of time series data with n observations When examining annual data, for example, the ACF at a lag of 1 would describe the annual phosphorus concentration related to the previous annual phosphorus concentration. The phosphorus concentration at a lag of 10 (i.e., 11th year of record) related to the relationships established at other lag times would encompass the correlations between all the successive years and up to the 11th year. T ime series models require stationary data before model building can begin (Box and Jenkins 1970). The examined trophic state data for the individual lake were determined to be of a stationary, random process. Random processes are defined as random variables, where each point is unique at a given point in time, meaning the random process reflects properties of random variables (e.g., mean and variance). If the statistical properties of the random process do not significantly vary with time, then the process is stationary. I f the process is nonstationary the random variables significantly

PAGE 32

32 vary with time; the majority of environmental and aquatic data are nonstationary. For the examined trophic state variable data, t he AugmentedDickey Fuller test (ADF) was used to determine whether the data were stationary. If the data (ytime) were nonstationary, a technique called differencing was used to remove the variance (i.e., variance due to season and/or patterns of change) making the data stationary. The removal of variance (i.e., differencing) was completed by calculating the difference time = ytime ytime 1,). The integrated (I) term in the time series model denoted the data were differenced (e.g., AR IMA) versus an ARMA model denoted the data were stationary and differencing was not necessary. The time series models were built and selected using the general principal of Box and Jenkings (1976); model parameter estimation, model identification, and diagnostic checks of the residuals of the fitted models. The autoregressive (AR) and the moving average (MA) terms were estimated from the autocorrelation function (ACF) and the partial autocorrelation function (Partial ACF) The AR and MA terms were identifi ed from the ACF and Partial ACF plots based on visual assessment over the lagged periods of time and also by determination of the last lag term outsi de the associated 95% confidence interval. The last lag term outside of the 95% denoted the lag value at which the data were statistically significant and the point at which the data were no longer dependent on past values. Using significant ACF and Partial ACF lag terms juxtaposed with the visual pattern in the lag values (e.g., sine wave, exponential decreas e, or strong initial peak(s)), the time series model param eters were estimated (Figure 2 2 ). The Akaikes Information Criterion (AIC) was used as the model selection criteria to

PAGE 33

33 determine the best fit model for the data series. To ensure the data were best represented by the selected model, anywhere from 50 to 80 model variants were completed. The models were compared using the AIC value. The model with the AIC value closest to zero was selected as the best fit model (Akaike 1974). The selected model was verified as a good fit by further examination of the residuals and testing for white noise using the Bartletts Kolmogorov Smirnov test. The selected time series model; therefore, had no pattern in the residuals or statis tically significant white noise over the examined period of record. The selected time series model tested the null hypothesis that the variance of the given trophic state variable over the examined period of record wa s equal to zero. If significant variance was detected over time (i.e., the p valu e of the time series model was < 0.05), then the data showed significant change over time. Estimates of trend were determined by the value of the constant estimate term generated from the time series model (similar to the slope estimate for linear regression analysis). A positive constant estimate indicated an increasing trend and a negative constant estimate indicated a decreasing trend. Unlike linear regression models, time series models can detect a significant change in variance, but the constant estimate may equal zero indicating no directional, monotonic trend. Seasonal time series models were used when monthly data were examined. Autoregressive terms were added at lag 1 and successively at lag s of 12 to reflect periodic movement of season in the time series models T he lag 1 term accounted for the deviation of the current month from the previous month, while the lag 12 term accounted for the deviation of the current month from that of the same month in the

PAGE 34

34 previous year. The ARMA, ARIMA, and seas onal ARMA/ARIMA time series models are statistically complex and require much time and patience. There are computer packages that offer all the necessary tools to complete time series model analysis. In addition, if it is necessary to use multiple computer packages the user will need to understand each program requiring additional time. Time series modeling of the transformed monthly data for the population of 27 Florida lakes (Table 27) detected significant increasing trends in 4% (1 lake) of the lakes f or total phosphorus, 15% (4 lakes) for total nitrogen, and 11% (2 lakes) for chlorophyll concentrations Significant decreasing trends were shown for 11% (2 lakes) for water clarity measurements (Table 27 ). Time series modeling of transformed annual means for the population of 27 Florida lakes (Table 28) detected increasing trends in 15% (4 lakes) of the lakes for total phosphorus, 7% (2 lakes) for total nitrogen, 15% (4 lakes) for chlorophyll concentrations and decreasing trends in 16% (3 lakes) for wat er clarity measurements The proportion of lakes showing significant trends by time series analysis was greater than expected by chance (i.e., probability of 0.05) for each trophic state variable, but the overall percentage of population of lakes with significant trends was much less than the number of trends detected by linear regression, if an R2 predictive power. (Table 2 8) Time series analysis did not detect significant trends for some Florida lakes, but estimated signi ficant change in the variance over the examined time period of record (Figure 2 3). These lakes were categorized with the lakes that had no significant trends when summarizing for the examined population of Florida lakes (Table 2 3 ).

PAGE 35

35 Specifically, f or the monthly data, trophic state variable change was detected in 3 lakes (Table 27) F or the annual data there were 5, 8, 9, and 7 lakes that showed significant change for total phosphorus, total nitrogen, chlorophyll, and water clarity measurements (Table 2 8) Linear regression analysis detected no significant trend for all of the lakes where ARMA/ARIMA time series models detected significant change in the variance. Alternative A pproach Prairie (1996) reiterated a limitation of the predictive power of lin ear regression analysis was the influence of the number of samples on determination of a significant relationship. Prairie (1996) suggested using intervals, determined by the intersection of the linear regression model line and the associated 95% confidenc e intervals (Figure 2 4 ) to decrease the number of samples thereby increasing the predictive power of the linear regression analysis. The empirical derivation of the number of classes was related to the R2 value, Equation 25, where an R2 value of 0.65 a nd greater providing a predictively powerful linear regression model (Figure 2 4 ). NC = ( 2 5) With: NC = the number of classes 1.32 = the t value for a bivariate regression at a p value of 0.05 R2 = the coefficient of determination for a bivariate regression Intervals of six classes (NC= 6 mean values) were used for the trophic state data as Bryhn and Dimberg (2011) demonstrated that aquatic environmental data have the highest R2 values when divided into 6 classes. Linear regression analysis was then completed across the 6 calculated mean values providing an estimation of trend that

PAGE 36

36 was predictively power ful (Prairie 1996) and statistically meaningful (Bryhn and Dimberg 2011). The proposed alternative approach, modified linear regression analysis, that included the suggestions of Prairie (1996) and Byrhn and Dimberg (2011) detected significant increasing monotonic trends in 26% (7 lakes) of the Florida lakes for total phosphorus, 26% (7 lakes) for total nitrogen, and 19% (5 lakes) for chlorophyll. Decreasing monotonic trends were shown in 37% (7 lakes) for water clarity measurements (Table 2 3 ). The results of this proposed modified linear regression analysis were more similar to the results of tr end detection by either time series model analysis than those obtained by linear regression or Kendall Tau analyses for the examined population of 27 Florida lakes (Table 2 3 ). Discussion It is disconcerting that the use of difference statistical methods provided different results when used to evaluate the same aquatic time series data. For instance, Floridas human population has grown from about 9.7 million people in 1980 to over 19 million in 2010, which r aises concerns about impact of anthropogenic sources of pollution on lake water quality. The results from linear regression or Kendall Tau analyses would support the statement that population growth has adversely affected water quality as measured by the t rophic state variables in the examined Florida lakes over the past 20 plus years (Table 2 3 ). Similar results, however, were not obtained by the use of time series modeling (Table 2 3 ). Rather, time series modeling showed trends of degradation in the trophic state variables for only a small proportion of the examined Florida lakes The important point is that depending on the method of statistical analysis used, different

PAGE 37

37 conclusions could be reached and these conclusions may have divergent implications for science, management, and society. Given that different conclusions could be reached depending on the statistical analyses used, ecologists need to explicitly define the experimental unit and scale of analysis to explain the results in terms of the var iance when addressing aquatic systems (Duarte and Kalff 1990). For example, variance component analysis of the examined population of Florida lakes indicated that laketo lake differences accounted for the majority of the variance in the trophic state vari ables When the variance within the individual lakes was examined, yearly (i.e., total phosphorus and total nitrogen) or monthly (i.e., chlorophyll and water clarity) differences accounted for the majority of the variance. Analogous results have been shown for a different population of Florida lakes (Brown et al. 2000). Thus, if a research objective is to address patterns of change or trend across a geographic region, like the State of Florida, inclusion of more sampling in the study would best account for temporal and spatial variance in the examined trophic state variables Hkanson and Duarte (2008) recognized station and laboratory error as major contributors to variance in the study of single ecological units. For studies completed within a lake, station and laboratory error explain more variance in the trophic state variables and should is considered in single ecological unit studies. The statistical assumptions of a given statistical analysis are additionally important to consider when examining aquat ic data and help to understand the different results obtained by the use of different statistical analyses. A major statistical assumption of liner regression analysis, and also for the rankings of the Kendall Tau analysis, is a linear r elationship between the dependent and independent variabl e. In

PAGE 38

38 this chapter linear regression and Kendall Tau analyses provided a reliable assessment of trend in data that exhibited a visually identifiable linear relationship, as was the case with total phosphorus concentrations in Lake Lorraine (Fig ure 21A ). However, when nonlinear trends were present linear regression and Kendall Tau analyses did not provide an appropriate assessment, as was the case with total phosphorus concentrations Little Orange Lake (Fig ure 21 B ). Another consideration is that linear regression and Kendall Tau analyses only account for variance in the dependent variable, whi le time series analysis account for variance i n both the dependent and independent variable. If nonlinear patterns were present, then variance in both the dependent and independent variable should be included so the number of trends is not overestimated. Due to the treatment of the dependent and independent variable and the number of lakes with nonlinear relationships in the examined dataset, linear regression and Kendall Tau analysis may have overestimated the number of trends. Furthermore, the number of samples influences statistical significant determination of a linear trend (Prairie 1996). The examined environmental data included a large number of samples ( n n annual mean data) of which linear regression and Kendall Tau analyses detected many statistically significant trends. But, although statistically significant, the relations hip of many trends was weak with data scattered from the trend line and low coefficients of determination (e.g., R2 = 0. 1 0 ). The detection of trends by time series analysis may have been influenced by the violation of the parametric statistic assumption of lack of serial correlation or in dependence of the error terms. Such a violation is common with environmental data

PAGE 39

39 and can cause the standard errors to be underestimated and ultimately leads to false conclusions because the t or F test may be incorrect (Abaurrear et al. 2011). Tim e series model analysis includes methods to remove serial correlation (e.g., the differencing technique), but the removal of serial correlation changes the response variable (Nickerson and Madsen 2005). For example, lagged values, which are a consequence of the removal of serial correlation, make the response variable a function of the past. As response variables are generally dependent on past values no change or trend would be detected when indeed a change or trend may have been present. In such cases, time series analysis may have underestimated the number of trends. Understanding the differences in linear regression, Kendall Tau, and time series model analyses is not simplistic and there are most likely additional considerations Statistical methods are tools to help guide ecologists. The alternative, modified linear regression analysis approach, which combined statistical methods outlined by Prairie (1996) and Bryhn and Dimberg (2011), resulted in a percentage of the population of lakes with increasing and decreasing trends similar to that obtained by t he use of statistically robust, yet complex ARMA/ARIMA time series modeling. The alternative approach offers a statistical took that provides predictively powerful results yet can be simply understood leading to increased application by more ecologists, including nonstatisticians (Murtaugh 2007). Comments and Recommendations Although it is important to integrate predictively powerful statistical approaches to detect trends in aquatic time series data, it is also important to detach from the statisti cs every now and again. Whether statistically complex or simplistic, a statistical method may not detect a change or trend in ecological time series data. Ther e may be certain

PAGE 40

40 events (i.e., stochastic, climatic, or ant hropogenic) that are recognized to drast ically impact an aquatic system. For example, time series analysis did not detect a significant trend in total phosphorus concentrations in Little Lake Santa Fe, but visual examination of the se data showed an order of magnitude change in total phosphorus concentrations (Fig ure 23 ). In Little Lake Santa Fe, an order of magnitude increase in total phosphorus concentrations was due to a large forest fire, a stoc hastic event that an ecologist or limnologist would recognize to have substantial impacts on an aquatic system ( Ruiz Bernard 2012). Therefore, despite the numerous statistical tools available to detect trends in environmental data, it is recommended to plo t and examine the data, then explore the use of statistics.

PAGE 41

41 Table 21. S ummary statistics (i.e., mean median, minimum, maximum, an d coefficient of variation) for untransformed total phosphorus (g/L), total nitrogen (g/L), chlorophyll concentrations (g /L) and water clarity measurements (m) among annual mean data for the examined population of Florida lakes. Trophic State Variable N Lakes Mean Median Minimum Maximum Coefficient of Variation Total Phosphorus 27 33 18 3.7 125 42% Total Nitrogen 27 1139 701 107.0 3729 28% Chlorophyll 27 31 9 1.6 169 71% Water Clarity 19 1.5 1.4 0.3 3.1 32%

PAGE 42

42 Table 22. R esults of variance component analysis and the percent of variance attributed to laketo lake differences, year to year differences, monthto month differences, and residual error (includes stationto station and laboratory differences) using monthly data f or the population of 27 Florida lakes. Within the individual 27 Florida lakes, the mean percent of variance attributed to year to year differences, monthto month differences, stationto station differences, and residual error (laboratory differences) usin g monthly data are presented. Trophic State Variable % variance lake to lake % variance year to year % variance month to month % variance station to station % variance residual error Population 27 lakes Total Phosphorus 82 9 6 3 Total Nitrogen 86 7 5 2 Chlorophyll 82 6 10 2 Water Clarity 83 8 8 1 Individual 27 lakes Total Phosphorus 45 39 6 10 Total Nitrogen 51 37 6 6 Chlorophyll 55 33 4 8 Water Clarity 48 46 2 4

PAGE 43

43 Table 23 P ercentage of the population of lakes with monotonic increasing trends, monotonic decreasing trends, and no trends over 20plus years detected by the use of linear regression models, Kendall Tau analysis, ARMA/ARIMA time series models, and the alternative approach following Prairie (1996) and Byrhn and Dimberg (2011). Monthly data and annual data were evaluated for total phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements for the examined population of Florida lakes. Trophic State Variable N Increasing Trend (%) Decreasing Trend (%) No Trend (%) Monthly Data Linear Regression Model Total Phosphorus 27 63 22 15 Total Nitrogen 27 55 30 15 Total Chlorophyll 27 44 38 18 Water Clarity 19 13 48 39 Time Series Model Total Phosphorus 27 4 7 89 Total Nitrogen 27 15 4 81 Total Chlorophyll 27 11 0 89 Water Clarity 19 0 11 89 Annual Data Linear Regression Model Total Phosphorus 27 44 19 37 Total Nitrogen 27 40 4 56 Total Chlorophyll 27 26 26 48 Water Clarity 19 5 37 58 Kendall Tau Total Phosphorus 27 37 19 44 Total Nitrogen 27 44 7 49 Total Chlorophyll 27 22 22 56 Water Clarity 19 5 3 7 58 Time Series Model Total Phosphorus 27 15 11 74 Total Nitrogen 27 7 4 89 Total Chlorophyll 27 15 4 81 Water Clarity 19 0 16 84 Practical Approach Total Phosphorus 27 26 7 67 Total Nitrogen 27 26 4 70 Total Chlorophyll 27 19 11 70 Water Clarity 19 0 3 2 6 8 *The percentage of the population of lakes reported with no trend for time series analysis included lakes with detection of significant change, but no significant monotonic trend.

PAGE 44

44 Table 2 4 Linear regression analysis detection of a significant monotonic trend ( and coefficient of determination ( R) of the monthly time series logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes. Linear regr ession analysis completed using the residual s of the best polynomial fit to remove variance due to season, is denoted after the slope value County Lake L 10 TP S lope L 10 TP R L 10 TN Slope L 10 TN R L 10 CHL Slope L 10 CHL R L 10 SD Slope L 10 SD R Alachua Alto 0.22 0.52 0.10 0.35 Alachua Little Orange 0.25 0.02 0.16 0.12 Alachua Little Santa Fe 0.43 0.35 0.06 0.26 Alachua Santa Fe 0.43 0.36 0.27 0.34 Alachua Wauberg 0.23 0.40 0.20 0.38 Hillsborough Brant 0 0 0.04 0 0.0001 0.01 Hillsborough Magdalene 0.07 0.04 0 0 0 0.00 Lake Beauclaire 0.57 0.05 0.03 0 0.00 Lake Crooked 0.19 0.10 0.08 0.22 Lake Dora East 0.45 0.08 0.09 Lake Dora West 0.30 0.09 0.08 Lake Grasshopper 0.24 0.05 0.06 Lake Harris 0.06 0.07 0.12 0.09 Lake Lorraine 0.62 0.57 0.50 Lake Sellers 0.33 0.30 0.36 Marion Charles 0.13 0.22 0.02 0.001 0.00 Marion Deerback 0 0.01 0 0 0.05 0.04 Marion Eaton 0.03 0.02 0 0 0 0.00 Marion Halfmoon 0 0 0.37 0.11 0.03 Orange Georgia 0.17 0.12 0.10 0.27 Orange Giles 0.01 0 0 0.02 0.00 Orange Ola 0.06 0.23 0.13 Orange Sarah 0.06 0.08 0.38 Putnam Como 0.26 0.35 0.07 Putnam Higgenbotham 0.02 0 0 0.23 0 0.01

PAGE 45

45 Table 24 Continued Co unty Lake L 10 TP Slope L 10 TP R L 10 TN Slope L 10 TN R L 10 CHL Slope L 10 CHL R L 10 SD Slope L 10 SD R Putnam Star 0.22 0.3 0 0 0.11 Putnam Winnott 0.5 0.48 0.22 0.55

PAGE 46

46 Table 2 5. Linear regression analysis detection of a significant monotonic trend ( and coefficient of determination ( R) of the annual mean time series logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes. County Lake L 10 TP Slope L 10 TP R L 10 TN Slope L 10 TN R L 10 CHL Slope L 10 CHL R L 10 SD Slope L 10 SD R Alachua Alto 0.47 0.75 0.22 0.56 Alachua Little Orange 0.32 0.002 0.04 0.34 0.26 Alachua Little Santa Fe 0.62 0.52 0.010 0.18 0.43 Alachua Santa Fe 0.64 0.52 0.45 0.55 Alachua Wauberg 0.35 0.54 0.45 0.005 0.19 Hillsborough Brant 0.000 0 0.005 0.06 0.002 0 0.002 0.01 Hillsborough Magdalene 0.005 0.13 0.002 0.10 0.000 0 0.001 0.01 Lake Beauclaire 0.81 0.004 0.10 0.006 0.10 0.000 0.00 Lake Crooked 0.37 0.007 0.14 0.26 0.46 Lake Dora East 0.58 0.004 0.13 0.009 0.16 Lake Dora West 0.39 0.004 0.15 0.008 0.16 Lake Grasshopper 0.39 0.013 0.09 0.011 0.12 Lake Harris 0.004 0.17 0.004 0.14 0.41 0.22 Lake Lorraine 0.86 0.81 0.78 Lake Sellers 0.56 0.55 0.65 Marion Charles 0.22 0.39 0.100 0.04 0.002 0.01 Marion Deerback 0.002 0.03 0.001 0 0.007 0.14 0.004 0.10 Marion Eaton 0.007 0.10 0.003 0.04 0.001 0 0.000 0.00 Marion Halfmoon 0.001 0 0.50 0.43 0.003 0.08 Orange Georgia 0.34 0.005 0.23 0.28 0.44 Orange Giles 0.002 0.05 0.001 0 0.007 0.11 0.006 0.15 Orange Ola 0.003 0.14 0.52 0.40 Orange Sarah 0.005 0.16 0.004 0.18 0.65 Putnam Como 0.42 0.57 0.010 0.15 Putnam Higgenbotham 0.003 0.10 0.000 0 0.56 0.003 0.03

PAGE 47

47 Table 25. Continued County Lake L 10 TP Slope L 10 TP R L 10 TN Slope L 10 TN R L 10 CHL Slope L 10 CHL R L 10 SD Slope L 10 SD R Putnam Star 0.31 0.55 0.002 0 0.24 Putnam Winnott 0.018 0.71 0.011 0.63 0.018 0.44 0.021 0.65

PAGE 48

48 Table 2 6. K endall Tau analysis detection of significant monotonic trends ( ) with the tau value for annual mean total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity (SD) data for the 27 Florida lakes. County Lake TP Tau TN Tau CHL Tau SD Tau Alachua Alto 0.28 Alachua Little Orange 0.28 0.04 0.3 Alachua Little Santa Fe Alachua Santa Fe 0.6 0 Alachua Wauberg 0.23 Hillsborough Brant 0.01 0.21 0.14 0.08 Hillsborough Magdalene 0.23 0.28 0.04 0.08 Lake Beauclaire 0.18 0.18 0.03 Lake Crooked Lake Dora East 0.20 0.26 Lake Dora West 0.18 0.22 Lake Grasshopper 0.14 0.19 Lake Harris 0.21 0.24 0.23 Lake Lorraine Lake Sellers Marion Charles 0.29 0.19 0.03 Marion Deerback 0.06 0.03 0.18 0.15 Marion Eaton 0.21 0.14 0.03 0.01 Marion Halfmoon 0.04 0.11 Orange Georgia 0.12 Orange Giles 0.10 0.06 0.18 0.29 Orange Ola 0.29 0.45 Orange Sarah 0.23 0.28 Putnam Como 0.1 Putnam Higgenbotham 0.09 0.01 0.12 Putnam Star 0.04 Putnam Winnott

PAGE 49

49 Table 27 T ) denotes significant change and denotes significant change and monotonic trend), the Akaikes Information Criterion (AIC) value, the time lag corresponding to the significant autocorrelation (AC), and the coefficient of determination (R2) using monthly logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes County Lake L 10 TP Mode l L 10 TP AIC L 10 TP AC L 10 TP R 2 Alachua Alto ARIMA(1,1,1) 432 48 0.49 Alachua Little Orange Seasonal ARIMA(2,1,2)(0,1,0)12 268 12 0.21 Alachua Little Santa Fe Seasonal ARIMA(0,2,0)(1,1,1)12 5 15 0.36 Alachua Santa Fe Seasonal ARIMA(0,1,0)(0,1,0)12 285 15 0.25 Alachua Wauberg 308 12 0.38 Hillsborough Brant Seasonal ARIMA(1,1,1)(0,1,0)12 384 9 0.68 Hillsborough Magdalene Seasonal ARIMA(0,0,0)(1,1,0)12 334 13 0.21 Lake Beauclaire 252 24 0.12 Lake Crooked Seasonal ARIMA(0,1,0)(0,1,0)12 180 15 0.38 Lake Dora East 340 25 0.26 Lake Dora West Seasonal ARIMA(0,2,0)(0,1,0)12 245 36 0.03 Lake Grasshopper Seasonal ARIMA(0,1,0)(1,1,0)12 8 10 0.50 Lake Harris ARIMA(2,2,2) 503 13 0.30 Lake Lorraine Seasonal ARIMA(0,1,0)(0,1,0)12 155 18 0.26 Lake Sellers Seasonal ARIMA(1,1,1,)(0,1,0)12 42 22 0.54 Marion Charles Seasonal ARIMA(0,1,0)(0,1,0)12 161 12 0.00 Marion Deerback ARI(1,1) 397 4 0.00 Marion Eaton Seasonal ARIMA(1,1,0)(0,1,0)12 107 11 0.27 Marion Halfmoon Seasonal ARIMA(1,0,1)(0,1,0)12 406 13 0.23 Orange Georgia Seasonal ARIMA(1,0,1)(0,1,0)12 356 12 0.12 Orange Giles Seasonal ARIMA(0,0,0)(1,1,1)12 327 30 0.16 Orange Ola Seasonal 545 7 0.02 Orange Sarah Seasonal ARIMA(2,1,0)(1,1,1)12 392 20 0.12 Putnam Como Seasonal ARIMA(0,1,0)(0,1,0)12 163 17 0.26

PAGE 50

50 Table 27 C ontinued County Lake L 10 TP Model L 10 TP AIC L 10 TP AC L 10 TP R 2 Putnam Higgenbotham IMA(1,1) 346 24 0.33 Putnam Star Seasonal ARIMA(1,1,1)(0,1,0)12 384 14 0.22 Putnam Winnott Seasonal ARIMA(0,0,0)(0,1,0)12 273 18 0.40 County Lake L 10 TN Model L 10 TN AIC L 10 TN AC L 10 TN R 2 Alachua Alto Seasonal ARIMA(1,1,1)(0,1,0)12 634 25 0.49 Alachua Little Orange Seasonal ARIMA(1,1,1)(0,1,0)12 552 4 0.21 Alachua Little Santa Fe Seasonal ARIMA(0,1,0)(0,1,0)12 413 17 0.36 Alachua Santa Fe 495 16 0.25 Alachua Wauberg 477 15 0.38 Hillsborough Brant Seasonal ARIMA(1,1,0)(0,1,0)12 531 10 0.68 Hillsborough Magdalene Seasonal ARIMA(1,1,1)(1,1,0)12 654 6 0.21 Lake Beauclaire Seasonal ARIMA(0,2,0)(1,1,1)12 455 10 0.12 Lake Crooked Seasonal ARIMA(0,1,0)(0,1,0)12 457 13 0.38 Lake Dora East I(2) 619 10 0.26 Lake Dora West 545 9 0.03 Lake Grasshopper Seasonal ARIMA(0,2,0)(1,1,0)12 11 10 0.50 Lake Harris Seasonal ARIMA(1,1,1)(0,1,0)12 597 12 0.30 Lake Lorraine Seasonal ARIMA(0,0,0)(0,1,0)12 464 18 0.26 Lake Sellers IMA(1,1) 22 15 0.54 Marion Charles Seasonal ARIMA(0,0,0)(1,1,1)12 359 10 0.00 Marion Deerback Seasonal 632 5 0.00 Marion Eaton Seasonal ARIMA(0,1,0)(0,1,0)12 314 6 0.27 Marion Halfmoon 520 12 0.23 Orange Georgia Seasonal ARIMA(1,1,1 )(0,1,0)12 515 11 0.12 Orange Giles Seasonal ARIMA(1,1,1)(1,1,1)12 236 27 0.16 Orange Ola Seasonal ARIMA(0,1,0)(0,1,0)12 590 23 0.02 Orange Sarah Seasonal ARIMA(1,1,0)(1,1,1)12 529 21 0.12 Putnam Como Seasonal ARIMA(0,0,0)(0,1,0)12 199 23 0.26 Putnam Higgenbotham Seasonal ARIMA(1,1,0)(0,1,0)12 478 4 0.33 Putnam Star Seasonal ARIMA(0,1,1)(0,1,0)12 459 15 0.22

PAGE 51

51 Table 27. Continued County Lake L 10 TN Model L 10 TN AIC L 10 TN AC L 10 TN R 2 Putnam Winnott Seasonal ARIMA(1,1,1)(0,1,0)12 551 19 0.40 County Lake L 10 CHL Model L 10 CHL AIC L 10 CHL AC L 10 CHL R 2 Alachua Alto Seasonal ARIMA(0,0,0)(1,1,0)12 91 48 0.07 Alachua Little Orange 6 14 0.43 Alachua Little Santa Fe Seasonal ARIMA(1,1,1)(1,1,1)12 16 14 0.26 Alachua Santa Fe Seasonal ARIMA(0,1,0)(0,1,0)12 78 12 0.26 Alachua Wauberg 165 24 0.25 Hillsborough Brant I(1) 10 12 0.53 Hillsborough Magdalene Seasonal ARIMA(0,1,0)(0,1,0)12 146 12 0.10 Lake Beauclaire Seasonal ARIMA(0,1,1)(0,1,0)12 181 12 0.28 Lake Crooked ARI(1,1) 11 13 0.24 Lake Dora East Seasonal ARIMA(0,2,0)(1,1,1)12 210 12 0.23 Lake Dora West Seasonal ARIMA(0,1,0)(0,1,0)12 273 12 0.25 Lake Grasshopper Seasonal ARIMA(0,1,0)(0,1,0)12 24 12 0.09 Lake Harris ARI(1,1) 72 36 0.11 Lake Lorraine IMA(1,1) 10 16 0.58 Lake Sellers Seasonal ARIMA(1,1,0)(0,1,0)12 50 24 0.13 Marion Charles ARIMA(1,1,1) 187 11 0.42 Marion Deerback ARI(2,1) 112 12 0.15 Marion Eaton ARIMA(2,1,1) 151 42 0.27 Marion Halfmoon ARI(1,1) 209 24 0.08 Orange Georgia 58 24 0.32 Orange Giles ARIMA(1,1,1) 10 1 0.20 Orange Ola 97 25 0.13 Orange Sarah ARI(1,1) 45 25 0.50 Putnam Como Seasonal ARIMA(1,1,0)(0,1,0)12 50 13 0.06 Putnam Higgenbotham ARI(1,1) 110 25 0.18 Putnam Star ARI(2,1) 69 12 0.10 Putnam Winnott 182 12 0.48

PAGE 52

52 Table 27. Continued County Lake L 10 SD Model L 10 SD AIC L 10 SD AC L 10 SD R 2 Alachua Alto Seasonal ARIMA(0,0,0)(0,1,0)12 217 36 0.12 Alachua Little Orange Seasonal ARIMA(1,0,1)(0,1,0)12 349 7 0.10 Alachua Little Santa Fe Seasonal ARIMA(0,0,0)(0,1,1)12 270 13 0.16 Alachua Santa Fe Seasonal ARIMA(0,0,0)(0,1,0)12 349 13 0.02 Alachua Wauberg Seasonal ARIMA(0,1,0)(1,1,0)12 410 13 0.22 Hillsborough Brant Seasonal ARIMA(0,1,0)(0,1,0)12 371 12 0.16 Hillsborough Magdalene ARIMA(1,1,1) 703 12 0.64 Lake Beauclaire Seasonal ARIMA(0,1,0)(1,1,1)12 366 11 0.12 Lake Crooked Seasonal ARIMA(1,1,0)(1,1,0)12 296 16 0.23 Lake Dora East Lake Dora West Lake Grasshopper Lake Harris Seasonal ARIMA(0,0,0)(1,1,0)12 135 24 0.01 Lake Lorraine Lake Sellers Marion Charles ARI(1,1) 159 5 0.30 Marion Deerback Seasonal ARIMA(0,1,0)(0,1,0)12 361 12 0.16 Marion Eaton Seasonal ARIMA(0,1,0)(0,1,0)12 3 4 0.13 Marion Halfmoon Seasonal ARIMA(0,1,1)(0,1,0)12 358 13 0.20 Orange Georgia Seasonal ARIMA(0,1,1)(0,1,0)12 493 12 0.32 Orange Giles ARI(1,1) 157 3 0.14 Orange Ola Orange Sarah Putnam Como Putnam Higgenbotham Seasonal ARIMA(0,1,0)(0,1,0)12 316 6 0.21 Putnam Star Seasonal ARIMA(0,0,0)(0,1,0)12 257 24 0.08 Putnam Winnott Seasonal ARIMA(0,2,0)(0,1,0)12 254 15 0.28

PAGE 53

53 Table 28 T denotes significant change and denotes significant change and monotonic trend), the Akaikes Information Criterion (AIC) value, the time lag corresponding to the significant autocorrelation (AC), and the coefficient of determination (R2) using annual mean logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes. County Lake L 10 TP Model L 10 TP AIC L 10 TP AC L 10 TP R 2 Alachua Alto I(1) 49 1 0.03 Alachua Little Orange 1 1 0.30 Alachua Little Santa Fe I(2) 16 1 0.46 Alachua Santa Fe ARI(2,2) 30 1 0.03 Alachua Wauberg 28 0 0.15 Hillsborough Brant ARI(1,1) 20 1 0.01 Hillsborough Magdalene ARIMA(2,2,1) 32 1 0.06 Lake Beauclaire 32 2 0.50 Lake Crooked ARIMA(2,2,2) 15 1 0.16 Lake Dora East 34 1 0.54 Lake Dora West ARIMA(2,2,2) 38 1 0.72 Lake Grasshopper 9 1 0.41 Lake Harris 53 0 0.10 Lake Lorraine 23 1 0.46 Lake Sellers ARIMA(2,2,2) 15 1 0.48 Marion Charles ARI(1,1) 9 1 0.08 Marion Deerback 46 0 0.17 Marion Eaton 10 1 0.29 Marion Halfmoon ARIMA(1,1,1) 48 1 0.13 Orange Georgia 38 1 0.04 Orange Giles ARIMA(1,1,1) 48 0 0.06 Orange Ola ARIMA(2,2,1) 44 1 0.02 Orange Sarah 42 1 0.23 Putnam Como 20 1 0.22 Putnam Higgenbotham 57 0 0.02

PAGE 54

54 Table 28 Continued County Lake L 10 TP Model L 10 TP AIC L 10 TP AC L 10 TP R 2 Putnam Star ARI(2,2) 18 1 0.08 Putnam Winnott ARIMA(2,2,1) 35 1 0.62 County Lake L 10 TN Model L 10 TN AIC L 10 TN AC L 10 TN R 2 Alachua Alto ARI(1,1) 83 2 0.77 Alachua Little Orange 57 0 0.06 Alachua Little Santa Fe 43 1 0.38 Alachua Santa Fe ARIMA(2,2,2) 47 1 0.48 Alachua Wauberg I(2) 44 1 0.41 Hillsborough Brant ARIMA(2,2,2) 18 1 0.08 Hillsborough Magdalene ARIMA(1,1,1) 67 0 0.04 Lake Beauclaire 48 1 0.40 Lake Crooked ARIMA(2,2,2) 43 1 0.66 Lake Dora East 50 1 0.19 Lake Dora West 53 1 0.20 Lake Grasshopper IMA(1,1) 2 1 0.21 Lake Harris IMA(1,1) 49 1 0.09 Lake Lorraine 48 2 0.42 Lake Sellers ARIMA(2,2,2) 1 1 0.58 Marion Charles ARMA(2,2) 32 1 0.14 Marion Deerback 81 0 0.09 Marion Eaton 33 0 0.11 Marion Halfmoon ARIMA(2,2,1) 39 1 0.20 Orange Georgia 51 1 0.20 Orange Giles 43 0 0.12 Orange Ola IMA(2,2) 59 1 0.20

PAGE 55

55 Table 28 Continued County Lake L 10 TN Model L 10 TN AIC L 10 TN AC L 10 TN R 2 Orange Sarah ARIMA(2,2,2) 48 1 0.39 Putnam Como IMA(1,1) 32 2 0.51 Putnam Higgenbotham 47 0 0.19 Putnam Star ARIMA(2,2,2) 40 1 0.25 Putnam Winnott ARIMA(2,2,1) 45 2 0.50 County Lake L 10 CHL Model L 10 CHL AIC L 10 CHL AC L 10 CHL R 2 Alachua Alto 30 0 0.05 Alachua Little Orange IMA(1,1) 1 1 0.03 Alachua Little Santa Fe 16 0 0.09 Alachua Santa Fe 8 1 0.16 Alachua Wauberg ARIMA(2,2,2) 17 1 0.02 Hillsborough Brant 1 1 0.24 Hillsborough Magdalene ARIMA(1,1,1) 22 1 0.07 Lake Beauclaire ARI(1,1) 19 1 0.01 Lake Crooked ARIMA(2,2,2) 5 1 0.23 Lake Dora East IMA(1,1) 21 1 0.01 Lake Dora West IMA(1,1) 27 1 0.02 Lake Grasshopper 10 1 0.42 Lake Harris ARI(2,2) 15 1 0.14 Lake Lorraine 4 2 0.43 Lake Sellers 15 2 0.47 Marion Charles 17 0 0.35 Marion Deerback 26 0 0.10 Marion Eaton 7 0 0.17 Marion Halfmoon ARIMA(2,2,2) 27 1 0.02

PAGE 56

56 Table 28 Continued County Lake L 10 CHL Model L 10 CHL AIC L 10 CHL AC L 10 CHL R 2 Orange Georgia 16 0 0.09 Orange Giles 23 0 0.12 Orange Ola ARIMA(2,2,2) 21 1 0.15 Orange Sarah I(2) 5 2 0.46 Putnam Como ARIMA(1,1,1) 16 1 0.14 Putnam Higgenbotham 20 0 0.04 Putnam Star 27 0 0.00 Putnam Winnott ARI(2,2) 6 1 0.03 County Lake L 10 SD Model L 10 SD AIC L 10 SD AC L 10 SD R 2 Alachua Alto 38 2 0.30 Alachua Little Orange 37 0 0.11 Alachua Little Santa Fe ARI(1,1) 30 1 0.12 Alachua Santa Fe 44 1 0.38 Alachua Wauberg ARIMA(2,2,2) 39 1 0.12 Hillsborough Brant ARIMA(1,1,1) 31 1 0.06 Hillsborough Magdalene ARIMA(2,2,1) 40 1 0.09 Lake Beauclaire 42 1 0.14 Lake Crooked ARIMA(2,2,1) 26 1 0.19 Lake Dora East Lake Dora West Lake Grasshopper 2 1 0.52 Lake Harris ARIMA(1,1,1) 18 0 0.04 Lake Lorraine Lake Sellers Marion Charles 17 0 0.20

PAGE 57

57 Table 28. Continued County Lake L 10 SD Model L 10 SD AIC L 10 SD AC L 10 SD R 2 Marion Deerback 42 1 0.37 Marion Eaton 9 0 0.14 Marion Halfmoon 34 1 0.03 Orange Georgia ARI(1,1) 41 1 0.03 Orange Giles ARIMA(1,1,1) 31 0 0.04 Orange Ola Orange Sarah Putnam Como Putnam Higgenbotham 35 0 0.01 Putnam Star ARIMA(2,1,2) 35 1 0.28 Putnam Winnott ARI(2,2) 20 1 0.37

PAGE 58

58 Table 29 Modified l inear regression analysis with six categories of annual mean data, to detect significant monotonic trends ( and coefficient of determination ( R) for logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD) data for the 27 Florida lakes. County Lake L 10 TP Slope L 10 TP R 2 L 10 TN Slope L 10 TN R 2 L 10 CHL Slope L 10 CHL R 2 L 10 SD Slope L 10 SD R 2 Alachua Alto 0.93 0.82 0.03 0.59 0.69 Alachua Little Orange 0.84 0 0.09 0.05 0.45 0.77 Alachua Little Santa Fe 0.67 0.71 0.73 0.74 Alachua Santa Fe 0.76 0.77 0.88 0.87 Alachua Wauberg 0.57 0.03 0.49 0.04 0.49 0.02 0.26 Hillsborough Brant 0 0 0.02 0.20 0 0 0 0.01 Hillsborough Magdalene 0.02 0.24 0.01 0.27 0 0 0.01 0.07 Lake Beauclaire 0.78 0.01 0.05 0.01 0.04 0 0.02 Lake Crooked 0.04 0.48 0.02 0.12 0.04 0.35 0.03 0.50 Lake Dora East 0.05 0.54 0.01 0.10 0.02 0.16 Lake Dora West 0.03 0.29 0.01 0.12 0.02 0.13 Lake Grasshopper 0.08 0.51 0.04 0.12 0.03 0.12 Lake Harris 0.01 0.23 0.01 0.24 0.05 0.52 0.02 0.20 Lake Lorraine 0.93 0.89 0.82 Lake Sellers 0.07 0.55 0.12 0.67 0.09 0.77 Marion Charles 0.04 0.22 0.03 0.65 0.01 0.01 0.01 0.06 Marion Deerback 0 0 0 0 0.02 0.46 0.01 0.04 Marion Eaton 0.03 0.14 0.01 0.07 0.01 0.02 0 0 Marion Halfmoon 0 0 0.77 0.70 0.01 0.04 Orange Georgia 0.03 0.64 0.02 0.49 0.03 0.43 0.81 Orange Giles 0 0.11 0 0.02 0.02 0.23 0.02 0.40 Orange Ola 0.01 0.31 0.02 0.66 0.71 Orange Sarah 0.01 0.05 0.01 0.24 0.1 0.49 Putnam Como 0.06 0.63 0.05 0.48 0.01 0.06 Putnam Higgenbotham 0.005 0.06 0 0 0.83 0.01 0.06 Putnam Star 0.90 0.92 0 0 0.03 0.58 Putnam Winnott 0.82 0.91 0.92 0.90

PAGE 59

59 Figure 21. Linear regression model analysis and the associated 95% confidence intervals for annual mean total phosphorus concentrations (g / L). A) Total phosphorus concentrations in Lake Lorraine located in Lake County, Florida (p < 0.0001, R2 = 0.77) B) Total phosphorus concentrations (g / L) in Little Orange Lake located in Alachua County, Florida (p = 0.03, R2 = 0.23). The Kendall Tau, ARMA/ARIMA time series model s, and the proposed alternative methods detected a significant trend in total phosphorus concentrations (g / L) in Lorraine Lake, but not in Little Orange Lake.

PAGE 60

60 Figure 22. Example of a time series plot of annual mean total phosphorus concentrations (g/L) for Lake Como (located in Putnam County, Florida) the corresponding autocorrelation function (ACF) plot, and the corresponding partial autocorrelation function (Partial ACF) plot against time with successively time units (years) lagged by one. The dotted lines on the ACF and PACF plot represent the upper and lower 95% confidence intervals. The statistically significant ACF and Partial ACF values along with the pattern of the lag terms were used to estimate the autocorrelation term (AR) and moving average (MA) terms of the time series model. The selected time series model for these data was A RIMA (1,1,1).

PAGE 61

61 Figure 22 Continued

PAGE 62

62 Figure 23 A nnual mean total phosphorus concentrations ( g/ L) in Little Lake Santa Fe located in Alachua County, Florida. Linear regression and Kendall Tau analyses detected significant increasing monotonic trends in total phosphorus ( g / L), while the time series model detected a significant change in total phosphorus concentrations ( g / L), but no signific ant trend over the examined 2 4 year record (19862009).

PAGE 63

63 Figure 24 N umber of classes determined (three intervals as pictured) using the bivariate linear regression model and the associated 95% confidence intervals (top figure). As the number of classes increases the coefficient of determination ( R2) values increase with a value of 0.65 noted as a pr edictively powerful linear regression model (bottom figure). Data and figures are from Prairie (1996)

PAGE 64

64 CHAPTER 3 D ECADAL SCALE TRENDS IN TROPHIC STATE VARIAB LES WITHIN A LARGE POPULATION OF FLORID A LAKES Background Long term data are invaluable to prov ide i mproved estimates of environmental changes and trends that offer a context in which to evaluate environmental management options (Stow et al. 1998; Hobbie 2003). Scientists, federal and state agencies, and policy makers recognize the importance of long term monitoring data and have consequently implemented several largescale lake monitoring programs (e.g., United States Geological Surveys National Water Quality Assessment programs, United States Department of Agricultures National Resources Inventory, and the United States Environmental Protection Agencys National Lakes Assessment Program). Many of these lake monitoring programs; however, do not meet the sampling frequency and duration suggested to best detect changes and trends in lake trophic state variables that represent the systems behavior (i.e., 6 consecutive years (Molot and Dillion 1991) to 12 years (Howden et al. 2011) to at least 20 years (Knowlton and Jones 2006)). As lake management decisions that target improvement of water quality or nutrient control are often based on trends, assessment of long term tr ophic state trends and understanding the magnitude of these trends are needed to facilitate advance ment of scientific research to develop the most appropriate management plans. The detection of trends has recently drawn much attention due to influences of mounting human pressur es and climatic drivers on lake trophic state variables (e.g., Carpenter et al. 1998; Williamson et al. 2009; Adrian et al. 2009; James et al. 2011). The focus of many of these studies has been on lake responses to anthropogenic infl uences at the individual lake level and across small populations of lakes. There is a

PAGE 65

65 paucity of studies in the literature, however, that examine trends in lak e changes (as measured by trophic state variables ) across a large population of lakes. As global change has been projected to impact freshwater systems (Kernan et al. 2010), examination of change across many lakes provides insight that individual lake studies or studies of a few lakes may not offer. For example, if long term trends in changes in trop hic state variables are documented across many lakes or a specific grouping of lakes then there may be dominant drivi ng factor(s) influencing change that may be more easily recognized in the lake ( s). Limnologists have long sought to explain variation of l ake systems (Hutchinson 1965) yet understanding whether lakes have changed over time prior to determination of factors driving lake change is a step that is many times overlooked. In the State of Florida, there are some well known individual lakes with long term data records that have been studied in great detail such as Lake O keechobee (e.g., James et al. 2011), Lake Apopka ( e.g., Bachmann et al. 1999), and Lake Annie (Gaiser et al. 2009). However, despite the State of Florida having over 7,700 lakes, of w hich 3,298 are named (Schafer et al. 1986), there has been only one study (Terrell et al. 2000) to use long term records to assess lake changes across a large population of Florida lakes. The study completed by Terrell et al. (2000) included data collected by different groups, a population of 127 Florida lakes, and was completed 16 years ago. T here is need for an updated and improved long term assessment of change within a large population of Florida lakes because, in the past 16 years, the State of Florida has experienced large human population growth ( about a 250% increase) and shifts in climatic patterns

PAGE 66

66 (Gaiser et al. 2009) and research studies have focused on identification of factors driving change in the trophic state of Florida lakes (e.g., James et al. 2011). The purpose of this chapter is to provide an updated and improved ass essment of long term changes in trophic state variables (i.e., total phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements) across a large population of Florida lakes. Long term trophic state data were consistently collected and analyzed by the Florida LAKEWATCH program for at least 15 years and up to 23 years for 193 Florida lakes that spanned the State of Flori da (Figure 3 1) The objectives of this chapter were to 1) identify long term trends in the lake trophic state variables for the population of 193 Florida lakes and 2) explore spatial distribution of lakes with similar identified trends in the examined tro phic state variables Due to the decadal length of the examined data (i.e., one and a half to two decades), this chapter describes trends and explores lake relationships in context of a decadal scale of time. The robust, decadal scale dataset provides not only a good estimation of trends in trophic state variables but also incorporates natural variability and the inclusion of stochastic events, such as hurricanes or extreme droughts and floods. Methods Monthly total phosphorus (TP), total nitrogen (TN), chlorophyll concentrations (CHL) and water clarity (SD) measurements ( obtained by the use of a Secchi disk) were collected by the well established Florida LAKEWATCH program and water chemistry analyses were completed by t he Florida LAKEWATCH laboratory co nsistent over time (Canfield et al. 2002). The Florida LAKEWATCH program began in 1986 and currently samples over 1,500 lakes across the State of Florida. A subset of the Florida LAKEWATCH database was used, which included lakes with all four trophic state

PAGE 67

67 variables collected for at least 15 years and up to 24 years for 193 Florida lakes (Figure 3 2 ). The majority of lakes were sampled monthly at 3 openwater stations and a monthly mean calculated among the stations. An annual mean was calculated from the m onthly means and then used to calculate an overall mean value for the lake. The subset of data examined included 31,050 monthly samplings. Data were analyzed with the statistical package JMP version 8.0 (SAS Institute Inc. 2007). Trends in a nnual mean t ot al phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements were evaluated for the 193 individual Florida lakes using a modified linear regression. The modified linear regression analysis used mean groupings of annual data, or intervals, to decrease the number of samples thereby increasing the predictive power of regression analysis (Prairie 1996). Prairie (1996) found the number of intervals was related to the coefficients of variation ( R2) and also R2 values of 0.65 or greater increased the predictive power of linear regression analysis. Six intervals were calculated f or each trophic state data time series (i.e., six mean data points, where each point represented the mean of three to four years depending on the length of the data series. Six categories were selected for use because Brhyn and Dimberg (2011) demonstrated that aquatic environmental data have the highest R2 values when divided into six classes. Linear regression analysis was completed across the six calculated mean values providing an estimation of trend in the trophic state variable that was of high predictive power (Prairie 1996) and statistically meaningful (Bryhn and Dimberg 2011). To meet the requirements of parametric statistics, logarithmic (base 10) transformations were completed using annual mean

PAGE 68

68 values when appropriate (Snedecor and Cochran 1980) and all statements of statistical significance were at a probability level of < 0.05. There are alternative statistical trend detection methods, such as th e ARMA/ARIMA time series model, that are suggested to handle variance in a manner that does not affect statistical determination of a trend Compared to the selected modified linear regression analysis, the ARMA/ARIMA time series analysis would have result ed in a more conservative estimate of the number of trends identified in the trophic state variables for the 193 Florida lakes The modified linear regression model was selected to determine trends because ARMA/ARIMA time series analysis have strict data limitations ( Bendat and Piersol 2010 ) that would have required the elimination of data and an overall reduction in the size of the dataset. The selected modified linear regression model offered an alternative method that has been shown t o provide a predictively powerful statistical assessment to detect decadal scale trends in annual mean trophic state variables (see Chapter 2) It is important to note, the terms degradation and improvement were used in this chapter to denote the direction of the trend for each trophic state variable and for ease of explanation. The term degradation was used to desc ribe increasing trends in total phosphorus, total nitrogen, and c hlorophyll concentrations and decreasing trends in water clarity measurements The term improvement was used to describe decreasing trends in total phosphorus, total nitrogen, and chlorophyll concentrations and increasing trends in water clarity measurements The use of the terms, degradation and improvement, with reference to lake water quality does not imply; however, the

PAGE 69

69 relationship was bad or good or implie any causeand effect relationships (e.g., degradation in the lake trophic state variables was due to an anthropogenic source). The spatial distribution of the Florida l akes was determined by visual identification of the geographic location of lakes with decadal scale trends in total phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements across the State of Florida. As an initial effort to understand the lakes with identified trends in one or more of the trophic state variables, the influence of natural characteristics (i.e., geology, soils, and hydrology) and nutrient characteristics (i.e., areas of Florida with high or low nutrient concent rations) were examined. The Florida Lake Regions (Griffin et al. 1997) were used as an estimate of the influence of natural characteristics and the total phosphorus zones (TP Zones) and total nitrogen zones (TN Zones) developed for the State of Florida by Bachmann et al. (2012a, 2012b ) were used as an estimate of the influence of nutrient characteristics ArcGIS 10.0 (ESRI 2011 ) was used to examine the lake relationships with the Florida Lake Regions, TP Zones, and TN Zones A one way analysis of variance was used to quantify the amount of variance the Florida Lake Regions, TP Zones, and TN Zones attributed to each trophic state variable. The analysis was completed for the 105 Florida lakes (out of the 193 Florida lakes) where at least one trophic state variable was documented to significantly increase or decrease over the examined period of record. Specifically, examination of the Florida Lake Regions was completed by placing the lakes with identified trends in at least one t rophic state variable into the respective Florida Lake Region (N=47 regions) Using an overall mean for each lake and trophic state variable, t he one way analysis of variance (ANOVA) identified the amount of variance in the given trophic state variable

PAGE 70

70 attribu ted to the Florida Lake Regions. Relationships between lakes with trophic state variable trends and TP Zones and TN Zones were examined by placing lakes into the respective total phosphorus zone (N=6 zones) and total nitrogen zone (N=5). A one way ANOV A estimated the amount of variance in the given trophic state variable accounted for by both the TP zones and TN zones. Bachmann et al. (2012a, 2012b) developed the TP and TN zones by grouping Florida Lake Regions with similar chemical characteristics, so it was assumed the information on soils, physiography, geology, vegetation, climate, and land use/ land cover associated with the Florida Lake Regions carried over to the TP and TN zones. Results There was a wide range in the concentrations of total phosphorus, total nitrogen, and chlorophyll concentrations and measurements of water clarity among the 193 Florida lakes (Table 31) and a broad range in other physical and chemical variables; mean depth, lake area, specific conductance, and color (Table 3 2 ). The 193 Florida lakes encompassed all lake trophic state categories (Table 33) as defined by Forsburg and Ryding ( 1980) Specifically, for total phosphorus 35% (67 lakes) would be classified as oligotrophic (TP < 15 g/L), 28% (54 lakes) m esotrophic (TP 1524.9 g/L), 30% eutrophic (2599.9 g/L), and 7% (13 lakes) hypereutrophic ( 100 g/L). For total nitrogen, 13% (25 lakes) were classified as oligotrophic (< 400 g/L), 21% (41 lakes) mesotrophic (400599.9 g/L), 53% (102 lakes) eutrophic (6001499. 9 g/L), 12% (24 lakes) hypereutrophic ( 1500 g/L). For chlorophyll concentrations 8% (15 lakes) were classified as oligotrophic (< 3 g/L), 33% (63 lakes) mesotrophic (36.9 g/L), 44% (86 lakes) eutrophic (739.9 g/L), and 15% (29 lakes) hypereutrophic ( 40 g/L). For water clarity measurements 7% (14 lakes) were classified as oligotrophic (> 3.96 m),

PAGE 71

71 21% (40 lakes) mesotrophic (2.433.96 m), 42% eutrophic (81 lakes) (0.91 2.44 m), and 29% (56 lakes) hypereutrophic (< 0.91 m). For the population of 193 Florida lakes, t he modified linear regression analysis of the logarithmic (base 10) transformed annual mean data indicated significant increasing monotonic trends in 21% of the population of Florida lakes (40 lakes) for total phosphorus, 26% (50 lakes) for total nitrogen, 12% (23 lakes) for chlorophyll concentrations and 4% (8 lakes) for water clarity measurements Statistically significant decreasing monotonic trends were identified in 7% (14 lakes) for total phosphorus, 6% (12 lakes) f or total nitrogen, 7% (14 lakes) for chlorophyll concentrations and 18% (34 lakes) for water clarity measurements (Table 3 4 ). For the individual 193 Florida lakes, t here were 88 lakes that did not shown statistically significant increasing or decreasing trends in total phosphorus, total nitrogen, chlorophyll concentrations, or water clarity measurements over the period of record (Table 3 5 ). There were 9 lakes with trends of degradation in all four lake trophic state variables (Table 36) Other individual lakes experienced trends of degradation in three, two, or one of the examined trophic state variables (Table 3 6 ). There were two lakes that experienced improving trends in all four of the trophic state variables ; other individual lakes exhibited trends of improvement for three, two, or one of the examined trophic state variables (Table 3 6 ). There was one lake (i.e., Blue North located in Polk County), with an increasing trend in total phosphorus and decreasing trends in total nitrogen and water clarity measurements that did not fit into the denoted trend categories of degradation or improvement.

PAGE 72

72 Individual lakes with similar trends of long term degradation or improvement in the examined trophic state variables were identified across the State of Florida. Clusters of lakes with similar trends in the trophic state variables were visually identified. There were 8 lakes located in close proximity in the panhandle region of Florida that showed trends of degradation as measured by the trophic state variables Specifically, two lakes experienced trends of degradation in all four of the trophic state variables (i.e., l akes Tallavana and Bradford), two lakes experienced trends of degradation in three of the trophic state variables (i.e., l akes Pine Hill, and Ov erstreet) and three lakes showed increasing t rends in total nitrogen (i.e., l akes Arrowhead, Hiawatha, and Monkey Business ) (Figure 3 3 A). There was one lake (i.e., Lake Hall) within this cluster of lakes that showed a decreasing trend in chlorophyll conc entration There was another cluster of lakes, located in the north central region of Florida, which also exhibited trends of degradation in the trophic state variables Specifically, four lakes showed trends of degradation in all four trophic state variables (i.e., l akes Little Santa Fe, Putnam, Santa Fe, and Sheelar), one lake had trends of degradation in three variables (i.e., Lake Hampton), four lakes showed degradation in two variables (i.e., l akes Alto, Cowpen, Lit tle Johnson, and Riley), and five lak es showed degradation in one trophic state variable (i.e., l akes Bivens Arm, Deer, Putnam, Rosa, and Star). Within this cluster of lakes, however, there was one lake that experienced a decreasing trend in total phosphorus (i.e., Lake McMeekin) and one lake that experienced a decreasing trend in chlorophyll concentrations (i.e., Lake Higgenbotham) (Figure 3 3 B). The last cluster of lakes with similar trends was located in southcentral Florida. This cluster of five lakes included two lakes with trends of im provement in three trophic state variables (i.e.,l akes

PAGE 73

73 Holden and Ivanhoe) two lakes with trends of improvement in two variables ( i.e., l akes Conway and Little Fairview), and one lake with a decrease in total nitrogen (i.e., Lake Porter) (Figure 3 3 C). One way analysis of variance demonstrated that for the lakes with identified trends in each trophic state variable from the examined population of 193 Florida lakes, the Florida Lake Regions explained 63% of the variance in total phosphorus, 59% in total n itrogen, 52% in chlorophyll, and 54% in water clarity. The Florida TP Zones explained 53% variance in total phosphorus, 31% in total nitrogen, and 38% in chlorophyll concentrations, and 30% in water clarity measurements The Florida TN Zones explained 41% of the variance in total phosphorus, 52% in total nitrogen, and 34% in chlorophyll concentrations and 33% in water clarity measurements Bachmann et al. (2012a ) also found the Florida Lake Regions to be the best predictor of lake trophic status across Florida lakes. Although the Bachmann et al. (2012 a ) did not complete a longterm analysis of trophic state variables among Florida lakes, their results were similar as the TP and TN zones were found to explain 40% of the variance in total phosphorus and total nitrogen in 1,387 examined Florida lakes. Discussion There were few lakes, a small percentage ( 26%) of the examined population of Florida lakes, with significant decadal scale trends across the measured variables total phosphorus, total nitrogen, chlor ophyll concentrations and water clarity measurements Considering the influence of the cumulative effects of the growing human population on freshwater systems (Carpenter et al. 1998; Carpenter and Lantrop 2008; James et al. 2011) and the subsequent projections of lake water quality to worsen in the future due to inputs of nutrients and changing climate conditions (Adrian et al. 2009; Williamson et

PAGE 74

74 al 2009), the small percentage of the population of 193 Florida lakes with decadal scale in the trophic stat e variables was contrary to the expectation of these statements. Other long term trend analysis studies completed at the statelevel have shown similar results; no significant trends of degradation in total phosphorus, total nitrogen, chlorophyll concentrations, and water clarity measurements for a population of lakes in Vermont (N=195 lakes; 11 years of data Smeltzer et al. 1989) and in Florida (N=127 lakes; varying periods of record, Terrell et al. 2000). Bachmann et al. (2012a,b,c) examined plant nutrie nt and chlorophyll concentrations for 1 ,387 Florida lakes across varying scales of analysis (e.g., paleolimnological comparisons or natural (unaltered by human activities) lake condition comparisons) and found the majority of examined Florida lakes had not experienced changes in plant nutrients and chlorophyll concentrations. Analogous conclusions have been drawn from national lake assessments as well. For example, using a subset of lakes that were sampled as a part of the National Eutrophication Survey in the 1970s, of which many received sewage pollution, the United State Environmental Protection Agency (USEPA) (2009) demonstrated that over a 30year period, total phosphorus concentrations did not change in 24% of the nations lakes (N= 800 lakes) and chlorophyll a concentrations did not change in 51% of the same subset of examined lakes. Carlson et al. (2012) examined water clarity measurements, ranging in record from at least 5 years and up to 40 years for 4,812 of the nations water bodies and determine d 83% of the lakes did not significantly change over t he examined period of record, respectively. T he percentage of the population of 193 Florida lakes that experienced decadal scale trends of degradation was greater than the percentage of the population of lakes

PAGE 75

75 with identified decadal scale trends of improvement in the examined trophic state variables Although the percentage of the lakes with identified trends was smal l compared to the percentage of lakes with no trends there were individual lakes that experienced decadal scale trends of increases in total phosphorus, total nitrogen, and chlorophyll concentrations, and decreases in water clarity measurements (Table 34 ) Compared to other trend analyses of trophic state variables for Florida lakes, the percentage of the population of 193 Florida lakes with identified trends in the trophic state variables was greater than determined by both Terrell et al. (2000) and Bach mann et al. (2012a,b,c). The divergent results could be attributed to examination of different populations of Florida lakes, varying availability of data, or the length of time used to detect trends and make comparisons to other data. The different conclus ions illustrate the need to consider the scale of analysis when interpreting trophic state variable trends, especially as limnologists and lake managers may interpret and value results differently depending on the scale (e.g., geological time scale versus decadal time scale). T here were 9 individual lakes out of the 193 examined Florida lakes with increasing trends in total phosphorus, total nitrogen, and chlorophyll concentrations and decreasing trends in water clarity measurements over the decadal scale period or record. These lakes offer the opportunity for limnologists and lake managers to better understand the mechanisms driving lake change because all four of the trophic state variables exhibited decadal scale trends. Following trophic state theory, the factors driving these trophic state variable trends, whether anthropogenic or natural factors may be more easily identified and understood in the 9 lakes compared to lakes where trends

PAGE 76

76 were found in one, two, or three of the other trophic state var iables Detailed e xamination of the two lakes out of the 1 93 Florida lakes that showed improving trends in all four of the trophic state variables may additionally be worthwhile to recognize why these improvements occurred and if these improvements could b e replicated in other lake ecosystems. Overall, the 9 lakes with trends of degradation in the examined trophic state variables was less than 5% of the population of examined 193 Florida lakes indicating the state of Florida lakes, as a whole, may not be as severe as hypothesized ( USEPA 2000). Another alternative to focus future research and management efforts would be to target spatial clusters of lakes with similar trophic state variable trends, either trends of degradation or tr ends of improvement. As t his chapter demonstrated, there were three spatial clusters of lakes across the State of Florida; two clusters where the lakes had trends of degradation and one cluster where the lakes had trends of improvement. Rather than exploring driving factors within an indivi dual lake, focus on a cluster of lakes or even the connectivity of the lake systems (i.e., the connection of lakes by other waters) incorporates a spatial aspect which may enhance the understanding of both anthropogenic and natural factors influen cing lake trophic state changes and trends. The major conclusion of this chapter is that many Florida lakes have not experienced decadal scale trends in total phosphorus, total nitrogen, chlorophyll concentrations and water clarity measurements. T here are, however, select individual Florida lakes that have experienced concerning trends in the lake trophic state variables Both research and lake management efforts should focus on these individual lakes and the identified clusters of lakes, to identi fy and understand the drivers that may

PAGE 77

77 be influencing these changes. If the goal is to provide a sustainable lake ecosystem for the future, then such attention and assessment of Floridas lakes are necessary.

PAGE 78

78 Table 31. Summary statistics (i.e., m ean, median, minimum, maximum, an d coefficient of variation (%)) for annual mean total phosphorus ( g/L), total nitrogen ( g/L), chlorophyll concentrations ( g/L), and water clarity measurements (m) for the population of 193 Florida lakes. Trophic State V ariable N Mean Median Minimum Maximum C oefficient of Variation Total Phosphorus 193 38.4 18.3 3.8 357.5 152 Total Nitrogen 192 92 1 71 6 109 3,7 80 75 Chlorophyll 193 23.4 9.2 1.6 199 149 Water Clarity 191 1.92 1.58 0.32 6.69 69

PAGE 79

79 Table 32. Summary statistics (i.e., mean, median, minimum, maximum, and coefficient of variation (%)) for supplemental data including mean depth (m), surface area (ha), specific conductance ( S), and true color (Pt Co Units) for the population of 193 Flor ida lakes. Supplemental Variable N Mean Median Minimum Maximum C oefficient of Variation Mean Depth (m) 132 3.2 3 0.8 9.9 45 Surface Area (ha) 167 525 63 1 19,808 344 Specific Conductance (S) 182 170 151 0.01 1316 91 True Color (Pt Co Units) 193 47 22 2.2 444 139

PAGE 80

80 Table 33 P ercentage of the population of 193 Florida lakes for total phosphorus, total nitrogen, chlorophyll concentration, and water clarity measurements within each trophic state classification (i.e., oligotrophic, mesotrophic, eutrophic, and hypereutrophic) following Forsburg and Ryding (1980). Percent of 193 Florida lakes Total Phosphorus Oligotrophic 35 Mesotrophic 28 Eutrophic 30 Hypereutrophic 7 Total Nitrogen Oligotrophic 13 Mesotrophic 21 Eutrophic 53 Hypereutrophic 12 Chlorophyll Oligotrophic 8 Mesotrophic 33 Eutrophic 44 Hypereutrophic 15 Water Clarity Oligotrophic 7 Mesotrophic 21 Eutrophic 42 Hypereutrophic 29

PAGE 81

81 Table 34 P ercentage of the population of Florida lakes with increasing trends, decreasing trends, and no trends detected in annual mean total phosphorus, total nitrogen, chlorophyll concentrations and water clarity measurements over a period or record of at least 15 years. Trophic State Variable N Increasing Trend (%) Decreasing Trend (%) No Trend (%) Total Phosphorus 193 21 7 72 Total Nitrogen 192 26 6 68 Total Chlorophyll 193 12 7 81 Water Clarity 191 4 18 78

PAGE 82

82 Table 3 5. Linear regression analysis using six data points where the point reflects the mean among the six annual mean data for the logarithmic base 10 (L10) total phosphorus (TP), total nitrogen (TN), chlorophyll (CHL), and water clarity measurements (SD). A s ignificant monotonic trend ( slope value, and coefficient of determination ( R) for each trophic state variable for t he 193 Florida lakes. County Lake L 10 TP Slope L 10 TP R 2 L 10 TN Slope L 10 TN R 2 L 10 CHL Slope L 10 CHL R 2 L 10 SD Slope L 10 SD R 2 Alachua Alto 0.75 0.79 0.02 0.29 0.04 0.54 Alachua Bivans Arm 0.74 0.06 0.59 0.06 0.28 0.07 0.60 Alachua Little Orange 0 0.37 0.0008 0 0.06 0.40 0.02 0.34 Alachua Little Santa Fe 0.68 0.71 0.73 0.74 Alachua Lochloosa 0.04 0.41 0.003 0 0.03 0.04 0 0 Alachua Newnan 0.05 0.45 0.007 0.01 0.05 0.24 0.02 0.15 Alachua Orange 0.07 0.23 0.04 0.30 0.05 0.10 0.05 0.50 Alachua Santa Fe 0.76 0.77 0.07 0.88 0.87 Alachua Wauberg 0.03 0.57 0.03 0.49 0.04 0.49 0.02 0.26 Bay Powell 0.004 0 0.02 0.54 0.01 0.30 0.02 0.50 Bradford Hampton 0.70 0.84 0.85 0.04 0.61 Brevard Forest 0.04 0.64 0.002 0.02 0.02 0.20 0.90 Citrus Henderson 0.01 0.04 0.002 0 0.002 0.01 0.01 0.02 Citrus Hernando 0.75 0.06 0.64 0.08 0.62 0.05 0.53 Citrus Little Henderson 0.005 0.02 0.001 0 0.02 0.17 0.02 0.26 Citrus Todd 0.72 0.81 0.90 0.05 0.48 Citrus Tsala Apopka 0.02 0.25 0.02 0.18 0.02 0.14 0.004 0.02 Clay Asbury North 0.02 0.25 0.01 0.38 0.02 0.24 0.003 0.02 Clay Crystal 0.004 0.01 0.003 0.01 0.04 0.18 0.03 0.40 Clay Deer 0.03 0.33 0.80 0.04 0.42 0.03 0.49 Clay Johnson 0.04 0.55 0.05 0.64 0.06 0.45 0.04 0.24 Clay Kingsley 0.009 0.09 0.02 0.49 0.03 0.38 0.01 0.25 Clay Lily 0.05 0.40 0.03 0.22 0.08 0.56 0.01 0.10 Clay Little Crystal 0.04 0.60 0.02 0.30 0.04 0.34 0.03 0.35

PAGE 83

83 Table 3 5. C ontinued County Lake L 10 TP Slope L 10 TP R 2 L 10 TN Slope L 10 TN R 2 L 10 CHL Slope L 10 CHL R 2 L 10 SD Slope L 10 SD R 2 Clay Little Johnson 0.73 0.83 0.05 0.58 0.07 0.54 Clay Sheelar 0.92 0.91 0.78 0.82 Flagler Disston 0.80 0.79 0.03 0.18 0.01 0.09 Flagler Ribbon North 0.02 0.58 0.01 0.38 0.06 0.52 0.02 0.26 Gadsden Tallavana 0.88 0.93 0.82 0.89 Highlands Charlotte 0.003 0.07 0.88 0.007 0.07 0.89 Highlands Clay 0.003 0.01 0.01 0.41 0.01 0.08 0.01 0.14 Highlands Francis 0 0 0.02 0.61 0.005 0.03 0.002 0.01 Highlands Grassy 0.72 0.02 0.51 0.04 0.21 0.02 0.13 Highlands Huntley 0.77 0.002 0.01 0.97 0.90 Highlands Jackson 0.02 0.16 0.01 0.22 0.03 0.09 0.02 0.31 Highlands Josephine East 0.01 0.11 0.002 0.01 0.008 0.08 0.86 Highlands June 0.02 0.34 0 0 0.02 0.25 0 0 Highlands Lillian 0.02 0.11 0.84 0.03 0.21 0.02 0.30 Highlands Little Jackson 0.001 0 0.82 0.02 0.12 0.76 Highlands Persimmon 0.85 0.81 0.02 0.60 0.01 0.31 Highlands Placid 0.01 0.22 0.01 0.23 0.01 0.14 0.008 0.08 Highlands Red Beach 0.02 0.48 0.02 0.37 0.002 0 0.02 0.61 Highlands Redwater 0.03 0.38 0.02 0.07 0.03 0.25 0.05 0.33 Highlands Sebring 0.01 0.10 0.01 0.26 0.83 0.03 0.35 Hillsborough Armistead 0.05 0.60 0.02 0.46 0.09 0.54 0.02 0.20 Hillsborough Brant 0.01 0.02 0.02 0.08 0.006 0 0.01 0.03 Hillsborough Carroll 0.02 0.33 0.004 0.17 0.03 0.17 0.02 0.25

PAGE 84

84 Table 3 5. C ontinued County Lake L 10 TP Slope L 10 TP R 2 L 10 TN Slope L 10 TN R 2 L 10 CHL Slope L 10 CHL R 2 L 10 SD Slope L 10 SD R 2 Hillsborough Crenshaw 0.001 0 0.006 0.06 0.02 0.03 0.004 0.01 Hillsborough Dead Lady 0.03 0.36 0.002 0.01 0.009 0.01 0.005 0.09 Hillsborough Hiawatha 0.03 0.53 0.009 0.15 0.01 0.05 0.03 0.57 Hillsborough Hobbs 0.81 0.78 0.12 0.64 0.87 Hillsborough James 0.01 0.04 0.01 0.08 0.05 0.42 0.02 0.08 Hillsborough Juanita 0.70 0.04 0.59 0.67 0.92 Hillsborough Keystone 0.03 0.55 0.02 0.43 0.04 0.34 0.04 0.46 Hillsborough Magdalene 0.02 0.21 0.01 0.41 0.001 0 0.003 0.01 Hillsborough Wilson 0.04 0.57 0.002 0.12 0.01 0.40 0.02 0.31 Lake Beauclaire 0.77 0.006 0.03 0.01 0.02 0.006 0.04 Lake Cherry 0.05 0.64 0.72 0.93 0.81 Lake Crooked 0.04 0.48 0.02 0.12 0.04 0.38 0.04 0.64 Lake Dora East 0.06 0.62 0.01 0.20 0.03 0.29 Lake Dora West 0.04 0.44 0.01 0.27 0.02 0.23 Lake Dorr 0.006 0.06 0.76 0.03 0.48 0.001 0 Lake East Crooked 0.02 0.44 0.80 0.006 0.51 0.005 0 Lake Emma 0.06 0.65 0.06 0.55 0.04 0.56 0.66 Lake Eustis 0.002 0.02 0.0003 0 0.03 0.45 0.008 0.02 Lake Gertrude 0.90 0.004 0.08 0.05 0.63 0.01 0.65 Lake Grasshopper 0.78 0.04 0.65 0.03 0.20 0.05 0.39 Lake Griffin 0.03 0.49 0.008 0.09 0.03 0.19 0.004 0.01 Lake Harris 0.01 0.13 0.01 0.13 0.06 0.59 0.02 0.19 Lake Joanna 0.009 0.25 0.82 0.01 0.08 0.01 0.36 Lake Little Harris 0.02 0.22 0.001 0 0.006 0.01 0.02 0.14 Lake Lorraine 0.91 0.87 0.79

PAGE 85

85 Table 3 5. Continued County Lake L 10 TP Slope L 10 TP R 2 L 10 TN Slope L 10 TN R 2 L 10 CHL Slope L 10 CHL R 2 L 10 SD Slope L 10 SD R 2 Lake May 0.02 0.46 0.01 0.17 0.93 0.02 0.41 Lake Minneola 0.84 0.73 0.09 0.54 0.1 0.65 Lake Peanut Pond 0.03 0.48 0.001 0 0.06 0.57 0.02 0.46 Lake Picciola 0.03 0.35 0.02 0.16 0.05 0.22 0.01 0.05 Lake Sellers 0.80 0.77 0.77 Lake Trout 0.04 0.40 0.76 0.04 0.12 0.79 Lake Unity 0.005 0.10 0.02 0.37 0.04 0.32 0.02 0.55 Lake Yale 0.71 0.89 0.74 0.68 Lee Little Murex 0.79 0.73 0.003 0.02 0.01 0.11 Leon Arrowhead 0.02 0.23 0.69 0.08 0.43 0.01 0.03 Leon Blairstone 0.006 0.01 0.008 0.01 0.02 0.03 0.01 0.10 Leon Blue Heron 0 0.03 0.004 0.01 0.09 0.46 0.02 0.17 Leon Bradford 0.66 0.76 0.67 0.89 Leon Diane 0.03 0.22 0.01 0.15 0.005 0 0.03 0.18 Leon Hall 0.03 0.49 0.007 0.08 0.73 0 0 Leon Hiawatha 0.05 0.61 0.68 0.04 0.40 0.02 0.18 Leon Minniehaha 0.02 0.56 0.03 0.58 0.007 0.04 0.008 0.02 Leon Monkey Business 0.002 0 0.72 0.03 0.11 0.02 0.16 Leon Overstreet 0.89 0.91 0.85 0.02 0.40 Leon Petty Gulf 0.007 0.10 0.01 0.06 0.02 0.10 0.007 0.02 Leon Pine Hill 0.02 0.64 0.77 0.87 0.69 Marion Charles 0.04 0.32 0.76 0.003 0.00 0.006 0.09 Marion Deerback 0.008 0.06 0 0 0.03 0.29 0.01 0.12 Marion Eaton 0.03 0.13 0.01 0.17 0.001 0 0 0 Marion Halfmoon 0.0005 0 0.69 0.02 0.61 0.005 0.03

PAGE 86

86 Table 3 5. Continued County Lake L 10 TP Slope L 10 TP R 2 L 10 TN Slope L 10 TN R 2 L 10 CHL Slope L 10 CHL R 2 L 10 SD Slope L 10 SD R 2 Marion Sunset Harbor 0.02 0.58 0.004 0.03 0.007 0.14 0.002 0.01 Marion Weir 0.02 0.39 0.009 0.25 0.01 0.59 0.02 0.34 Orange Adair 0.07 0.41 0.02 0.49 0.04 0.17 0.05 0.30 Orange Bay 0.01 0.08 0.02 0.53 0.03 0.15 0.03 0.28 Orange Bennett 0.004 0.05 0.69 0.02 0.06 0.005 0.01 Orange Bessie 0.73 0.71 0.67 0.75 Orange Burkett 0.008 0.07 0.006 0.03 0.03 0.41 0.79 Orange Conway North 0.005 0.04 0.02 0.52 0.92 0.96 Orange Conway South 0.004 0.05 0.002 0.01 0.06 0.34 0.02 0.30 Orange Down 0.001 0 0.008 0.41 0.04 0.40 0.03 0.59 Orange Eola 0.001 0 0.001 0 0.02 0.08 0.003 0.01 Orange Estelle 0.001 0 0.002 0.01 0.004 0.02 0.03 0.53 Orange Farrar 0.001 0.01 0.003 0 0.0005 0 0.002 0.01 Orange Formosa 0.03 0.45 0.01 0.20 0.03 0.48 0.04 0.64 Orange Georgia 0.03 0.45 0.02 0.32 0.03 0.27 0.03 0.64 Orange Giles 0.008 0.36 0.006 0.04 0.02 0.32 0.02 0.60 Orange Hickorynut 0.66 0.03 0.29 0.75 0.78 Orange Holden 0.86 0.03 0.36 0.68 0.69 Orange Ivanhoe East 0.79 0.03 0.61 0.77 0.76 Orange Ivanhoe Middle 0.78 0.001 0 0.80 0.73 Orange Ivanhoe West 0.79 0.007 0.10 0.73 0.87 Orange John's 0.04 0.26 0.007 0.07 0.67 0.04 0.33 Orange Little Fairview 0.77 0.84 0.11 0.55 0.09 0.64 Orange Little Hickorynut 0.78 0.77 0.88 0.92 Orange Lurna 0.03 0.28 0.01 0.27 0.04 0.25 0.02 0.17

PAGE 87

87 Table 3 5. Continued County Lake L 10 TP Slope L 10 TP R 2 L 10 TN Slope L 10 TN R 2 L 10 CHL Slope L 10 C HL R 2 L 10 SD Slope L 10 SD R 2 Orange Marsha 0.02 0.25 0.02 0.62 0.05 0.40 0.03 0.60 Orange Mary Jane 0.82 0.02 0.41 0.72 0.02 0.29 Orange Minnehaha 0.01 0.27 0.01 0.20 0.01 0.04 0.003 0 Orange Moxie 0.01 0.14 0.02 0.32 0.02 0.02 0.01 0.03 Orange North Lotta 0.03 0.36 0.01 0.27 0.02 0.26 0.02 0.35 Orange Ola 0.01 0.30 0.74 0.04 0.58 Orange Olympia 0.006 0.07 0.02 0.37 0.75 0.02 0.18 Orange Peach 0.005 0.02 0.74 0.03 0.05 0.006 0.02 Orange Porter 0.01 0.10 0.78 0.002 0 0.006 0.01 Orange Primavista 0.02 0.60 0.86 0.03 0.54 0.01 0.21 Orange Rowena 0.03 0.52 0.002 0.02 0.04 0.64 0.04 0.55 Orange Sarah 0.01 0.15 0.02 0.32 0.11 0.63 Orange Shannon 0.01 0.20 0.02 0.06 0.03 0.56 Orange South Lotta 0.02 0.19 0.006 0.11 0.68 0.02 0.37 Orange Spring 0.75 0.01 0.27 0.006 0 0.03 0.42 Orange Starke 0.83 0.02 0.56 0.04 0.65 0.03 0.50 Orange Susannah 0.03 0.48 0.04 0.51 0.1 0.52 0.06 0.43 Orange Waunatta 0.02 0.62 0.79 0.73 0.87 Orange Willis 0.01 0.10 0.001 0.01 0.03 0.14 0.006 0.05 Osceola Alligator 0.81 0.03 0.43 0.007 0.04 0.82 Osceola Brick 0.73 0.04 0.60 0.01 0.13 0.01 0.32 Osceola Center 0.01 0.04 0.04 0.60 0.02 0.07 0.02 0.17 Osceola Coon 0 0 0.02 0.27 0.004 0.01 0.02 0.43 Osceola Kissimmee 0.03 0.54 0.67 0.05 0.45 0.02 0.53 Osceola Lizzie 0.02 0.39 0.03 0.33 0.01 0.15 0.04 0.56

PAGE 88

88 Table 3 5. C ontinued County Lake L 10 TP Slope L 10 TP R 2 L 10 TN Slope L 10 TN R 2 L 10 CHL Slope L 10 CHL R 2 L 10 SD Slope L 10 SD R 2 Osceola Trout 0.01 0.07 0.02 0.32 0.02 0.23 0.004 0.01 Polk Big Bass 0.01 0.11 0.69 0.05 0.57 0.78 Polk Blue North 0.66 0.81 0.09 0.50 0.69 Polk Boca Cove 0.01 0.10 0.72 0.04 0.52 0.78 Polk Dexter 0.80 0.002 0.02 0.002 0.01 0 0 Polk Fauna 0.86 0.002 0 0.03 0.29 0.02 0.26 Polk Flora 0.009 0.06 0.02 0.63 0.03 0.55 0.68 Polk Gaskin's Cut 0.01 0.10 0.03 0.61 0.04 0.39 0.70 Polk Little Bass 0.02 0.14 0.04 0.60 0.05 0.46 0.75 Polk Weohyakapka 0.05 0.64 0.87 0.08 0.53 0.83 Putnam Blue 0.003 0 0.06 0.56 0.02 0.06 0.02 0.14 Putnam Broward 0.69 0.82 0.04 0.42 0.005 0.12 Putnam Chipco 0.003 0.01 0.04 0.65 0.02 0.15 0 0 Putnam Como 0.76 0.67 0.02 0.09 Putnam Cowpen 0.77 0.80 0.02 0.06 0.03 0.40 Putnam Fanny 0.02 0.14 0.03 0.22 0.03 0.14 0.02 0.18 Putnam Gillis 0.005 0.02 0.04 0.22 0.02 0.23 0.004 0.01 Putnam Higgenbotham 0.003 0.03 0.001 0 0.86 0.006 0.04 Putnam McMeekin 0.75 0.006 0.19 0.06 0.50 0.02 0.25 Putnam Punchbowl 0.81 0.02 0.50 0.04 0.27 0.04 0.59 Putnam Riley 0.78 0.86 0.06 0.44 0.02 0.48 Putnam Rosa 0.002 0.01 0.85 0.08 0.39 0.01 0.32 Putnam Star 0.03 0.35 0.70 0.004 0.02 0.02 0.36 Putnam Winnott 0.84 0.91 0.94 0.96 Seminole Adelaide 0.02 0.35 0.02 0.21 0.02 0.12 0.73

PAGE 89

89 Table 3 5. C ontinued County Lake L 10 TP Slope L 10 TP R 2 L 10 TN Slope L 10 TN R 2 L 10 CHL Slope L 10 CHL R 2 L 10 SD Slope L 10 SD R 2 Seminole Bear 0.002 0.01 0.90 0.05 0.60 0.03 0.44 Seminole Florida 0.03 0.40 0.01 0.14 0.03 0.05 0.03 0.40 Seminole Little Bear 0.02 0.47 0.74 0.05 0.28 0.04 0.36 Seminole Mary 0.03 0.55 0.02 0.56 0.04 0.25 0.02 0.46 Seminole Orienta 1 0.003 0.01 0.001 0 0.01 0.06 0.03 0.26 Seminole Orienta 2 0.01 0.13 0.004 0.01 0.02 0.22 0.02 0.11 Seminole Rock 0.93 0.90 0.03 0.30 0.96 Seminole Seminary 0.01 0.53 0.74 0.006 0.02 0.01 0.22 Seminole Spring 0.88 0.90 0.88 0.79 Seminole Woods 0.02 0.64 0.01 0.16 0.83 0 0.00 St Lucie Margaret 0.01 0.10 0.05 0.39 0.03 0.39 0.03 0.66 Sumter Panasoffkee 0.08 0.61 0.78 0.08 0.55 0.83 Volusia Ashby 0.02 0.18 0.90 0.03 0.34 0.03 0.56 Volusia Beresford 0.002 0.01 0.01 0.43 0.01 0.15 0.75 Volusia Bethel 0.80 0.02 0.21 0.07 0.45 0.07 0.58 Volusia Broken Arrow 0.66 0.04 0.40 0.71 Volusia Charles 0.88 0.88 0.06 0.60 0.91 Volusia Harney 0.05 0.34 0.01 0.15 0.01 0.01 0.03 0.22 Volusia Winnemissett 0.68 0.02 0.23 0.05 0.46 0.02 0.11 Walton Camp Creek 0.03 0.44 0.006 0.13 0.003 0 0.006 0.10 Walton Spring 0.003 0.02 0.02 0.49 0.01 0.06 0.03 0.52

PAGE 90

90 Table 36 N umber of lakes (no. of lakes) out of the examined annual mean data for 193 Florida lakes with: 1) decadal scale increasing trends in total phosphorus (TP) total nitrogen (TN) and chlorophyll concentrations (CHL) and decreasing trends in water clarity measurements (SD) and 2) decadal scale decreasing trends in total phosphorus (TP), total nitrogen (TN), and chlorophy ll concentrations (CHL) and increasing trends in water clarity measurements (SD) and 3) no trend in any of the four trophic state variables No. of Lakes Trend No. of Lakes Trend No. of Lakes Trend No. of Lakes Trend 9 Increase TP, TN, CHL 3 Increase TP, TN 8 Increase TP, TN 11 Increase TP Decrease SD Decrease SD 2 Increase TP, CHL 17 Increase TN 2 Increase TP, CHL 1 Increase TP 3 Increase CHL Decrease SD Decrease SD 9 Decrease SD 3 Increase TN, CHL 7 Increase TN Decrease SD Decrease SD 2 Decrease TP, TN, CHL 5 Decrease TP, CHL 3 Decrease TP, TN 4 Decrease TP Increase SD Increase SD 1 Decrease CHL 5 Decrease TN Increase SD 5 Decrease SD 88 No Trend

PAGE 91

91 F igure 31. D istribution of the examined population of 193 Florida lakes.

PAGE 92

92 Figure 32 N umber of years each lake was sampled (N= 193 Florida lakes). The numbers above the bars denote the number of lakes that were sampled for the respective number of years.

PAGE 93

93 Figure 33. Florida lakes (N=105 lakes) with detected decadal scale trends of degradation (i.e., increases in total phosphorus (TP), total nitrogen (TN), and chlorophyll concentrations (CHL) and decreases in water clarity (SD) measurements, and trends of improvement (i.e., decreases in TP, TN, and CHL concentrations and increases in SD). Spatial clusters of individual lakes exhibiting similar trends of degradation in one or more of the examined variables were identified (A and B) and a spatial cluster of lakes exhibiting similar trends of improvement in one or more of the variables were identified (C).

PAGE 94

94 CHAPTER 4 SEASONAL PATTERNS OF PHYTOPLANKTON BIOMAS S AND RESPONSES TO CLIMATE IN SUBTROPICAL, FLORIDA LAKE S Background Primary production occ urs across the biosphere following annual cycles of growth and senescence driven by the climatic system. Lake ecosystems are particularly sensitive to climaterelated changes and many recent lake studies have focused on the examination of response variables like water temperature, dissolved organic carbon, or planktonic composition as indicators of climate changes (Williamson et al. 2009; Adrian et al. 2009). Phytoplankton biomass, in particular, is projected to increase (Kernan et al. 2012; Jeppesen et al. 2007a, 2010) with continuing trends of changes in climate (Mann et al. 1998, Magnuson et al. 2000, IPCC 2007). Yet, some scientists argue projections related to a changing climate do not account for the role of natural variability (Battarbee 2010). Season al patterns in phytoplankton biomass, which are naturally driven by climate at both a local and global scale, are many times disregarded or the seasonal variability is even removed prior to data analyses (Bendat and Piersol 2010). The absence of identifying seasonal variability and understanding how these seasonal patterns vary among regions could have major consequences in the interpretation of linkages in phytoplankton biomass to climate. Recurrent (i.e., how phytoplankton biomass compare from year to year) and synchronous (i.e., how phytoplankton biomass compare at a given time within a year) seasonal patterns (Clocern and Jassby 2008) have been well documented in temperate lakes. The seasonal pattern of phytoplankton in temperate lakes is generally defined by a distinct spring bloom followed by a summer depression, a subsequent fall bloom, and

PAGE 95

95 low levels through the winter months (Hutchinson 1967; Marshall and Peters 1989). The seasonal patterns observed in temperate lakes, however, do not necessarily reflect the annual cycles of phytoplankton biomass i n all lakes. Comparatively, s ubtropical lakes are located at lower latitudes and sustain longer periods of warm er water temperatures Generally, subtropical lakes are shallow in depth (i.e., < 3m), experi ence temporary thermal stratification, and have short water r esidence times (Lewis 1973; Scheffer 1998). Also, subtropical lakes do not have a period of winter ice cover as do the temperate lakes (Bachmann et al. 2012c) High levels of phytoplankton biomas s experienced in subtropical lakes; therefore, would extend greater in duration over an annual cycle compared to temperate lakes Moss (1973) determined the optimal temperature for phytoplankton growth was 23 C and above. There are many months of the year where subtropical lakes experience temperatures 23 C and greater supporting the postulation that subtropical lakes experience longer periods of high phytoplankton biomass annually compared to temperature lakes. There is limited documentation of seasonal patterns of phytoplankton biomass in subtropical lakes (Brown et al. 1998), restricting ability to compare seasonal patterns of phytoplankton biomass across changes in latitude and longitude. The range in latitude a ffects phytoplankton biomass at the large scale (e.g., across North America), especially as latitude is linked to variation in local climate conditions (Brylinsky and Mann 1973). Similarly, phytoplankton biomass (as measured by chlorophyll conce ntrations ) vary with longitude (Soranno et al. 1999). The influence of differences in climate, associated with changes in latitude and longitude, on levels of phytoplankton biomass has not been thoroughly explored in the literature, but offers an opportunity to

PAGE 96

96 und erstand the latitude and longitudinal differences observed in phytoplankton biomass. The identification of recurrent and synchronous seasonal patterns in phytoplankton biomass in subtropical lakes facilitates the ability to explore contrasting seasonal pat terns at the largescale, across changes in both latitude and longitude. W ater is a dynamic entity, marked my measurable changes in the biology, chemistry, and physical aspects that fluctuate within a year and among years. To understand these changes, limnologists frequent ly classify waters Classification by lake trophic state category is commonly used to encompass the continuum of movement towards a more biologically pr oductive system (Carlson 1977). Chlorophyll concentrations are often used as a proxy for phytoplankton biomass (Canfield et al. 1985) and to summarize lake trophic conditions (Forsburg and Ryding 1980). There is a strong, positive relationship between chlorophyll concentrations and inter annual variance (Knowlton et al. 1984); therefore, s easonal variation in chlorophyll concentrations has been suggested to be greater in more eutrophic or biologically rich lakes (Marshall and Peters 1989; Brown et al. 1998). Generally, the range of chlorophyll values per lake trophic state categorization (F orsburg and Ryding 1980) is smaller in temperate lakes (Marshall and Peters 1989) compared to subtropical lakes that include the whole range of chlorophyll values within a trophic category (e.g., chlorophyll concentrations of 740 g/L for eutrophic classifications) Classifying trophic state categories using the annual mean (i.e., referred to as lake year ) versus a mean value per individual lake, captures a wider range of chlorophyll values within each trophic state category and also reflects the change waters experience year to year.

PAGE 97

97 The study objectives were to evaluate seasonal patterns of phytoplankton biomass or chlorophyll variability to 1) identify recurrent and synchronous seasonal patterns, 2) determine whether seasonal patterns differ among waters classified by trophic state category, 3) examine the influence of the climatic factors temperature and rainfall on seasonal patterns of phytoplankton biomass, and 4) determine if the frequency of occurrence of extreme chlorophyll events has chang ed over the years of record. Methods Data sets Two datasets were used in this chapter. The first dataset included c ontinuous monthly chlorophyll concentrat ions (January to December) for 27 Florida lakes. The 27 Florida lakes were obtained from the Florida LAKEWATCH database and ranged in record from 20 to 2 4 years. The second dataset which included monthly chlorophyll concentrations f or 193 Florida lakes ranging in record from 15 to 24 years, was also obtained from the Florida LAKEWATCH database. The 27l ake database was used for all analyses presented in this chapter. The 193lake database was used only to compare seasonal patterns in chlorophyll concentrations to the 27lake database. The 193lake database was not selected for use in other analyses because there were missing monthly data and the major objective of the chapter was to understand inter and intraseasonal patterns among the years of record. The examined 27 lakes provided a representative subset of Florida lakes that ranged in chlorophyll conc entrations (Table 41); therefore, the 27 lake dataset provided enhanced analyses of seasonal pat terns in subtropical lakes. The analytical methods for chlorophyll for both the 27 lake and 193lake datasets, did not correct for pheophytins; therefore, the

PAGE 98

98 estimates of chlorophyll were considered total chlorophyll concentrations (Canfield et al. 2002). Both the Florida LAKEWATCH sampling and chlorophyll analytical methods were consistent over time (Canfield et al. 2002). Total chlorophyll concentrations (g /L) were used to estimate phytoplankton biomass Approach to I dentify Seasonal P atterns i n Phytoplankton B iomass Although t here are many ways to identify seasonal patterns in phytoplankton biomass (Clocern and Jassby 2008) three approaches were used The first approach examined the variability attributed to season by ARMA/ARIMA t ime series model analysis. The second approach identified the month in which maximum chlorophyll values occurred each year a common approach used to understand seasonal patter ns across ecological systems; subtropical lakes (Brown et al. 1998), temporal lakes (Marshall and Peters 1989), coastal and pelagic oceanic systems (Clocern and Jassby 2008), and terrestrial vegetation systems (Myneni et al. 1998). The third approac h ident ified the number of extreme chlorophyll events that occurred each month within each year of examination. Extreme events included elevated chlorophyll concentrations and did not include low extreme chlorophyll values because a component of the chapter was t o identify whether climaterelated factors increased chlorophyll concentrations in subtropical lakes. The three approaches used to identify seasonal patterns were also completed categorizing waters by the annual mean chlorophyll concentration into the tr ophic state categories outlined by Forsburg and Ryding (1980). The trophic state classification by the annual mean chlorophyll concentration was referred to as lakeyear T rophic state classification of waters by lakeyear allowed an estimate of changes in chlorophyll concentrations by trophic category both within and among years.

PAGE 99

99 Statistical A nalyses The JMP software (version 8.0) was used f or all statistical analyses (SAS 2007) and all statements of statistical significance were determined at a probability of < 0.05. Monthly chlorophyll means were det ermined by averaging the three, openwater stations sampled by Florida LAKEWATCH on the same day for the individual month. A nnual chlorophyll means were determined by averaging the 12 monthly means. The data were evenly distributed throughout the year as each of the 27 lakes had 12 consecutive months of c hlorophyll data per year for 20plus years. There were a few missing value s due to restricted sampling. In such cases, a missing monthly datum was replaced by the calculated mean from the month prior and following the missing datum For the 193lake dataset, the number of missing datum was great; therefore, the data presented include missing monthly datum. Chlorophyll concentrations were not normally distributed for the 27 Florida lakes or the majority of the 193 Florida lakes (KSL Goodness of Fit Test). B ecause the goal of the paper was to examine patterns in seasonal variabilit y, logarithmic transformations were not completed unless required for statistical analysis (i.e., ARMA/ARIMA time series modeling) Identification of seasonal patterns was completed using the three outlined approaches The goal of the first approach was t o identify seasonal patterns from examination of the variance that attributed to season and removed prior to application of ARMA/ARIMA time series model analysis. Time series models identify variance that is attributed to a seasonal component and display t his variance through a plot called a spectral density plot The spectral density plot represent s a function of the period and frequency of the chlorophyll concentrations with the integral of the plot equal to the variance exhibited by the examined variable over the entire period of record (i.e., 240-

PAGE 100

100 252 total chlorophyll observations for each i ndividual Florida lake). T ime series models must meet the requirements of parametric statistics, so logarithmic transformed (base 10) chlorophyll concentrations were used in this analysis (Snedecor and Cochran 1980). To determine whether the chlorophyll variance had a periodic component (i.e., the variance was attributed to season) the Fishers Kappa Statistic was used to test the null hypothesis that the data were fr om a normal distribution (JMP 2007). If the Fishers Kappa Statistic showed there was a period component in the variance of the chlorophyll concentrations, the spectral density plot was additionally examined to visually determine if the variance was attributed to season. A peak at 12 (i.e., 12 months) indicated the variance in the chlorophyll data series was attributed to season. Typically, the variance identified by the spectral density plot of the time series analysis is the variance that is removed prior to ARMA/ARIMA time series model analysis. The goal of the second approach used to identify seasonal patterns in chlorophyll concentrations was to examine the monthly variance within the individual lake and compared the monthly variances among the years of record. The monthly mean percent (mean %) difference from the annual mean was calculated for the individual lake. Specificall y, t he monthly mean % differences were determined by subtracting the annual mean value from the monthly mean, dividing this value by the annual mean, and multiplying by 100. A positive % difference indicated the mean chlorophyll concentration for the month was greater than the annual mean for the respective year. A negative % difference indicated the mean chlorophyll concentration for the month was greater than the annual mean for the respective year. T he monthly mean % difference was summarized among years and then among the 27 Florida lakes by month. T he 95%

PAGE 101

101 confidence intervals were calculated around the monthly mean. A one way analysis of variance was used (the residuals for each lake were homoscedastic) to determine whether a significant difference exis ted among t he mean % difference among the month s of the year using the summarized data for the 27 Florida lakes. The goal of the third approach used to identify seasonal patterns was to evaluate the frequency of occurrence of extreme chlorophyll values. E xtreme chlorophyll values were defined as 1) the maximum chlorophyll value that occurred in each month of each year of record and 2) a mean chlorophyll value that exceeded the grand mean by double in value for each month and year of record. The frequency of occurrence of the maximum chlorophyll values and the frequency of occurrence of the chlorophyll values exceeding the grand mean by double were summarized by month among years for the 27 Florida lakes (Brown et al. 1998). Histograms illustrating the frequency of occurrence of the maximum chlorophyll values and the chlorophyll values exceeding the grand mean by double were generated to identify the months with a higher probability of experienc ing extreme chlorophyll values, respectively. Due to the continuous and lengthy record of the examined dataset, more extreme chlorophyll conce ntrations were likely captured. Pareto distributions (Pareto 1897) were used to describe whether there was a change in the occurrence of extreme chlorophyll events over the 20 plus year record of examination. Pareto distributions, which have great potential to describe a wide range of aquatic variables (Vidondo et al. 1997), use a semi logarithmic (base 10) plot where the slope of the regression line corresponds to the probability function. The percent number of records greater than x (the corresponding chlorophyll concentration) represented the dependent variable and was

PAGE 102

102 plotted against the corresponding logarithmic (base 10) chlorophyll value (x) that represented the indep endent variable. Linear regression analysis was fit to these data and the slope value determined. The slope value was determined for every year of record for the individual 27 Florida lakes. The annual slope values were plotted against the corresponding year and linear regression analysis was used to determine whether the re was a trend in the slope values over the examined period of record for the individual lake. I f linear regression analysis indicated a significant increasing trend in the annual slope val ues, then it was concluded that there was an increase in the frequency of occurrence of extreme chlorophyll concentrations over the examined period of record. If linear regression analysis of the annual slope values indicated a significant decreasing trend, then it was concluded that there was a decrease in the frequency of occurrence of extreme chlorophyll concentrations over the examined period of record. Climate R elationships In lake water temperature data (i.e., monthly or annual data) were not available for the examined Florida lakes. Water temperatur es have been correlated with air temperatures across Florida lakes (Coenen 2005); therefore, air temperature data were used. Monthly mean air temperature data were obtained from the Florida Climate Center (http://climatecenter.fsu.edu/) with the closet location of collection to the individual lake. The longterm mean (i.e., 24 year period) and the associated 95% confidence intervals, w ere calculated among the 27 Florida lakes for each month of the year. The 24year period corresponded to the longest record of trophic state variable collection and used to make consistent water temperature comparisons among the 27 Florida lakes The long term monthly means were determined monthly to examine the

PAGE 103

103 relationship with the monthly mean chlorophyll concentrations among the 27 Florida lakes. Th e rainfall data were gathered by the monthly sum of rainfall (cm) and obtained from the Florida Climate Ce nter using the closest location of collection to the individual lake. The sum of r ainfall dat a obtained matched the month of trophic state variable collection for the individual, 27 Florida lakes. The long term mean (i.e., the length of years depends on the record for the individual lake) was calculated for the individual lake for each month of the year. The long term, mean chlorophyll concentration was also calculated for the individual lake for each month of the year. The long term monthly sum of rainfall and the long term monthly chlorophyll concentrations were compared for the individual, 27 Florida lakes. Comparisons between temperature and rainfall data were not completed for the 27 Florida lakes because the location of the collection sites for tempera ture and rainfall data varied for some of the individual 27 Florida lakes. Results Magnitude of P hytoplankton B iomass The chlorophyll concentration database of 27 Florida lakes included 19,836 individual measure ments, 6,612 monthly mean measurements, and 551 lakeyears, providing a comprehensive inventory of phytoplankton biomass for subtropical Florida lakes. The distribution of the monthly mean chlorophyll concentrations for the 27 Florida lakes ranged three orders of magnitude with 75% of the measurements below 23 g/L. The distribution of chlorophyll concentrations for the 551 lakeyears of chlorophyll data (i.e., annual means) also ranged three orders of magnitude with 75% of the values below 24 g/ L, but the annual maximum chlorophyll values were about half (29 2 g/L) that observed among the monthly mean values (436 g/L).

PAGE 104

104 Variability of Seasonal Patterns a cross F lorida Lakes S easonal period components were identified with spectral density analysis i n the chlorophyll concentrati ons for each of the 27 Florida lakes. Seasonal patterns were identified in the chlorophyll variance, which would have been removed prior to ARMA/ARIMA t ime series modeling for 21 of the 27 subtropical, Florida lakes (Table 43 ). The seasonal patterns were statistically identified by the Fishers Kappa Test of variance and also visually identified by a peak (i.e., high varia nce ) of the chlorophyll concentrations at 12 months (Figure 4 1) The re were six lakes with no seasonal pattern in the chlorophyll concentration time series, meaning the variance was not statistically different from zero and there was no visual peak at 12 months. Examination of the mean % difference in monthly chlorophyll concentrations refined identification of seasonal patterns in the variability of chlorophyll concentrations by quantifying monthly changes over an annual cycle (Figure 4 2). The monthly mean % difference among the 27 Florida lakes ranged from 18% (January) to 19% (September). The cooler months (i.e., November through May) had lower chlorophyll concentrations (negative % mean difference), while the warmer months (June through October) had higher chlorophyll concentrations (positive % mean difference). A oneway analysis of variance (ANOVA) with multiple comparisons showed August and September monthly mean % differences were significantly different from all other months of the year a mong the 27 Florida lakes The frequency of occurrence of extreme maximum chlorophyll events provided an alternativ e detection of seasonal patterns in the variability of chlorophyll concentrations. Among the population of examined Florida lakes, the months with the higher number of maximum chlorophyll events occurred during the months of June

PAGE 105

105 through October (Figure 4 3), identical to the months with the positive % mean differences in monthly chlorophyll concentrations. The highest frequency of maximum chlorophyll events occurred in October. The same seasonal pattern was also present among all Florida lakes when chlorophyll concentrations were characterized using a more conservative estimate of an extreme chlorophyll value. The conservative estimate, the frequency of occurrence of chlorophyll values that exceeded the doubling of the grand mean of an individual lake, als o showed the higher frequency of events during the months June through October, with September having the highest number of values that exceeded the grand mean by double (Figure 4 3). Variability of Seasonal Patterns by Lakeyear Trophic State C ategory T he annual seasonal patterns of monthly chlorophyll concentrations were different in hypereutrophic lakeyear classifications. Seasonal patterns were evident as positive monthly % mean differences in chlorophyll concentrations occurred during the months of June through October across oligotrophic lakeyears classifications (N=79 lake years), mesotrophic (N=171 lakeyears), and eutrophic (N= 197 lakeyears) (Figure 4 4 A C). The seasonal pattern in lakeyears classified as oligotrophic mesotrophic, and eutrophic visually exhibited the same seasonal curve exhibited among the 27 Florida lakes. The monthly % mean differences in chlorophyll concentrations of the hypereutrophic lakeyears (N= 104 lakeyears) similarly have high chlorophyll concentrations in October, but the annual curve was bimodal with chlorophyll concentrations peaking in April and in October (Figure 4 4 D). The magnitude of chlorophyll concentrations varied among the lakeyear trophic state classifications. The range of monthly % mean difference in oligotrophic lakeyears ranged from 16% (February) to 19% (September), from 21% (January) to 26%

PAGE 106

106 (September) in mesotrophic lakeyears, from 20% (January) to 23% (August) in eutrophic lakelakes, and from 12 % (January) to 9% (October) in hypereutrophic lakeyears. The range of the monthly % mean differences in oligotrophic and hypereutrophic lake years was smaller than the monthly % mean difference in mesotrophic and eutrophic lake year classifications. Not only does this r esult suggest mesotrophic and eutrophic lakeyear classifications experience a wider range of chlorophyll concentrations on an annual basis but also eutrophic lakeyear classifications had more months with statistically higher mean % differences in chloro phyll concentrations. Specifically, for oligotrophic, mesotrophic, and the eutrophic lakeyears the mean % differences were above the annual mean during the warmer months of June through October and below the annual mean during the remaining months (Figur e 4 4 A C). The ANOVA and a multiple comparison test s showed the months August and September (oligotrophic), September (mesotrophic), and June, July, August, and September (eutrophic) had mean % difference in chlorophyll significantly different from the ot her months. The monthly % mean difference in the hypereutrophic lake years exceeded the annual mean for half the year (i.e., March, April, August, September, October and November) and was below the annual mean for the half the year (i.e., January, February May, June, July, and December ). The ANOVA and multiple comparison test s showed the % mean difference in chlorophyll concentrations in January was significantly less than other months The frequency of extreme events which showed seasonal patterns also differed across trophic lake years classifications. The frequency of maximum value extreme events varied annually among months in oligotrophic, mesotrophic, and eutrophic lake-

PAGE 107

107 years with the highest frequency of events occurring during the months June thr ough October (Figure 4 5 A and 4 5 B). The month with the highest frequency of maximum chlorophyll events shifted later in the year with a shift towards a more biologically productive system. Specifically, the highest frequency of maximum chlorophyll values occurred in June in oligotrophic lakeyears, September in mesotrophic lakeyears, and October in eutrophic lakeyears (Figure 4 5 A C). Hypereutrophic lakes did not follow the patterns of the occurrence of extreme chlorophyll events exhibited in the other trophic categories. Instead, the highest frequency of maximum chlorophyll values occurred in April, which exceeded the number of events occurring in any other month by double (Figure 4 5 D). A similar distinction of seasonal patterns of extreme events was shown when extreme events were defined as the chlorophyll value that exceeded the grand mean by double. Oligotrophic, mesotrophic, and eutrophic lakeyears had high frequencies of extreme chl orophyll events during June through October (Figure 4 5 A C). The month with the hi ghest number of extreme events, exceeding the grand mean by double, differed from the months with the highest number of maximum chlorophyll events (Figure 4 5 A C). The high est number of extreme events that exceeded the grand mean by double occurred in October in oligotrophic lakeyears and in September in m esotrophic and eutrophic lakeyears. The frequency of extreme doubling events in hypereutrophic lakeyears showed the s ame pattern as the number of extreme maximum events, with the most extreme events occurring in May (Figure 4 5 D). Temporal Shifts in the Occurrence of Extreme Chlorophyll E vents The occurrence of extreme chlorophyll events, based on the results of the Par eto analysis, did not change annually over the past two decades in the majority ( N= 23

PAGE 108

108 lakes) of the examined Florida lakes (Table 42) Neither extended periods of higher phytoplankton biomass nor an increase in extreme chlorophyll events were identified in 23 lakes. However, a significant lessening of the annual slope value across the years of record was determined from the Pareto analysis in three lakes (i.e., l akes Deerback, Harris, Sarah) indicating these three lakes experienced an increase in the occ urrence of extreme chlorophyll concentrations over the past two decades. A significant increase in the slope value occurred in one lake (i.e., Lake Lorraine) indicated a decrease in the occurrence of extreme chlorophyll ev ents over the past two decades. Climate R elationships Monthly temperature showed the same annual pattern as chlorophyll concentration where months of higher temperature corresponding to months of higher chlorophyll concentrations (i.e., June through October) and the months o f lower temperature corresponding to months of lower chlorophyll concentrations (i.e., November through May) (Figure 46). The amount of rainfall was related to seasonal patterns in chlorophyll concentrations with months of higher rainfall corresponded to higher chlor ophyll concentrations (i.e., June through October) and months of lower rainfall corresponded to lower chlorophyll concentrations (i.e., November through May). There were some lakes where month(s) of high rainfall were followed by high chlorophyll concentrations 13 months thereafter, a lagged effect (Figure 4 7). Overall, among the 27 Florida lakes, there were three groupings of the rainfall chlorophyll patterns : 1) monthly chlorophyll concentrations either increased with increased monthly rainfall amounts (N= 19 lakes), 2) monthly chlorophyll concentrations decreased with increased monthly rainfall

PAGE 109

109 amounts (N=5 lakes), or 3) there was no distinct pattern between monthly chlorophyll and rainfall (N=3 lakes). Discussion Seasonal patterns in chlorophyll concentrations, used as an estimate of phytoplankton biomass, followed cycles of phytoplankton biomass growth and senescence that were recurrent from year to year and synchronous across the examined population of subtropical, Florida lakes. The magnitude of chlorophyll concentrations varied among the individual lakes and there was high inter annual variance around the mean within the individual lake. T here may be some apprehension that a subset of 27 Florida lakes was not r epresentative of subtropical, Florida lakes. Examination of annual seasonal patterns in chlorophyll concentrations using a larger population of 193 subtropical, Florida lakes showed a similar annual seasonal pattern as observed across the 27 Florida lakes (Figure 4 8). The difference in the annual seasonal pattern being the period of higher chlorophyll values (i.e., positive mean % differences) was extended by one month (i.e., June through November ) among the population of 193 Florida lakes versus June thro ugh October as identified among the population of 27 Florida lakes. The additional month of observed higher chlorophyll concentrations may be due to the expanded geographic range of lakes or an artifact of an unequal dataset. The 193 l ake dataset did not have consistent monthly samples or the number of years sampled compared to the 27lake dataset. The number of years sampled was found to be of great er importance when identifying seasonal patterns. The inter annual variance around the monthly mean was larger for the population of 193 lakes yet the range of chlorophyll values observed among the 27 lakes was larger than observed among the 193 lakes. Specifically, the

PAGE 110

110 mean coefficient of variation among the 193 lakes was 43% and ranged from 23% to 76% across t he lakes whereas the mean coefficient of variation among the 27 lakes was 33% and ranged from 21% to 84%. These results suggested that more extreme chlorophyll events were captured with the inclusion of more years of sampling. Therefore, the 27 Florida l akes were not only representative of subtropical, Florida lakes, but also the consistent (i.e., monthly) and lengthy (i.e., greater than 20 years) dataset provided sound identification of seasonal patterns in phytoplankton biomass. Overall, the 193and 27 lake datasets show ed an extended period of phytoplankton growth (i.e., high chlorophyll concentrations) over an annual cycle from July through October Understanding the temporal and spatial structure of seasonal variability in the chlorophyll concentrat ions of subtropical Florida lakes was best described by the annual climate cycle. The climatic influence on seasonal dynamics is well documented in lakes with phytoplankton responses directly linked to changes in solar radiation and temperature (Wetzel 2001). The 27 examined Florida lakes were located in the humid subtropical climate zone (using the Kppen climate zones) characterized by monthly average temperature above 18 C and a rainy season from June through September (Henry 1998). The annual seasonal patterns in the examined lakes reflected the climatic characteristics of the subtropical region with higher phytoplankton biomass persisting over an extended period of the year (i.e., June through October) that closely followed annual mean monthly air temperature (Figure 4 6). Terrestrial plants in Florida have an al l year growing season (USDA 2012), meaning that depending on the plant species, something is always growing in Florida. This idea transfers to aquatic systems as

PAGE 111

111 warmer lakes have prolonged growi ng seasons with a greater probability of prolonged algal bloom s, or extreme chlorophyll events (Jeppesen et al. 2007b). In the exam ined Florida lakes, maximum chlorophyll concentrations were found to occur at any time during the year over the examined per iod of record. The highest monthly chlorophyll concentrations, however, consistently occurred during the months where the air temperature exceeded 23 C, an ideal temperature for phytoplankton growth (Moss 1973). These months (i.e., air temperature exceedi ng 23 C) correspond ed to months of higher solar radiation, indicating the seasonal patterns in subtropical lakes, like temperate lakes (Wetzel 2001), are driven by annual solar radiation and temperature cycles. There is contradicting evidence that tempera ture variation within the State of Florida affects annual seasonal patterns in phytoplankton biomass. Beaver et al. (1998) suggested lake biological processes differ across temperature differences observed in the State of Florida, from north Florida to south Florida (e.g., annual mean maximum temperatures in Jacksonville and Key West differ by 0.6 C, Coenen 2005). Examination of a large population of Florida lakes, however, did not find any significant changes in annual chlorophyll concentrations due to t emperature changes across Floridas latitudinal gradient (Coenen 2005). Brown et al. (1998) further suggested that even light and temperature conditions were similar throughout the year and across Florida lakes. The results of this chapter show ed the influ ence of temperature on seaso nal patterns, but the temperature had a greater influenced on chlorophyll concentrations over an annual cycle than latitude differences across the State of Florida Other temperature driven seasonal patterns support this conclus ion as well. For

PAGE 112

112 example, temperature driven seasonal patterns in dissolved oxygen and consequent fish kills have been documented across the State of Florida over an annual cycle (Hoyer et al. 2009). Rainfall cycles climatically described the annual vari ability in the chlorophyll concentrations for the examined subtropical, Florida lakes. High rainfall events have been linked to increased input of nutrients, sediments, and dissolved organic carbon in many lake systems (Deevey 1988; Gaiser et al. 2009), but the response of phytoplankton biomass to rain driven fluctuations in nutrients, sediments, and dissolved organic carbon is not consistent among lakes. W hen rainfall and phytoplankton biomass relationships were examined on an annual basis among the 27 examined Florida lakes phytoplankton biomass consistently responded to high rainfall amounts with high biomass levels over the 20plus year record in many of the examined subtropical, Florida lakes (Figure 4 7). An inverse relationship of phytoplankton biomass and rainfall was identified in five of the lakes; months with low phytoplankton biomass and high rainfall (Figure 4 7). The different annual relationships of phytoplankton biomass and rainfall may be dri ven by differences in the water entering an individual lake from the surrounding watershed, both surface water and groundwater. T he observed differences in the phytoplankton and rainfall relationships among the 27 Florida lakes may also be a result of flus hing rates. The amount of water that enters a lake per a given amount of time could be high, inhibiting phytoplankton use of available nutrients or could be low, promoting phytoplankton use of available nutrients (Vollenweider 1968). Overall, o n an annual basis, the identified relationships between monthly rainfall amounts and monthl y

PAGE 113

113 phytoplankton biomass provided an understanding of seasonal patterns that may become important if climate patterns change in the future. The lake trophic state concept explained contrasting annual seasonal patterns identified across the examined Florida lakeyears. Seasonal patterns of chlorophyll concentrations and the occurrence of extreme chlorophyll events (i.e., maximum chlorophyll concentrations and values exceeding the grand mean by double) among oligotrophic, mesotrophic, and eutrophic classified lakeyears were similar, following the temperature and rainfall patterns (Figures 4 6 and 4 7). Seasonal patterns of hypereutrophic lakeyears, however, did not follow the seasonal pattern exhibited by the other l ake year trophic categories. Rather, the seasonal patterns in hypereutrophic lakeyears were more typical of seasonal patterns observed in a temperate lake with peak chlorophyll concentrations occurring in April and Oc tober like the spring and fall overturn of temperate lakes. The mechanisms driving the phytoplankton biomass peaks turn over events in the spring (i.e., April/May) and fall (i.e., September/October) are initiated by an upwelling of hypolimnetic phosphorus and disruption of the thermocline (N rnberg 1985; Marshall and Peters 1989). Although temporary stratification (Lewis 1973) does occur in some Florida lakes, lake mechanisms associated with stratification most likely do not drive the chlorophyll peaks observed in the Florida hypereutrophic lakeyears as the examined Florida lakes are shallow (< 3m) and polymictic. Differences in temperate and subtropical limnological mechanisms are well acknowledged (Hu tchinson 195 7; Scheffer 1998), yet it is difficult to resolve the reason for contrasting seasonal patterns in the hypereutrophic lakeyears with the available data for the examined lakes. One explanation that could be supported by data was the

PAGE 114

114 effect of wind resuspension of bottom sediments. Relationships of chlorophyll and nutrients have been shown to follow wind patterns in Florida lakes ( Carrick et al. 1993; Havens et al. 1999; Bachmann et al. 2000) and the bottom sediments of productive systems are nutrient rich as these systems accumulate large amounts of organic material, marked by the large nutrient rich flocculent layers (Brenner et al. 1996) Although the resuspension of bottom sediments could limit light availability, it has been shown this effect is lessened in productive Florida systems by the shall ow lake depth, low concentration of light attenuating inorganic particles, or high concentrations of soluble nutrients in the sediments that simulate phytoplankton growth (Havens et al. 1999). Wind driven resuspension of the bottom sediments may explain th e observed annual April and October peaks in chlorophyll concentrations across hypereutrophic lake years because April and October are months with historically high wind velocities (NOAA 1996). An increase in extreme chlorophyll concentrations has been suggested to occur across aquatic ecosystems as changing global climatic patterns are projected to increase growing conditions (Vitovesk et al. 1997). The three lakes where an increase in occurrence of extreme chlorophyll concentrations was identif ied over the past two decades would support this hypothesis. D espite the strong link recognized between the seasonal influence of the phenological traits temperature and rainfall on seasonal patterns of phytoplankton biomass, the Pareto analysis showed no change in the frequency of occurrence of elevated chlorophyll concentrations over the years of record in 23 (85%) of the examined subtropical, Florida lakes. One consideration relevant to the importance of regional climate on shifts in the occurrence of extreme chlorophyll

PAGE 115

115 events is that the Florida phytoplankton levels reach maximum capacity, meaning phytoplankton reach a threshold of self shading and light limitation (Agust et al. 1990). At levels of maximum capacity, the percent biomass contributi on of phytoplankton to the community plateaus, which is dependent on the algal species and trophic state of the Florida lake (Duarte et al. 1992) and offers a future consideratio n for future examination of the shifts in the occurrence of ex treme chlorophy ll events over a period of record. It is evident that seasonal patterns i n Florida lakes are affected by climate as estimated by temperature and rainfall. With the exception of the annual seasonal patterns identified by hypereutrophic lakeyear category, the annual seasonal patterns in phytoplankton biomass do not follow the typical, annual seasonal patterns of temperate lakes with peaks in phytoplankton biomass occurring with the spring and fall turnover events S ubtropical lakes demonstrate an extended g rowing period of phytoplankton biomass (i.e., June through October) where extreme chlorophyll events could occ ur during any month of the year. Comparisons between subtropical and temperate annual seasonal patterns in phytoplankton biomass indicate differences occur at the largescale and latitudinal and longitudinal considerations should be considered. This study further highlights the importance of defining seasonal patterns and incorporates not remove the variance due to seasonal patterns Understanding that natural seasonal variability is an important determinant of recurrent and synchronous lake dynamics and that this seasonal variability may differ depending on the scale of analysis will enhance the understanding of limnology.

PAGE 116

116 T able 4 1. Summary st atistics of monthly chlorophyll sam ples (g/L) collected over a 20 plus year period for 27 Florida lakes. County Lake N samples Mean Chlorophyll Min imum Chlorophyll Max imum Chlorophyll Standard error C oefficient of Variation Alachua Alto 252 11.3 2.7 57.3 0.4 54 Alachua Little Orange 240 18.5 2.7 148.7 1.0 83 Alachua Little Santa Fe 276 9.7 1.0 54.7 0.5 78 Alachua Santa Fe 276 8.1 1.3 37.3 0.3 71 Alachua Wauberg 240 96.2 29.7 240.3 2.9 46 Hillsborough Brant 240 21.2 1.7 216.0 1.4 100 Hillsborough Magdalene 240 4.3 1.0 12.0 0.1 45 Lake Beauclaire 240 169.1 38.7 435.7 4.5 41 Lake Crooked 240 8.6 2.0 36.0 0.4 68 Lake Dora East 240 160.4 26.0 344.7 3.8 36 Lake Dora West 240 148.3 45.0 310.7 3.1 32 Lake Grasshopper 240 2.8 1.0 14.7 0.1 77 Lake Harris 240 56.6 4.0 121.3 1.5 42 Lake Lorraine 240 23.4 1.7 105.0 1.4 90 Lake Sellers 240 1.6 0.1 7.7 0.1 69 Marion Charles 240 6.4 0.3 296.7 1.3 314 Marion Deerback 252 4.7 1.0 21.7 0.2 55 Marion Eaton 240 5.8 1.0 41.7 0.3 94 Marion Halfmoon 240 9.1 2.0 20.7 0.2 36 Orange Georgia 252 5.2 0.1 28.3 0.2 57 Orange Giles 240 32.2 4.3 125.3 1.3 64 Orange Ola 240 4.1 1.0 14.0 0.1 54 Orange Sarah 240 13.5 2.0 59.0 0.6 66 Putnam Como 252 2.5 0.7 10.0 0.1 55 Putnam Higgenbotham 240 3.3 1.0 14.3 0.1 57 Putnam Star 252 7.0 1.3 30.0 0.3 58 Putnam Winnott 240 4.1 1.0 31.3 0.2 81

PAGE 117

117 Table 42. Linear regression analysis of slope values by year. The annual slope values were derived from determination of the percent number of records greater than a given chlorophyll concentration against the corresponding logarithmic (base 10) transformed chlorophyll concentration. Significant liner relationships indicate a change in the frequency of occurrence of extreme chlorophyll concentrations over the examined period of record for the individual 27 Florida lakes. County Lake Slope R 2 Alachua Alto 0.38 0 Alachua Little Orange 1.66 0.05 Alachua Little Santa Fe 2.85 0.08 Alachua Santa Fe 2.8 0.04 Alachua Wauberg 0.18 0 Hillsborough Brant 5.5 0.1 0 Hillsborough Magdalene 3.42 0.06 Lake Beauclaire 3.5 0.04 Lake Crooked 2.68 0.11 Lake Dora East 7.24 0.07 Lake Dora West 3.74 0.01 Lake Grasshopper 0.1 0 Lake Harris 0.25 Lake Lorraine 0.42 Lake Sellers 13.75 0.16 Marion Charles 0.24 0 Marion Deerback 0.21 Marion Eaton 0.71 0.04 Marion Halfmoon 2.244 0.05 Orange Georgia 2.243 0.05 Orange Giles 1.61 0.04 Orange Ola 2.86 0.04 Orange Sarah 0.23 Putnam Como 2.1 0.03 Putnam Higgenbotham 3.48 0.06 Putnam Star 1.7 0.04 Putnam Winnott 4.4 0.11

PAGE 118

118 Table 43.Statistical identification of periodic component of variability in chlorophyll concentrations by spectral density analysis generated by the time series model analysis, by the Fishers Kappa Test (p values listed). V isual identification (Y= yes and N= no of a peak at 12 months in the chlorophyll variance indicated the individual Florida lake exhibited a seasonal periodic component across the examined period of record. County Lake Fisher's Kappa p value Visual Peak Alachua Alto 0 Y Alachua Little Orange 0 Y Alachua Little Santa Fe 0 Y Alachua Santa Fe 0 Y Alachua Wauberg 0 Y Hillsborough Brant 0 Y Hillsborough Magdalene 0 N Lake Beauclaire 0 N Lake Crooked 0 N Lake Dora East 0 Y Lake Dora West 0 Y Lake Grasshopper 0 Y Lake Harris 0 N Lake Lorraine 0 N Lake Sellers 0 Y Marion Charles 0 Y Marion Deerback 0 Y Marion Eaton 0 Y Marion Halfmoon 0.001 Y Orange Georgia 0 Y Orange Giles 0.005 N Orange Ola 0 Y Orange Sarah 0 Y Putnam Como 0 Y Putnam Higgenbotham 0 Y Putnam Star 0 Y Putnam Winnott 0 Y

PAGE 119

119 Figure 4 1. S pectral density plot generated f rom time series model analysis for monthly chlorophyll concentrations over a 21year period in Lake Alto located in Alachua County, Florida. The peak at a period of 12, indicates there is a seasonal component in the c hlorophyll concentrations in Lake Alto. 0 10 20 30 40 50 60 70 80 Spectral Density 0 10 20 30 40 50 60 70 Period

PAGE 120

120 Figure 42. M ean percent ( % ) difference of monthly chlorophyll concentrations over an annual cycle for 27 subtropical, Florida lakes. The bars represent the 95% confidence intervals around the mean of the monthly mean % difference in chlorophyll concentrations from the annual mean. Positive differences indicate concentrations greater than the mean and negative differences indicate conc entrations less than the mean.

PAGE 121

121 Figure 43. F requency of occurrence of extreme chlorophyll events represented as the maximum chlorophyll concentrations (light grey bars) and the chlorophyll concentrations exceeding two times the grand mean (dark grey bars) summarized for each month among the years sampled for the 27 Florida l akes (N = 611 total maxima values and N = 373 values exceeding the grand mean by double).

PAGE 122

122 Figure 44. M ean percent ( % ) difference of monthly chlorophyll concentrations calculated over an annual cycle by classification into lakeyear trophic state categories A) oligotrophic, B) mesotrophic, C) eutrophic, and D) hypereutrophic classification. The bars represent the 95% confidence intervals associated with the mean for the monthly mean % difference in chlorophyll concentrations.

PAGE 123

123 Figure 44. Continued

PAGE 124

124 Figure 45. F requency of occurrence of extreme chlorophyll events represented as the maximum chlorophyll concentrations (light grey bars) and the chlorophyll concentrations exceeding the grand mean by double (dark grey bars) summarized for each month among the years sampled by classification into lake year trophic categories A) oligotrophic (N= 948 total lake years) B) mesotrophic (N = 2051 total lake years ), C) eutrophic (N=2365 total lake years ), and D) hypereutrophic (N=1248 total lake years.

PAGE 125

125 Figure 45 C ontinued

PAGE 126

126 Figure 46. Monthly air temperature (C) data averaged over a 24year period for the five nearest collection site s (solid line) and the corresponding mean % difference of monthly chlorophyll concentrations over an annual cycle (dotted line) for the examined 27 Florida lakes. -20 -15 -10 -5 0 5 10 15 20 Mean % Difference Chlorophyll Concentrations 10 15 20 25 30 Temperature (C) 1 2 3 4 5 6 7 8 9 10 11 12 Month

PAGE 127

127 Figure 47. A verage monthly rainfall sum (cm) (dotted line) and the average monthly ch lorophyll concentrations (g/L) (solid line) calculated among the annual data for the individual Florida lake, which are represented as Lake Name (County of location).

PAGE 128

128 5 10 15 20 Rainfall Sum (cm) 7 8 9 10 11 12 13 14 15 16 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 10 15 20 25 30 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 5 6 7 8 9 10 11 12 13 14 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 5 6 7 8 9 10 11 12 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 70 80 90 100 110 120 130 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 10 15 20 25 30 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month Alto (Alachua) Little Orange (Alachua) Brant (Hillsborough) Wauberg (Alachua) Little Santa Fe (Alachua) Santa Fe (Alachua)

PAGE 129

129 5 10 15 20 Rainfall Sum (cm) 3.5 4 4.5 5 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 130 140 150 160 170 180 190 200 210 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 6 7 8 9 10 11 12 13 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 140 150 160 170 180 190 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 130 135 140 145 150 155 160 165 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 2 2.5 3 3.5 4 4.5 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month Magdalene (Hillsborough) Beauclarie (Lake) Crooked (Lake) Dora East (Lake) Dora West (Lake) Grasshopper (Lake)

PAGE 130

130 5 10 15 20 Rainfall Sum (cm) 35 40 45 50 55 60 65 70 75 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 19 20 21 22 23 24 25 26 27 28 29 30 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 1 1.2 1.4 1.6 1.8 2 2.2 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 0 5 10 15 20 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 2.5 3 3.5 4 4.5 5 5.5 6 6.5 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 2 3 4 5 6 7 8 9 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month Harris (Lake) Lorriane (Lake) Charles (Marion) Deerback (Marion) Eaton (Marion) Sellers (Lake)

PAGE 131

131 5 10 15 20 Rainfall Sum (cm) 8 8.5 9 9.5 10 10.5 11 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 26 28 30 32 34 36 38 40 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 2.5 3 3.5 4 4.5 5 5.5 6 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 9 10 11 12 13 14 15 16 17 18 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 1.75 2 2.25 2.5 2.75 3 3.25 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month Halfmoon (Marion) Ola (Orange) Georgia (Orange) Sarah (Orange) Giles (Orange) Como (Putnam)

PAGE 132

132 5 10 15 20 Rainfall Sum (cm) 2 2.5 3 3.5 4 4.5 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 3 4 5 6 7 8 9 10 11 12 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month 5 10 15 20 Rainfall Sum (cm) 3 4 5 6 7 Chlorophyll (g/L) 1 2 3 4 5 6 7 8 9 10 11 12 Month Higgenbotham (Putnam) Star (Putnam) Winnott (Putnam)

PAGE 133

133 Figure 48. M ean percent ( % ) difference of monthly chlorophyll concentrations over an annual cycle for a population of 193 Florida lakes (open circles connected by dotted line) and the population of 27 Florida lakes (closed circles connected by solid line). The bars represent the 95% confidence intervals around the me an of the monthly mean % difference in chlorophyll concentrations from the annual mean.

PAGE 134

134 CHAPTER 5 SUMMARY, RECOMMENDATIONS, AND MAJOR CONCLUSIONS Summary and Recommendations Freshwater lakes are sensitive to changes in the environment, such as nutrient lo ading or climatic events, and the response of the lake provides the ability to identify and understand the impact of changes in the surrounding watershed and landscape (Carpenter et al. 2007; Adrian et al. 2009; Schindler 2009). One of the biggest issues scientists, environmental managers, and policy makers currently face is how to assess changes over multiple scales of time and space (Williamson et al. 2009). Using a robust, long term dataset composed of lake trophic state variabl es, statistical methods, r anging from simplistic to complex, were used to identif y lake trophic state trends and proposed alternative trend detection methods ( Chapter 2) A simple alternative approach, which provided comparable results to the more complex statistical models was developed to determine trends in the trophic state variables and to examine spatial clusters of lakes with identified trends for a large populati on of Florida lakes (Chapter 3). C omponents of lake variability were examined by identification of seasonal patterns in phytoplankton biomass and the influence of climate factors, temperature and rainfall, on seasonal patterns (Chapter 4). The results from the above inquiries contributed to improving assessments of lake change in Florida over multiple scales of time and space. Improved assessments of lake change provide a platform to enhance future studies that target understanding the linkages of anthropogenic and natural factors to lake change. Evaluation of simple least squares linear regression, Kendall Tau, and ARMA/ARIMA time series models produced different results of trend detection when the

PAGE 135

135 same data were analyzed (Chapter 2). The ARMA/ARIMA time series models are suggested to best account for variance around the mean and also extreme values common character istics of aquatic time series data. Compared to the other evaluated statistical methods, ARMA/ARIMA time series models provided a conservative estimate of lakes exhibiting decadal scale trends in the examined trophic state variables and population of 27 Fl orida lakes. The ARMA/ARIMA time series models, however, are statistically complexity and have strict data requirements. Therefore, an alternative, mo dified linear regression method was developed, offering a statistical meaningful (Bryhn and Dimberg 2011) and predictively powerful approach (Prairie 1996) that provided similar results as the ARMA/ARIMA time series model analysis. The divergent d etermination of long term trends in lake trophic state variables using various statistical analyses elucidates the point that statistical methods are tool s useful to guide interpretation. Lakes are variable in nature and sometimes statistical determination of trends may not capture the variability in a data time series appropriately Thus, it is important to plot and examine data prior to using statistical tools. The examination of long term trends in lake trophic state variables is essential to describe a lakes behavior, but frequently not completed for a large populations of lakes. Determination of long term tre nds within and among lakes facilitates a context in which future limnological studies and evaluation of environmental management options can be accomplished. A comprehensive evaluation of trophic state variable trends for a large population of 193 Florida lakes was completed (Chapter 3). Due to postulated worsening conditions of freshwater systems in response to shifts in global climate (Kernan 2010), trends of degradation (i.e., increases in total phosphorus, total nitrogen,

PAGE 136

136 and chlorophyll concentrations and decreases in water clarity measurements) were expected to be documented in a number of the 193 examined Florida lakes over the decadal scale ( 15 years) period of record (Chapter 3). F or the population of 193 Florida lakes, increasing trends in total phosphorus (21%), increasing trends in total nitrogen (26%), increasing trends in chlorophyll concentrations (12%), and decreasing trends in water clarity measurements (18%) were determined using the alternative modified linear regression method (Chapter 2 ). T rends of improvement (i.e., decreases in total phosphorus, total nitrogen, and chlorophyll concentrations, and increases in water clarity measurements) were found in 7%, 6%, 7%, and 4% of the population of examined Florida lakes, respectively. The m ajo r conclusion is that not many of the examined Florida lakes experienced decadal scale trends in total phosphorus, total nitrogen, chlorophyll concentrations or water clarity measurements. The lakes with identified decadal scale trophic state trends (Chap ter 3) s hould be recognized and offer a valuable opportunity to focus future research and management efforts. For example, the nine lakes where trends of degradation were documented in all of the four trophic state variables or the three spatial clusters of lakes with similar trends among the trophic state variables (Chapter 3) should help to focus research and management effort s. Directing research and management efforts to lakes of interest or clusters of lakes of interest, wh ere decadal scale trends in the trophic state variables have been documented, would enhance allocation of time, money, and resources. The dynamic nature of lakes cofounded by the codependence of limnological mechanisms limits recognition of the factors influencing identified change in lakes. The influence of seasonal patterns in phytoplankton biomass, an important aspect of lake

PAGE 137

137 variability is often disregarded or even removed prior to statis tical trend analysis For the population of 27 Florida lakes with at least 20 years o f consistent monthly data, seasonal patterns in phytoplankton biomass (chlorophyll concentrations were used as an esti mate of phytoplankton biomass ) were identified (Chapter 4). The seasonal patterns in phytoplankton biomass were found to follow cycles of phytoplankton biomass growth and senescence that were recurrent and synchronous (Chapter 4). Annual elevated chlorophyll concentrations occurred June through October, an extended length of the year compared to the peak growing period of phytoplankton biomass in temperate lakes (Marshall and Peters 1989). There were contrasting seasonal patterns, which were best explained by the classification of waters by lakeyear trophic category of chlorophyll concentrations (Forsburg and Ryding 1980). Oligotrophic, mesotrophic, and eutrophic classified waters experienced patterns of elevated chlorophyll concentrations during June through October. Hypereutrophic classified water, however, showed the largest range in chlorophyll concentrations with elevated chlorophyll concentrations occurring in April and October. The reason for the difference in the seasonal patterns observed in the hy pereutrophic classified waters was not determined, but warrants an important question to address The overall results do illustrate the con tribution of seasonal components to lake variability and the importance of incorporating seasonal variability in to lake assessments and statistical analyses. Incorporation and interpretation of seasonal variability in lake assessments depends on the scale of analysis. Annual seasonal patterns of phytoplankton biomass, as measured by chlorophyll concentrations, are driven by annual climate cycles of solar radiation (Wetzel 2001), air temperature (Chapter 4), and rainfall (Chapter 4). Climate

PAGE 138

138 patterns vary w ith latitude and longitude suggesting seasonal patterns of phytoplankton biomass would vary at the large scale of analysis, along changes in latitude and longitude. Latitudinal and longitudinal variation is essential to recognize when comparing seasonal patterns of phytoplankton biomass across lakes. For example, examination of Florida lakes, indicated fluctuations of monthly chlorophyll values followed monthly air temperature changes with the higher chlorophyll levels occurring during the months where temperatures exceeded 23 C (Chapter 4) as temperatures of 23 C and above are optimal for phytoplankton growth (Moss 1973). Due to varying latitude, Florida lakes experience longer annual periods of phytoplankton growth versus temperature lakes where air temperatures do not reach or exceed 23 C for as many months, meaning a decreased annual period of phytoplankton growth, comparatively. The influence of climate warming conditions on phytoplankton has increased in interest because changing temperatures are anti cipated to alter levels of phytoplankton production and community structure in lakes (Hering et al. 2010). Warming temperatures have already been documented to change the limnology of lakes, such as the documented shift in patterns of ice out occurring ear lier in the year (Magnuson et al. 2000). The influence of warming temperatures on seasonal patterns of phytoplankton biomass may be of a greater magnitude for lakes located in northern latitudes over southern latitudes. The identification of annual seasonal patterns in phytoplankton biomass, link of seasonal patterns to annual temperature and rainfall patterns, and lack of determination of an increase in the occurrence of extreme chlorophyll events (i.e., only three of the 27 Florida lakes experienced an increase in the occurrence of extreme chlorophyll events over the past 20plus years) provide anecdotal evidence of a

PAGE 139

139 potentially greater impact of climate warming on seasonal patterns of phytoplankton in northern latitude lakes. It would be interesting, therefore, to see whether the same results were obtained in lakes located in more northern latitudes. Overall, the research completed outlines an alternative, modified linear regression method to detect decadal scale trend in trophic state variable time series data, offers suggestions as to where to focus future research efforts, and provides a framework t o address global factors driving lake changes. These contributions, however, would not have been possible without the involvement of citizen scientists whose efforts developed an excellent, robust long term database of trophic state variables available for a large population of lakes. The involvement of citizen scientists, like the Florida LAKEWATCH volunteers, allows scientists to answer broad ecological questions that may otherwise not have been possible (Ecological Society of America 2012). The use of ci tizen scientists to monitor and gather long term databases are invaluable to science and society. Major Conclusions 1 Different statistical methods used to detect trends in time series data provide different results. 2 For the population of 193 Florida lakes increasing decadal scale trends were detected for total phosphorus (21%) total nitrogen (26%) and chlorophyll concentrations (12%) and decreasing decadal scale trends for water clarity measurements. 3 Annual patterns in phytoplankton biomass were found to follow cycles of phytoplankton biomass growth and senescence that were recurrent and synchronous with elevated concentrations occurring in June through October.

PAGE 140

140 4 Hypereutrophic classified waters showed a wide range of chlorophyll concentrations with elev ated concentrations occurring in April and October, which differed from waters classified as oligotrophic, mesotrophic, or eutrophic with elevated concentrations in June through October. 5 Annual patterns of phytoplankton biomass follow monthly air temperatures with higher chlorophyll concentrations occurring during the months exceeding 23 C. 6 Annual patterns of phytoplankton biomass showed either similar or inverse relationships to monthly rainfall patterns.

PAGE 141

141 LIST OF REFERENCES Abaurrea, J., J. Asn, A. C. Cebrin, and M. A. GarcaVera. 2011. Trend analysis of water quality series based on regression models with correlated errors. J ournal of Hydrol ogy 400: 341352 [doi: 10.1016/j.jhydrol.2011.01.049]. Adrian, R., C.M. OReilly, H. Zagarese, S.B. Bain es, D.O. Hessen, W. Keller, D.M. Livingstone, R. Sommaruga, D. Straile, E. Van Donk, G.A. Wehenmeyer, and M. Winder. 2009. Lakes as sentinels of climate change. Limnology and Oceanog raphy 54(6, part 2):22832297 [doi:10.4319/lo.2009.54.6_part_2.2283]. Aka ike, H. 1974. A new look at the statistical model identification. Automatic Control, IEEE Transactions 19: 716723 [doi:10.1109/TAC.1974.1100705] Agust, S., C.M. Duarte, and D.E. Canfield Jr. 1990. Phytoplankt on abundance in Florida lakes: evidence for the frequent lack of nutrient limitation. Li mnology and Oceanography 35: 181183 [doi:10.4319/lo.1990.35.1.0181] Bachmann, R.W., M.V. Hoyer, and D.E. Canfield, Jr. 1999. The restoration of Lake Apopka in relation to alternative stable states. Hydrobiologi a 394:219232. Bachmann, R.W., C.A. Horsburg, M.V. Hoyer, L.K. Mataraza, and D.E. Canfield, Jr. 2002. Relations between trophic state indicators and plant biomass in Florida l akes. Hyd robiologia 470:219 234 [doi:10.1 023/A : 1015660922494]. Bachmann, R.W., D.L. Bigham, M.V. Hoyer, and D.E. Canfield, Jr. 2012a. Factors determining the distributions of total phosphorus, total nitrogen, and chlorophyll a in Florida lakes. Lake and Res ervoir Management 28:1026 [doi: 10.1080/07438141.2011.646458]. Bachmann, R.W ., D.L. Bigham, M.V. Hoyer, and D.E. Canfield, Jr. 2012b. A strategy for establishing numeric nutri ent criteria for Florida lakes. Lake and Reservoir Management 28:8491 [doi: 10.1080/07438141.2012.667053]. Battarbee, R.W. 2010. Aquatic ecosystem variability and climate change a palaeoecological perspective. In Climate change impacts on freshwater ecosystems. (Eds. Kernan, M., R.W. Battarbee, and B. Moss.) Wiley Black well Oxford, United Kingdom and New Jersey. Beaver, J.R., T.L. Crisman, and J.S. Bays. 1981. Thermal regimes of Florida lakes: a comparison with biotic and climatic transitions. Hydrobiologia 83:267 273 [ doi:10.1007/BF00008277] Bendat, J. S. and A. G. Piersol. 2010. Random data, 4th ed. John Wiley & Sons, Inc. Hoboken, New Jersey.

PAGE 142

142 Box, G .E.P. and G.E. Jenkins 1970. Time series analysis forecasting and control, HoldenDay San Francisco, California. Box, G.E.P. and G.M. Jenkins, 1976, Time series analysis forecasting and c ontrol 2nd ed. HoldenDay, San Francisco, California. Brenner, M., T.J. Whitmore, and C.L. Schelske. 1996. Paleolimnological evaluation of historic trophic state conditions in hypereutrophic Lake Thonotosassa, Florida USA. Hydrobiologia 331:143 152 [doi:10.1007/BF00025415]. Brown, C.D., D.E. Canfield Jr., R .W. Bachmann, and M.V. Hoyer. 1998. Seasonal patterns of chlorophyll, nutrient concentrations and Secchi disk transparency in Florida lakes. Lake and Reservoir Management 14:60 76 [ doi:10.1080/07438149809354110 ] Brown, C. D., D. E. Canfield Jr R. W. Ba chmann, and M. V. Hoyer. 1999. Evaluation of surface sampling for estimates of chlorophyll, total phosphorus, and total nitrogen concentrations in shallow Florida lakes. Lake and Reservoir Management 15: 121 132 [doi:10.1080/07438149909353956]. Brown, C.D. M.V. Hoyer, R.W. Bachmann, and D.E. Canfield, Jr. 2000. Nutrient chlorophyll relationships: an evaluation of empirical nutrient chlorophyll models using Florida and northtemperate lake data. Canadian Journal of Fisheries and Aquatic Sciences 57:15741583 [doi:10.1139/f00090]. Bryhn, A.C. and P.H. Dimberg. 2011. An operational definition of a statistically meaning ful trend. PLos ONE 6:19 [doi: 10.1371/journal.pone.0019241]. Brylinsky, M. and K.H. Mann. 1973. An analysis of factors governing productivity in lakes and reservoirs. Limnology and Oceanography 18:1 14 [ doi:10.4319/lo.1973.18.1.0001]. Burkholder, J.M., D.A. Dickey, C.A. Kinder, R.E. Reed, M.A. Mallin, M.R. McIver, L.B. Cahoon, G. Melia, C. Brownie, J. Smith, N. Deamer, J. Springer, H.B. Glasgow, and D. Toms. 2006. Comprehensive trend analysis of nutrients and related variables in a large eutrophic estuary: A decadal study of anthropogenic and climatic influences. Li mno logy Oceanogr aphy 51: 463487 [ doi:10.4319/lo.2006.51.1_part_2.0463]. Burns, N. M., J. C. Rutherford, and J. S. Clayton. 1999. A monitoring and classification system for New Zealand lakes and reservoirs. Lake and Reservoir Manage ment 1 5: 255 271 [doi:10. 1080/07438149909354122] Canfield Jr., D.E. and R.W. Bachmann. 1 981. Prediction of total phosphorus concentrations, chlorophyll a and Secchi depths in natural and artificial lakes. Canadian Journal of Fisheries and Aquatic Sciences 38 :414423 [ doi:10.113 9/f81058].

PAGE 143

143 Canfield Jr., D.E., S.B. Linda, and L.M. Hodgson. 1985. Chlorophyll biomass nutrient relationships for natural assemblages of Florida phytoplankton. Water Resources Bulletin 21:381391 [doi:10.1111/j.17521688.1985.tb00148.x]. Canfield Jr D. E., and M. V. Hoyer. 1988. Regional geology and the chemical and trophic state characteristics of Florida lakes. Lake and Reservoir Management 4: 21 31 [ doi:10.1080/07438148809354375]. Canfield Jr D. E., C. D. Brown, R. W. Bachmann, and M. V. Hoyer 2002. Volunteer lake monitoring: testing the reliability of data collected by the Florida LAKEWATCH program. Lake and Reservoir Management 18 : 1 9 doi:10.1080/07438140209353924]. Carlson, R.E. 1977. A trophic state index for lakes. Limnology and Oceanogr aphy 22:361369 [doi:10.4319/lo.1977.22.2.0361]. Carlson, R.E., D.E. Canfield Jr., D.L. Bigham, R.W. Bachmann, M.V. Hoyer, and C.A. Horsburgh. 2012. The Secchi DipIn: expanding our knowledge of Nort h American Lakes. LakeLine 32:3841. Carpenter, S.R., N.F. Caraco, D.L. Correll, R.W. Howarth, A.N. Sharpley and V.H. Smith. 1998. Nonpoint pollution of surface waters with phosphorus and nitrog en. Ecological Applications 8:559 569 [doi:10.1890/10510761(1998)008[0559:NPOSWW]2.0.CO]. Carpenter, S.R. and R. C. Lathrop. 2008. Probabilistic estimate of a threshold for eutrop hication. Ecosystems 11:601613 [ doi:10.1007/s10021008 9145 0]. Chandler, D.C. 1944. Limnological studies of Western Lake Erie IV. Relation to limnological and climatic factors to the phyt oplankto n of 1941. Transactions of the American Microscopical Society 63:203236 [ doi:10.2307/3223145]. Carrick, H.J., F.J. Aldridge, and C.L. Schelske. 1993. Wind influences phytoplankton biomass and composition in a shallow, productive lake. Limnology and Oceanography 38:11791192 [ doi:10.4319/lo.1993.38.6.1179]. Chatfield, C. 2004. The analysis of time series: an introduction, 6th ed Chapman and Hall/CRC LLC, Boca Raton, Florida. Cloern, J.E. and A.D. Jassby. 2008. Complex seasonal patterns of primary producers at the landsea interface. Ecological Letters 11:12941303 [doi:10.1111/j.14610248.2008.01244.x]. Conen, D. 2005. Interannual lake temperature trends as indicators for cli matic change in Florida. Thesis presented to the University of Florida, Gainesville, Florida.

PAGE 144

144 Deevey E.S. 1988. Estimation of downward leakage from Florida lakes. Limnology and Oceanography 3:13081320 [ doi:10.4319/lo.1988.33.6.1308]. Dillon, P.J. and F.H. Rigler. 1975. A simple method for predicting the capacity of a lake for development based on lake trophic status. Journal of Fisheries Resear ch Board of Canada 32:15191531 [ doi:10.1139/f75178]. Duarte, C., and J. Kalff. 1990. Patterns in submerged macrophyte biomass of lakes and the importance of the scale of analysis in the interpretation Canadian Journal of Fisheries and Aquatic Sciences 47: 357 363 [doi: 10.1139/f90037] Duarte, C.M., S. Agust, and D.E. Canfield, Jr. 1992. Patterns in phytoplankton community structure in Florida lakes. Limnology and Oceanography 37: 55 161 [doi:10.4319/lo.1992.37.1.0155] Downing J.A., Y.T. Prairie, J.J. Cole, C.M. Duarte, L.J. Tranvik, R.G. Striegl, W.H. McDowell, P. Kortelainen, N.F. Caraco, J.M. Melack, and J.J. Middelburg. 2006. The global abundance and size distribution of lakes, ponds, and impoundments. Limnology and Oceanography 51:23882397 [doi:10.4319/lo.2006.51.5.2388]. Downing, J.A. 2009. Global limnology: upscaling aquatic services to processes to plane t Earth. Verh. Internat. V erein. Limnol. 30:11491166 [doi: 0368 0770/09/1149 ] Ecological Society of America. 2012. Frontiers in Ecology and the Environment 10 (13 articles on volunteer monitoring in ecology). ( ESRI ) Environmental Systems Research Institute. 2011. ArcGIS Desktop: Release 10. Redlands, C alifornia : Environmental Systems Research Institute. Esterby, S.R. 1997. Review of methods for detection and estimation of trends with emphasis on water quality appli cations, p. 326. In Peters N.E., O.P. Bricker, and M.M Kennedy [eds.], Water Quality Trends and Geochemical Mass B alance. John Wiley & Sons, Inc., Hoboken, New Jersey. Florida Climate Center. 2012. http://climatecenter.fsu.edu/ (FDEP) Florida Department of Environmental Protection. 2008. Integrated water quali ty assessment for Florida: 305(b) report and 303(d) list update. Tallahassee, Florida. Forsberg, C.S. and S.R. Ryding. 1980. Eutrophication parameters and trophic state indices in 30 Swedish waste receiving lakes. Arch Hydrobiologia 80:189207. Gaiser, E.E., N.D. Deyrup, R.W. Bachmann, L.E. Battoe, and H.M. Swain. 2009. Multidecadal climate oscillations detected in a transparency record from a subtropical

PAGE 145

145 Florida lake. Limnology and Oc eanography 54:22282232 [ doi:10.4319/lo.2009.54.6.2228]. Griffith, G.E., D.E. Canfield Jr., C.A. Horsburg and J.M. Omernik. 1997. Lake regions of Florida. EPA (U.S. Environmental Protection Agency ), report R 97/127 Corvallis, Oregon. Hkanson, L., and C. Duar te. 2008. Data variability and uncertainty limits the capacity to identify and predict critical changes in coastal systems A review of key concepts. Ocean and Coastal Management 51: 671 688 [doi: 10.1016/j.ocecoaman.2008.07.003]. Havens, K.E., H.J. Carrick, E.F. Lowe, and M.F. Coveney. 1999. Contrasting relationships between nutrients, chlorophyll a and Secchi transparency in two shallow subtropical lakes: Lake Okeechobee and Ap opka (Florida, USA). Lak e and Reservoir Management 15: 298 309 [doi:10.1080/07438149909354125]. Henry, J.A. 1998. Weather and climate. In: E.A. Fernald and E.D. Purdum (eds.), Water resources atlas of Florida. Florida State University Institute of Science and Public Affairs. Tallahassee, Florida. Hobbie, J. E. 2003. Scientific accomplishments of the long term ecological r esearch program: a n i ntroduction. BioScience 53: 17 20 [doi:10.1641/00063568(2003)053[0017:SAOTLT]2.0.CO;2]. Howden, N.J. K., T. P. Burt, F. Worrall, and M.J. Whelan. 2011. Monitoring fluvial water chemistry for trend detection: hydrological variability masks trends in datasets covering fewer than 12 years. J ournal of Environmental Monitoring 13: 514 521 [doi:10.1039/c0em00722F]. Hoyer, M.V., D.L. Watson, D.J. Willis, and D.E. Canfield Jr. 2 009. Fish kills in Floridas canals, creeks/rivers, and ponds. Journal of Aq uatic Plant Management 47:5356 Hoyer, M.V., N. Wellendorf, R. Frydenborg, D. Bartlette, and D.E. Canfield Jr. 2012. A comparison between professionally (Florida Department of E nvironmental Protection) ad volunteer (Florida LAKEWATCH) collected trophic state chemistry data in Florida. Lake and Reservoir Management 2 8 : 277 281 [ doi: 10.1080/07438141.2012.736016] Hutchinson, G.E. 1957. A Treatise on Limnology. I. Geography, Physics, and Chemistry, John Wiley & Sons, Inc. New York, New York. ( IPCC) Intergovernmental Panel on Climate Change. 2007. Summary for policymakers 1. In climate change 2007: the physical science basis. Contribution of working group 1 to the forth assessment report of the intergovernmental panel on climate change (Eds. Solomon S., M. Manning, Z. Chen, et al.) Cambridge University Press, Cambridge and New York.

PAGE 146

146 James, T.R., K.E. Havens, P. Mc Co rmick, B. Jones, and C. Ford. 2011. Water quality trends in shallow south Florida lakes and assessment of regional versus local forcing functions. Critical Reviews in Environmental Science and Technology 41(S1):576607 [ doi: 10.1080/10643389.2010.530581] Jeppesen, E., M. Sndergaard, M. Meerhoff, T.L. Lauridsen, and J.P. Jensen. 2007a. Shallow lake restoration by nutrient loading reductionsome recent findings and chall enges ahead. Hydrobiologia 584:239252 [doi:10.1007/s10750007 05967]. Jeppesen, E., M. Meerhoff, B.A. Jacobsen, R.S. Hansen, M. Sndergaard, J.P. Jensen, T.L. Lauridsen, N. Mazzeo, and C.W.C. Branco. 2007b. Restoration of shallow lakes by nutrient control and biommanipulation: the successful strategy varies with the lake size and cli mate. Hydrobiologia 581:269285 [doi:10.1007/s1075000605073]. Jeppensen, E., M. Sndergaard, K. Holmgren, et al. 2010. Impacts of climate warming on lake fish community structure and potential effects on ecosystem function. Hydrobiologia 646:7390 [ doi:10.1007/s1075001001715]. Jones, J.R. and R.W. Bachmann. 1976. Predi ction of phosphorus and chlorophyll levels in lakes. Journal of Water Pollution Control Federation 48: 2176 2175. Jones, J.R. and R.W. Bachmann. 1978. Trophic status of Iowa lakes in relation to origin and glacial geology. Hydrobiologia 57: 267273 [ doi:1 0.1007/BF00014580]. Jumars, P.A. 1990. W(h)ither limnology? Limnology and Oceanography 35:12161218 [ doi:10.4319/lo.1990.35.5.1216]. Kendall, M. 1938. A new measure of rank correlation. Biometrika 30: 8189 [doi:10.2307/2332226]. Kernan, M., R.W. Batta rbee, and B. Moss. 2010. Climate change impacts on freshwater ecosystems. Wiley Blackwell, Oxford, United Kingdom. Knowlton, M.F., M.V. Hoyer, and J.R. Jones. 1984. Sources of variability in phosphorus and chlorophyll and their effects on use of lake sur vey data. Water Resources Bulletin 20:397407 [ doi:10.1111/j.17521688.1984.tb04722.x]. Knowlton, M.F., and J.R. Jones. 2006. Temporal variation and assessment of trophic state indicators in Missouri reservoirs: implication for lake monitoring and manage m ent. Lake and Reservoir Management 22:261 271 [doi: 10.1080/07438140609353904]. Kratzer, C. R., and P. L. Brezonik. 1981. A Carlsontype trophic state index f or nitrogen in Florida lakes. Journal of the American Water Resource Association 1 7: 713 715 [doi:1 0.1111/j.17521688.1981.tb01282.x].

PAGE 147

147 Kreindler, D., and C. Lumsden. 2006. The effects of the irregular sample and missing data in time series analysis. Nonlinear Dynamics Psychol ogy Life Sci ences 10: 187214. Kundzewicz, Z.W. and A.J. Robson. 2004. Change in detection in hydrological records a re view of the methodology. Hydrological Sciences Journal 49:7 19 [doi:10.1623/hysj.49.1.7.53993]. Lewis Jr., W.M. 1973. The thermal regime of Lake Lanao (Philippines) and its theoretical implications for tropical l akes. Limnology and Oceanography 18: 200217 [doi:10.4319/lo.1973.18.2.0200]. Loftis, J., C. Taylor, A. Newell, and P. Chapman. 1991. Multivariate trend testing of lake water quality. Water Resources Bulletin 2 7: 461 473 [ doi:10.1111/j.17521688.1991.tb01446.x]. Magnuson, J.J. et al. 2000. Historical trends in lake and river ice cover in the Norther n Hemisphere. Science 289:17431746 [ doi:10.1126/science.289.5485.1743]. Mann, M.E., R.S. Bradley, and M.K. Hughes. 1998. Glo bal scale temperature patterns and climate forcing over the past six centur ies. Nature 392:779787 [doi:10.1038/33859]. Marshall, T.C. and R.H. Peters. 1989. General patterns in the seasonal development of chlorophyll a for temperate lakes. Limnology and Oceanography 34:856867 [ doi:10.4319/lo.1989.34.5.0856]. MClear, R. and R.A. Hay Jr. 1980. Applied time series analysis for the social sciences. Sage Publications, Beverly Hills, California Molot, L., and P. Dillon. 1991. Nitrogen phosphorus ratios and the prediction of chlorophyll in phosphorus limited lakes in central Ontario. Canadian Journal of Fisheries and Aquatic Sciences 48: 140 145 [doi: 10.1139/f91019]. Moss. B. 1973. The influence of environmental factors on the distribution of freshwater al gae: an experimental study: III. Effects of temperature, vitamin requirements and inorganic nitrogen compounds on growth. Journal of Ecology 61:179 192 [doi:10.2307/2258926]. Moyle, J.B. 1956. Relationships between the chemistry of Minnesota surface water s and wildlife. Journal of Wildlife Management 20: 303320 [ doi:10.2307/3796967]. Murtaugh, P. 2007. Simplicity and complexity in ecological data analysis. Ecology 88: 56 62 [doi: 10.1890/00129658(2007)88[56:SACIED]2.0.CO;2].

PAGE 148

148 Myneni, R.B., C.J. Tucker, G. Asrar, and C.D. Keeling. 1998. Interannual variations in satellite sensed vegetation index data from 19811991. Journal of Geophysical Research 103:61456160 [doi:10.1029/97JD03603]. National Oceanic and Atmospheric Administration (NOAA). 1996. National Climatic Data C enter. Asheville, North Carolina Nickerson, D., and B. Madsen. 2005. Nonlinear regression and ARMA models for precipitation chemistry in East Central Florida from 1978 to 1997. Environmental Pollution 135: 371 379 [doi: 10.1016/j.envpol.2004.11.010|10.1016/j.envpol.2004.11.010]. Naumann E. 1919. A few comments on limnoplankton ecology with special reference to phytoplankton. Scen Bot Tidskr 13: 129163. on rich anoxic hypolimnia. Arch Hydrobiologia 104:459476. Prairie, Y. 1996. Evaluating the predictive power of regression models. Canadian Journal of Fisheries and Aquatic Sciences 53: 490492 [doi: 10.1139/cjfas 53 3 490]. Pareto, V. 1897. Cours deconomie politique, V. 2, book 3. Lausanne. R Deve lopment Core Team. 2008 R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 390005107 0, URL http://www.R projec t.org Rigler, F.H. and R.H. Peters 199 5. Science a nd Limnolgoy. Ecology Institute, Oldendoff/Luhe, Germany. Ruiz Bernard, I. 2012. Effects of a catastrophic forest fire on trophic parameters of an outstanding Florida water. Thesis presented to the University of Florida, Gainesville, Florida. Sakamoto, M. 1966. Primary production by phytoplankton community in some Japanese lakes and its dependence on lake depth. Archiv fur Hydrobiologie 62:1 28. Schafer, M.D. R.E. Dickinson, J.P Heaney, and W.C. H uber 1986. Gazetteer of Florida l akes. Pub 96. Florida Water Resource Research Center, Gainesville, Florida. Shannon, E.E. and P.L. Brezonik. 1972. Limnologic characteristics of north and central Florida lakes. Limnology and Oceanography 17:97110 [ doi:10.4319/lo.1972.17.1.0097]. SAS Institute Inc. 2007. JMP Users Guide. SAS Institute Inc., Cary, North Carolina. Scheffer, M. 1998. Ecology of shallow lakes. (Ed. Usher, D.L.) Chapman and Hall, London, United Kingdom

PAGE 149

149 Smeltzer, E., W.W. Walker Jr., and V. Garrison. 1989. Eleven years of lake eutrophication monitoring in Vermont: a critical evaluation. Enhancing States Lake Management Programs:5362. Snedecor, G.W. and W.C. Cochran. 1980. Statistical Methods, 7th ed. The Iowa State University Press, A mes, Iowa. Soranno, P.A., K.E. Webster, J.L. Riera, T.K. Kratz, J.S. Baron, P.A. Bukaveckas, G.W. Kling, D.S. White, N. Caine, R.C. Lantrop, and P.R. Leavitt. 1999. Spatial variation among lakes within landscapes: ecological organization along lake chains. Ecosystems 2:395410 [ doi:10.1007/s100219900089]. Stow, C., S. Carpenter, K. Webster, and T. Frost. 1998. Long term environmental monitoring: Some perspectives from lakes. Ecological Applications 8 : 269 276 [doi: 10.2307/2641066]. Terrell, J. B., D. L. Watson, M. V. Hoyer, M. S. Allen, and D. E. Canfield, Jr. 2000. Temporal water chemistry trends (1967 1997) for a sample (127) of Florida waterbodies. Lake and Reservoir Management 16: 177 194 [ doi:10.1080/07438140009353962 ] Thienemann, A. 1925. Die Binnengewsser Mitteleuropas (The inland waters of Central Europe). E. Schweizer bartsche Verlagsbuchhandlung. Vidondo, B., Y.T. Prairie, J.M. Blanco, and C.M. Duarte. 1997. Some aspects of analysis of size spectra in aquatic ecol ogy. Limnology and Oce anography 42:184192 [ doi:10.4319/lo.1997.42.1.0184]. Vitousek, P.M., H.A. Mooney, J. Lubchenco, and J.M. Melillo. 1997. Human domination of Earth s ecosystems. Science 277: 494499 [ doi:10.1126/science.277.5325.494]. Vollenweider, R.A. 1968. Scientific fundamentals of eutrophication of lakes and flowing waters with particular reference to nitrogen and phosphorus as factors of eutrophication. Paris(France): Organization for Economics CoOperation and Development: Tech Report No. DAS/CSI/68.27. Wetzel, R. G. 2001. Limnology, 3rd edition. Academic Press San Diego, California. Williamson, C.E., J.E. Saros, W.F. Vincent, and J.P. Smol. 2009. Lakes and reservoirs as sentinels, integrators, and regulators of climate change. Limnology and Oceanog raphy 54(6, p art 2):22732282 [ doi:10.4319/lo.2009.54.6_part_2.2273]. Worrall, F., and T. Burt. 2004. Time series analysis of long term river dissolved organic carbon records. Hydrologic Processes 18: 893911 [doi:10.1002/hyp.1321].

PAGE 150

150 (USEPA) United States Environmental Protection Agency. 2000. Nutrient criteria technical guidance manual: lakes and reservoirs. Office of Water EPA 822B00 001. (USEPA) United States Environmental Protection Agency. 2010. National lakes assessment: [cited 7 J uly 2012]. Available from http://www.epa.gov/lakesurvey/ (USDA) United State Department of Agriculture. 201 2. http://planthardiness.ars.usda.gov/PHZMWeb/ Urquhart, S.N., S.G. Paulsen, and D.P. Larsen. 1998. Monitoring for policy relevant regional trends over time. Ecological Applications 8:246 257 [doi:10.2307/2641064].

PAGE 151

151 BIOGRAPHICAL SKETCH Dana Bigham received her Bachelor of Science degree in zoology and biological aspects of conservation in 2001 from the University of Wisconsin Madison. As an undergraduate, she worked at the U niversity of Wisconsin Ophthalmology Department laboratory assisting in eye melanoma research. Dana also had the opportunity to work as a field research assistant examining two evolutionary distin ct stickleback (Gasterosteidae sp. ) in British Columbia lakes Before returning to graduate school, Dana worked for the Department of Wisconsin Natural Resources completing macrophyte surveys of Wisconsin lakes. In 2008, Dana completed a Master of Science degree at the University of Florida in Fisheries and Aquatic Sciences Her thesis examined concentrations of the cyanobacterial toxin, microcystin, ac ross Florida lakes and more specifically in the Harris Chain of Lakes, located in Lake County, Florida. Th ereafter, Dana began her dissertation research examining temporal changes in trophic state variables for a large population of Florida lakes. As a graduate student, Dana was an active member in various societies serving on the board of directors for the N orth American Lake Management Society and the Florida Lake Management Society. Dana received her Doctor of Philosophy from the University of Florida in December 2012