<%BANNER%>

Evaluating Temporal and Spatial Land Use Influences Affecting Nutrient Water Quality in the Biscayne Bay Watershed, Florida

Permanent Link: http://ufdc.ufl.edu/UFE0025165/00001

Material Information

Title: Evaluating Temporal and Spatial Land Use Influences Affecting Nutrient Water Quality in the Biscayne Bay Watershed, Florida
Physical Description: 1 online resource (179 p.)
Language: english
Creator: Carey, Richard
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: canals, emergy, florida, gis, landscape, nutrients, urbanization, water
Interdisciplinary Ecology -- Dissertations, Academic -- UF
Genre: Interdisciplinary Ecology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Biscayne Bay, a tropical estuary along the southeastern Florida coastline, drains both the Miami metropolitan complex and South Dade Agricultural Area. Water management systems protect the watershed from seasonal floods but have also disrupted historical freshwater flows to the bay, which requires minimal phosphorus and nitrogen inputs. Watershed discharges therefore have a controlling influence on bay water quality and can degrade sensitive estuarine habitats. To explore watershed land use and water quality variability, this study evaluated temporal and spatial land use influences on nutrient concentrations measured in canals discharging to Biscayne Bay. Disturbance indicators for 1995, 1999, and 2004 (landscape metrics, Landscape Development Intensity index LDI, and imperviousness) suggested urban sub-basins were stable and characterized by complex residential areas, corresponding to greater anthropogenic intensity compared to agricultural and mixed land use sub-basins. Historical nutrient data (1992 to 2006), analyzed using multiple methods (trend analysis, load estimation, and a new water quality index), revealed water quality has generally improved. This improvement was likely a response to implementation of agricultural and urban best management practices as well as repair of leaky wastewater systems. The Pollutant Empower Density (PED) index assesses proportional impacts from point source discharges and two discharge locations (MW04 and LR06) had the greatest potential to degrade Biscayne Bay water quality. Land use and water quality relationships were evaluated at multiple spatial extents (sub-basins, canal buffers, and site buffers) and regressions suggested nitrate/nitrite-nitrogen loads were most related to land use variables at the sub-basin level. Development patterns in a smaller zone (1000 m canal buffer) were important factors for total phosphorus loads, reflecting watershed nutrient transport processes. Rapid urbanization is ongoing in south Florida and both the intensity and spatial distribution of land uses affect nutrient discharges that could alter Biscayne Bay. Disturbance indicators can link land use to water quality parameters for improved watershed management, such as increasing treatment efficiency of established pollutant-control strategies (e.g., detention and retention systems) and guiding zoning regulations. Combined assessment of multiple indicators also provides a more holistic interpretation of water quality, which is necessary for optimizing resources to preserve water quality.
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 Richard Carey.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Migliaccio, Kati W.
Local: Co-adviser: Kiker, Gregory.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0025165:00001

Permanent Link: http://ufdc.ufl.edu/UFE0025165/00001

Material Information

Title: Evaluating Temporal and Spatial Land Use Influences Affecting Nutrient Water Quality in the Biscayne Bay Watershed, Florida
Physical Description: 1 online resource (179 p.)
Language: english
Creator: Carey, Richard
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: canals, emergy, florida, gis, landscape, nutrients, urbanization, water
Interdisciplinary Ecology -- Dissertations, Academic -- UF
Genre: Interdisciplinary Ecology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Biscayne Bay, a tropical estuary along the southeastern Florida coastline, drains both the Miami metropolitan complex and South Dade Agricultural Area. Water management systems protect the watershed from seasonal floods but have also disrupted historical freshwater flows to the bay, which requires minimal phosphorus and nitrogen inputs. Watershed discharges therefore have a controlling influence on bay water quality and can degrade sensitive estuarine habitats. To explore watershed land use and water quality variability, this study evaluated temporal and spatial land use influences on nutrient concentrations measured in canals discharging to Biscayne Bay. Disturbance indicators for 1995, 1999, and 2004 (landscape metrics, Landscape Development Intensity index LDI, and imperviousness) suggested urban sub-basins were stable and characterized by complex residential areas, corresponding to greater anthropogenic intensity compared to agricultural and mixed land use sub-basins. Historical nutrient data (1992 to 2006), analyzed using multiple methods (trend analysis, load estimation, and a new water quality index), revealed water quality has generally improved. This improvement was likely a response to implementation of agricultural and urban best management practices as well as repair of leaky wastewater systems. The Pollutant Empower Density (PED) index assesses proportional impacts from point source discharges and two discharge locations (MW04 and LR06) had the greatest potential to degrade Biscayne Bay water quality. Land use and water quality relationships were evaluated at multiple spatial extents (sub-basins, canal buffers, and site buffers) and regressions suggested nitrate/nitrite-nitrogen loads were most related to land use variables at the sub-basin level. Development patterns in a smaller zone (1000 m canal buffer) were important factors for total phosphorus loads, reflecting watershed nutrient transport processes. Rapid urbanization is ongoing in south Florida and both the intensity and spatial distribution of land uses affect nutrient discharges that could alter Biscayne Bay. Disturbance indicators can link land use to water quality parameters for improved watershed management, such as increasing treatment efficiency of established pollutant-control strategies (e.g., detention and retention systems) and guiding zoning regulations. Combined assessment of multiple indicators also provides a more holistic interpretation of water quality, which is necessary for optimizing resources to preserve water quality.
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 Richard Carey.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Migliaccio, Kati W.
Local: Co-adviser: Kiker, Gregory.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0025165:00001


This item has the following downloads:


Full Text

PAGE 1

1 EVALUATING TEMPORAL AND SPATIAL LAND USE INFLUENCES AFFECTING NUTRIENT WATER QUALITY IN THE BISCAYNE BAY WATERSHED, FLORIDA By RICHARD O. CAREY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009

PAGE 2

2 2009 Richard O. Carey

PAGE 3

3 To Florence and Judene

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank the chai r of my supervisory committee, Kati W. Migliaccio, for her helpful suggestions and invaluab le insight throughout my doctora l program. I would also like to thank committee members Gregory A. Kiker, Ma rk T. Brown, Yuncong Li, and Bruce Schaffer for allocating time to my proj ect and providing recommendations on various aspects of my dissertation research. The School of Natural Resources and Envi ronment, Agricultural and Biological Engineering department, Institute of Food and Ag ricultural Sciences, and the Tropical Research and Education Center were important resour ces during my program. In addition, financial support from the Agricultural and Biological Engin eering department facilitated my research. Overall, my project required data from numerous sources and I would like to thank the South Florida Water Management District, Miami-Dade Department of Environmental Resources Management, Florida Geographic Data Library, and the Florida Department of Environmental Protection for allowi ng me to access available data. My family also contributed to the successful completion of my program and I am grateful for their strength, support, and sacrifice.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4TABLE OF CONTENTS ............................................................................................................. ....5LIST OF TABLES ...........................................................................................................................8LIST OF FIGURES .......................................................................................................................10LIST OF ABBREVIATIONS ........................................................................................................ 12ABSTRACT ...................................................................................................................... .............131 INTRODUCTION .................................................................................................................. 15Background .................................................................................................................... .........15Land Use Analysis ..................................................................................................................17Configuration and Composition ...................................................................................... 17Emergy Index: Landscape Development Intensity .......................................................... 20Imperviousness ................................................................................................................ 22Water Quality Analysis ........................................................................................................ ...23Nutrient Enrichment ........................................................................................................23Trends ........................................................................................................................ ......24Loads ......................................................................................................................... ......25Emergy Index: Pollutant Empower Density .................................................................... 27Regression Analysis ........................................................................................................... .....28Background .................................................................................................................... ..28Variable Selection ...........................................................................................................29Model Validation and Assessment .................................................................................. 31Statement of Problem .......................................................................................................... ...32Objectives .................................................................................................................... ...........33Significance of Study ..............................................................................................................33Scope of Study ........................................................................................................................352 EVALUATING LAND USE CHANGE (1995 TO 2004) USING MULTIPLE DISTUR BANCE INDICATORS IN THE BISCAYNE BAY WATERSHED, FLORIDA ....................................................................................................................... ........44Introduction .................................................................................................................. ...........44Methods ..................................................................................................................................47Study Area .......................................................................................................................47Land Use Data .................................................................................................................48Landscape Metrics ........................................................................................................... 49LDI Index ........................................................................................................................51Imperviousness ................................................................................................................ 51

PAGE 6

6 Results .....................................................................................................................................52Landscape Metrics ........................................................................................................... 52LDI Index and Land Use Percentages .............................................................................53Imperviousness ................................................................................................................ 54Discussion .................................................................................................................... ...........54Landscape Metrics ........................................................................................................... 54LDI Index and Imperviousness ....................................................................................... 57Management Implications ............................................................................................... 59Conclusion .................................................................................................................... ..........623 NUTRIENT DISCHARGES TO BISCAY NE BAY, FLORI DA (1992 TO 2006): WATER QUALITY TRENDS, LOADS, AND A POLLUTANT INDEX ........................... 76Introduction .................................................................................................................. ...........76Methods ..................................................................................................................................79Study Area .......................................................................................................................79Water Quality Data ..........................................................................................................80Trend Analysis .................................................................................................................81Nutrient Loads .................................................................................................................82Pollutant Index .................................................................................................................83Results .....................................................................................................................................86Trend Analysis .................................................................................................................86Nutrient Loads .................................................................................................................86Pollutant Index .................................................................................................................87Discussion .................................................................................................................... ...........88Trends ........................................................................................................................ ......88Loads ......................................................................................................................... ......90Pollutant Index .................................................................................................................92Management Implications ............................................................................................... 94Conclusion .................................................................................................................... ..........954 LAND USE INFLUENCES (1995 TO 2004) AFFECTING NUTRIE NT WATER QUALITY IN THE BISCAYNE BAY WATERSHED, FLORIDA ................................... 108Introduction .................................................................................................................. .........108Methods ................................................................................................................................111Study Area .....................................................................................................................111Land Use Data ...............................................................................................................112Landscape Metrics ......................................................................................................... 113Landscape Development Intensity Index ...................................................................... 114Imperviousness .............................................................................................................. 115Water Quality Data ........................................................................................................115Nutrient Loads ...............................................................................................................116Stepwise Regressions .................................................................................................... 117Results ...................................................................................................................................119Landscape Metrics ......................................................................................................... 119LDI and Imperviousness ................................................................................................120

PAGE 7

7 Loads and Stepwise Regressions ...................................................................................121Discussion .................................................................................................................... .........122Land Use Variables and Nutrient Loads ....................................................................... 122Uncertainty Analysis ..................................................................................................... 125Management Implications ............................................................................................. 127Conclusion .................................................................................................................... ........1295 SUMMARY ....................................................................................................................... ...159Objective 1 ............................................................................................................................159Objective 2 ............................................................................................................................160Objective 3 ............................................................................................................................161Research Synthesis ...............................................................................................................161LIST OF REFERENCES .............................................................................................................165BIOGRAPHICAL SKETCH .......................................................................................................179

PAGE 8

8 LIST OF TABLES Table page 1-1 Land use/land cover data for the Bi scayne Bay watershed (1995, 1999, and 2004). ........ 372-1 Description of land use/land cover classes. .......................................................................642-2 Description of landscape metrics. ...................................................................................... 652-3 Land use/land cover coefficients for La ndscape Development Intensity (LDI) index and percent imperviousness. .............................................................................................. 662-4 Land use/land cover data for five study sub-basins in the Biscayne Bay watershed. ........673-1 Energy sources and conversion factors used to calculate background productivity (emergy signature) for Biscayne Bay. ............................................................................... 973-2 Trend analysis results for nutrient concentr ations at six water quality monitoring sites in the Biscayne Bay watershed. .........................................................................................983-3 Average Nash-Sutcliffe Efficiency (NSE) coefficients for six water quality monitoring sites in the Biscayne Bay watershed after comparing LOADEST simulated loads to measured loads (1992 to 2006). ........................................................... 993-4 Summary statistics for nutrient concentra tions (1992 to 2006) and flow data at six water quality monitoring sites in the Biscayne Bay watershed. ...................................... 1004-1 Land use/land cover data (500 m canal buffer). ..............................................................1304-2 Land use/land cover data (1000 m canal buffer). ............................................................1334-3 Land use/land cover data (1500 m canal buffer). ............................................................1364-4 Land use/land cover data (500 m water quality monitoring site buffer). .........................1394-5 Land use/land cover data (1000 m wate r quality monitori ng site buffer). ....................... 1424-6 Land use/land cover data (1500 m wate r quality monitori ng site buffer). ....................... 1454-7 Summary statistics for nutrient concentra tions (1992 to 2006) and flow data at eight water quality monitoring sites in the Biscayne Bay watershed. ...................................... 1484-8 Class-level factors and metrics describi ng spatial variability in the Biscayne Bay watershed. .................................................................................................................... ....1494-9 Summary statistics for Landscape Development Intensity (LDI) index values and Directly Connected Impervious Area (DCIA) percentages (1995 to 2004) considering multiple spatial extents in the Biscayne Bay watershed. ............................. 150

PAGE 9

9 4-10 Average Nash-Sutcliffe Efficiency (N SE) coeffi cients for eight water quality monitoring sites (1992 to 2006) in the Biscayne Bay watershed after comparing LOADEST simulated loads to measured loads. .............................................................. 1514-11 Median annual loads (1992 to 2006) at eight water quality monitoring sites in the Biscayne Bay watershed. .................................................................................................1524-12 Validation results for NOX-N stepwise regression mode ls (forward and backward selection and forward direction only) using three quantitative statistics. ........................ 1534-13 Validation results for TP stepwise re gression models (forward and backward selection and forward direction only) using three quantitative statistics. ........................ 154

PAGE 10

10 LIST OF FIGURES Figure page 1-1 Biscayne Bay watershed in southeastern Florida ...............................................................381-2 Examples of water management systems in south Florida ................................................ 391-3 Canals and drainage sub-basins in the Biscayne Bay watershed ....................................... 401-4 Watershed land use/land cover map (1995) ....................................................................... 411-5 Watershed land use/land cover map (1999) ....................................................................... 421-6 Watershed land use/land cover map (2004) ....................................................................... 432-1 Five study sub-basins in the Biscayne Bay watershed....................................................... 702-2 Landscape-level factors describing spatial va riability in the Biscayne Bay watershed ..... 712-3 Selected class-level metrics for row crops in the Biscayne Bay watershed. ..................... 732-4 Selected class-level metrics for medium density single family residential (MSR) land use class in the Bis cayne Bay watershed ........................................................................... 742-5 Landscape Development Intensity (LDI) index values and Directly Connected Impervious Area (DCIA) percentages for five study sub-basins in the Biscayne Bay watershed.. ................................................................................................................... ......753-1 Six water quality monitoring site s in the Biscayne Bay watershed. ................................ 1013-2 Annual discharge (1992 to 2006) from si x water quality monitoring sites in the Biscayne Bay watershed. .................................................................................................1023-3 Nash-Sutcliffe Efficiency (NSE) coeffici ents for selected water quality monitoring sites in the Biscayne Bay watershed after comparing LOADEST simulated loads to measured loads (1992 to 2006) ........................................................................................ 1033-4 Estimated annual nutrient loads (1992 to 2006) at six water quality monitoring sites in the Biscayne Bay watershed. .......................................................................................1043-5 Pollutant Empower Density (PED) index va lues (1992 to 2006) at six water quality monitoring sites in the Bi scayne Bay watershed. ............................................................ 1063-6 Flow chart illustrating the combined use of trend analysis, load estimation, and the Pollutant Empower Density index to evaluate water quality variability in watersheds. 1074-1 Buffers (500, 1000, and 1500 m) for five can als in the Biscayne Bay watershed. .......... 155

PAGE 11

11 4-2 Buffers (500, 1000, and 1500 m) for eight water quality m onitoring sites in the Biscayne Bay watershed. .................................................................................................1564-3 Land use variables influencing NOX-N loads in the Biscayne Bay watershed. ............... 1574-4 Land use variables influencing TP lo ads in the Biscayne Bay watershed ....................... 1585-1 Flow chart illustrating an overall process that can be us ed to evaluate land use-water quality relationships in watersheds. .................................................................................164

PAGE 12

12 LIST OF ABBREVIATIONS AWMSI Area weighted mean shape index DCIA Directly connected impervious area HSR High density single family residential LDI Landscape Development Intensity index LIC Low intensity commercial LPI Largest patch index LSI Landscape shape index LSR Low density single family residential MDL Minimum detection limit MNN Mean nearest neighbor MPI Mean proximity index MPS Mean patch size MSR Medium density single family residential NSE Nash-Sutcliffe Efficiency index PBIAS Percentage bias PED Pollutant empower density index PSCoV Patch size coefficient of variation PSSD Patch size standard deviation RSR Ratio of the root mean square error to the standard deviation of measured data TIA Total impervious area UEV Unit emergy value

PAGE 13

13 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EVALUATING TEMPORAL AND SPATIAL LAND USE INFLUENCES AFFECTING NUTRIENT WATER QUALITY IN THE BISCAYNE BAY WATERSHED, FLORIDA By Richard O. Carey December 2009 Chair: Kati W. Migliaccio Cochair: Gregory A. Kiker Major: Interdisciplinary Ecology Biscayne Bay, a tropical estuary along the southeastern Florid a coastline, drains both the Miami metropolitan complex and South Dade Ag ricultural Area. Water management systems protect the watershed from seasonal floods but have also disrupted hi storical freshwater flows to the bay, which requires minimal phosphorus and nitr ogen inputs. Watershed discharges therefore have a controlling influence on ba y water quality and can degrade se nsitive estuarine habitats. To explore watershed land use and water qualit y variability, this study evaluated temporal and spatial land use influences on nutrient concen trations measured in canals discharging to Biscayne Bay. Disturbance i ndicators for 1995, 1999, and 2004 (landscape metrics, Landscape Development Intensity index [LDI], and impervious ness) suggested urban s ub-basins were stable and characterized by complex resi dential areas, corresponding to greater anthropogenic intensity compared to agricultural and mixed land use subbasins. Historical nutrient data (1992 to 2006), analyzed using multiple methods (trend analys is, load estimation, and a new water quality index), revealed water quality has generally improved. This improvement was likely a response to implementation of agricultural and urban best management practices as well as repair of leaky wastewater systems. The Polluta nt Empower Density (PED) inde x assesses proportional impacts

PAGE 14

14 from point source discharges and two discharg e locations (MW04 and LR 06) had the greatest potential to degrade Biscayne Bay water quality. Land use and wate r quality relationships were evaluated at multiple spatial extents (sub-basins, canal buffers, and site buffers) and regressions suggested nitrate/nitritenitrogen loads were most related to land use variables at the sub-basin level. Development patterns in a smaller zone (1000 m canal buffer) were important factors for total phosphorus loads, reflecting wate rshed nutrient transport processes. Rapid urbanization is ongoing in south Fl orida and both the in tensity and spatial distribution of land uses affect nutrient discharges that could alter Biscayne Bay. Disturbance indicators can link land use to water quality parameters for improved watershed management, such as increasing treatment efficiency of establ ished pollutant-control st rategies (e.g., detention and retention systems) and guiding zoning regu lations. Combined assessment of multiple indicators also provides a more holistic interpretation of water quality, which is necessary for optimizing resources to preserve water quality.

PAGE 15

15 CHAPTER 1 INTRODUCTION Background The south Florida landscape has changed dram atically in recent decades. Originally dominated by wetlands, pine forests, estuarie s, bays, and hardwood hammocks, anthropogenic influences have transformed the region. Drainage projects, agriculture, and urban development have fragmented the landscape, threatened th e Florida Everglades, fostered exotic species invasions, and impaired coasta l resources (McPherson and Halle y 1996) such as the Biscayne Bay, a tropical barrier-island estuary located alon g the southeastern Florida coastline (Figure 11). Surrounded by urbanized Miami-Dade to the north, Homestead to the west, and the Florida Keys to the south, Biscayne Bay is ecologically and economi cally important to the Miami metropolitan area because its tropical reefs a nd mangroves support numerous species (manatees, dolphins, wading birds etc.) as we ll as fishing and recreational industries (BBPI 2001). The state designated Biscayne Bay as an Outstanding Fl orida Water in 1978 to prevent environmental degradation but extensive development in its 2,500 km2 watershed has altered the estuary. Problems include local fish extinctions, hypers alinity, algal blooms related to eutrophication, point source pollution, and seagrass deaths (Alleman et al. 1995). A historical perspective of na tural resource management in south Florida is required to understand water-related issues in the bay a nd elsewhere in the region. For example, the Everglades Reclamation program that began in 1906 sparked an intense real estate boom that laid the foundation for south Floridas future deve lopment. Over 5,000 years of accumulated peat deposits created a vast, 12,000 km2 Everglades, but in the 20th century, drainage projects reduced this area to 6,000 km2 (Gleason and Stone 1994). Local governments wanted to encourage settlement but could not afford to dr ain the land; instead, speculators bought millions

PAGE 16

16 of acres of cheap land from the state, drained it, and attracted settlers (Hanna and Hanna 1948). Between 1910 and 1930, the population in three sout h Florida counties (Miami-Dade, Broward and Palm Beach) increased 165% and accelerated again after the Great Depression (1930s) and World War II (1939-1945). However, the Florida r eal estate crash in 1926, two deadly hurricanes in 1926 and 1928, and record floods in 1947 and 1948 slowed regional growth and shifted the focus to flood control (Schultz 1991). Political pressure from concerned reside nts in the 1950s gave government agencies a mandate to prevent future flooding problems and to develop suitable drainage projects (Blake 1980). The new era of multipurpose water manageme nt thus attempted to control the hydrology of south Florida. Inland levees in Broward County, for example, helped to lower groundwater levels and improve development opportunities on fo rmer cypress and pine forests (Renken et al 2005). South Floridas population conseque ntly doubled during the 1950s (724,000 to 1.5 million). Another important factor contributing to population growth during this period was the expansion of the transportation system. Three major highways that were built between 1930 and 1950 U.S. Route 1, U.S. Route 27, and the Ta mpa-Miami (Tamiami) Trail supported the extensive road network esta blished in the 1950s and 1960s (Solecki and Walker 2001). From 1900 to the early 1950s, agricultural pr oduction was predominantly responsible for land use conversions in the region but this pattern was beginning to change. Urban land uses in Miami-Dade and Broward Counties were increasing rapidly as the demand to shift land to urban uses outweighed the demand to shift land to ag ricultural uses over a large area (Solecki and Walker 2001). Agricultural product ion was still vital to the south Florida economy, however, as demand for winter vegetables and fruits grew in Northern and Midwestern states (Winsberg 1991). In addition, the Cuban Revolution (1959) in creased tariffs on imported sugar and this

PAGE 17

17 stimulated intensive sugar cane production in so uth Florida (Salley 1986). As the coast became more urbanized (1950-1970), farmers sought new ag ricultural land south of Lake Okeechobee in the Everglades Agricultural Area (EAA). Two main factors have therefor e driven land use change in so uth Florida from the 1970s to the present: increased migrat ion and significant demand for Fl oridas agricultural produce (Walker et al. 1997). Although the overall population growth rate of south Florida has declined, absolute population growth in the region has c ontinued to increase rapidly during the last 30 years; south Floridas population increased from 2.5 million in 1970 to more than 5 million in 2000. Northeastern and Midwestern states supplied the majority of new residents but immigrants from Caribbean and Latin American countries have also been substantial (Schultz 1991). From 1973 to1986, 28.3% of all land conversions were from agriculture to urban use (Solecki and Walker 2001) and the EAA has subsequently expa nded deeper into the Ev erglades (Salley 1986). With intensive urbanization and agricultural opera tions, water consumption in south Florida has increased in recent decades and when coupled with hydrological modifications, this increased water demand has the potential to deteriorate water quality (McPherson and Halley 1996). Land Use Analysis Configuration and Composition Dow (2000) defined human-dominated landscapes (e.g., the Biscayne Bay watershed) as complex mosaics where heterogeneous human activities gradually transform biophysical characteristics. For example, numerous studies have linked land use with water quality (e.g., Osborne and Wiley 1988; Johnson et al. 1997; Harm an-Fetcho et al. 2005) as the proportion and spatial arrangement of land use and land cover (L ULC) within watersheds can have significant impacts (e.g., Hunsaker and Levine 1995; Roth et al. 1996; Johnson et al. 2001). The field of landscape ecology provides a conceptual fr amework to understand these anthropogenic

PAGE 18

18 influences because it is primarily concerned with land use patterns within defined areas, interactions among different lands cape elements, and the effect s of changes in the spatial heterogeneity complex over time (Risser et al. 1984; Haines-Young et al. 1994). German geographer Carl Troll originally developed the term landscap e ecology in the 1930s and the discipline grew in Europe as a corollary to land planning (Schreiber 1990); beginning in the 1980s, landscape ecology became prevalent in the North American literature (Turner 2005). A fundamental concept in la ndscape ecology is that patter ns influence processes and several studies have emphasized methods to qua ntify spatial heterogeneity (e.g., Forman and Godron 1986; ONeill et al. 1988; Turner and Gardner 1991). Metrics (var iables) have been developed for landscape composition (relative am ounts of different elements in the landscape) and configuration (arrangement of these elements) that aid an alysis and interpretation of landscape processes (Turner 1989; Li and Wu 2004). The most widely used software package to calculate landscape metrics is FRAGSTATS (McGarigal and Marks 1995); several different categories of metrics can be calculated as FRAGST ATS generates values that can be useful to understanding changes occurring in an area such as a watershed. Patch Analyst (Elkie et al. 1999) uses a modified form of FRAGSTATS and provides an integrated user interface that enables metrics to be calculated for land use layers at both landscape and class levels within a Geographic Information System (GIS) software package, ArcGIS (ESRI 2005). GIS software, which can integrate multiple historical datasets (e.g., land use and water quality data), can be used to study ecological relationshi ps within heterogeneous landscapes. Hundreds of metrics can be obtained from tem poral and spatial land us e analyses utilizing mathematical operations in GIS. For a large data set containing multiple variables, it is often easier to analyze this information if the number of variables is reduced to a smaller set of

PAGE 19

19 components that retains all the important information. Principal Component Analysis (PCA) is a data reduction technique that iden tifies linear combina tions of the original variables explaining all of the variance in a data set (Nichols 1977; Bengraine a nd Marhaba 2003). These linear combinations, or principal components, describe va riability in the dataset which are not directly measured. Factor analysis (FA) is another statis tical procedure used with multi-variable datasets to identify factors cont ributing to the overall variance (M cDonald 1985). The difference between PCA and FA is that while PCA attempts to simplify variable interpretation through data reduction, FA primarily focuses on identif ying significant, underlying factors. PCA and FA have been used toge ther in land use anal ysis to elucidate hi storical or ongoing processes at different spatial exte nts, such as an entire landscapes or individual land use classes. For example, Ritters et al (1995) calculated 55 landscap e metrics for 85 LULC maps representing different U.S. physi ographic regions to determine th e best combination of metrics that describe landscape pattern and structure. Landscape metr ics are often highly correlated because multiple variables can quantify the same phenomena but present the information in alternate forms (McGarigal and Marks 1995; Li and Wu 2004). Using correlation analysis, PCA, and FA, Ritters et al. (1995) re duced the initial set of 55 metric s to six factors, or composite variables, responsible for 87% of the variation in the dataset. Similarly, Cushman et al. (2008) analyzed LULC in eastern, central, and western U.S. regions and reduced both landscape (54 to 17) and class (49 to 24) level metrics into in dependent variables that described landscape configuration and structure. Other studies have calculated metrics and used PCA and FA to evaluate landscape structure at multiple scales (Griffith et al. 2000), to investigate development patterns in watersheds (Cifaldi et al. 2004; Kearns et al. 2005) and to assess the relationship between metrics and sediment contam ination levels (Paul et al. 2002).

PAGE 20

20 Emergy Index: Landscape Development Intensity Emergy, or embodied energy, is a concept de veloped by H.T. Odum (Odum 1971; Odum 1996) that derived from decades of analysis on th e differential ability of various forms of energy to do work (Brown and Ulgiati 2004). In the environmental context, available energy from sources such as the sun, wind, and rain are transformed within natural systems as it is used to maintain functional processes and/or to create new resources. However, energy is degraded during these transformations and the resultant fo rms of energy have unequal ability to do work (Odum 1996). To account for this variability, em ergy analysis expresses different forms of energy within a system in units of solar emergy (the available solar energy used during energy transformations; solar emjoules) to assess the total amount of energy required to achieve a particular output. Emergy, in essence, looks back upstream to record what energy went into the train of transformation processes (Odum 1996). Through emergy analysis, both natural and anthropogenic energy i nputs required to generate products and services can be evaluated. Brown and Viva s (2005) explored this benefit of emergy analysis by developing a Landscape Deve lopment Intensity (LDI) index that analyzed human disturbance gradients with in watersheds. The LDI index co mpares urban and agricultural areas based on energy signatures associated with land use classes that expanded on the earlier work of Brown (1980), and an evaluation of the relationship between development intensity and water quality in the St. Marks Watershed, Florid a (Brown et al. 1998). Brown and Vivas (2005) evaluated the nonrenewable areal empower intens ity (emergy per unit area per unit time) of various land uses using an area-weighted formula: LDITotal = % LUi LDIi (1-1) where LDITotal is the LDI ranking for a landscape, %LUi is the percent of the total area of influence in land use i and LDIi is the landscape development intensity coefficient for land use

PAGE 21

21 i Several other studies have uti lized this LDI index to quantif y disturbance gradients. Mack (2006) evaluated Ohio wetlands and found that th e LDI index compared favorably with other assessment tools. In Florida, the LDI index was strongly correlated to a wetland biological integrity index (Reiss 2006; Reiss and Brown 2007) Lane and Brown (2006) determined that the LDI index explained more of the variation in be nthic diatom species within Florida freshwater marshes compared to other landscape metrics (e .g., percent agriculture and percent urban). As the LDI concept was applied under di fferent conditions, an important emerging limitation of the index was that human disturba nce intensity was not related to background conditions in the landscape. Another limitation was that each land use had predetermined LDI coefficients. Reiss et al. (2009) developed a revised LDI me thod to address these limitations. In the revised method, LDI values star t at zero (i.e., nonrenewable em power intensity is equal to the renewable empower of the landscape unit), the overall impact of the nonrenewable empower intensity in a landscape is reduced as the background renewable empower intensity increases, and there is no maximum value. The following equation illustrates th e revised LDI method: LDI = 10 log10 (emPITotal /emPIRef) (1-2) where LDI [unit less] is the Landscape Devel opment Intensity index for a landscape, emPITotal [sej ha-1 yr-1] is the total empower intensity (sum of renewable background empower intensity and nonrenewable empower intensity of land uses), and emPIRef is the renewable empower intensity of the background e nvironment within a particular landscape. For example, the renewable empower intensity of Florida is 1.97 E15 sej ha-1 yr-1 (Reiss et al. 2009). The total empower intensity (emPITotal) was calculated as follows: emPITotal = emPIRef + (%LUi emPIi ) (1-3)

PAGE 22

22 where %LUi is the percent of the total area in LULC class i and emPIi [sej ha-1 yr-1] is the nonrenewable empower inte nsity for LULC class i Imperviousness Urban development creates impervious surfaces that can have multiple hydrological, physical, and ecological effects within a watershed. Arnold and Gibbons (1996) described four qualities of imperviousness: (1) impervious surfaces contribute to hydrologic changes that impair water quality; (2) imperviousness is a characte ristic of intensive land use activities; (3) impervious surfaces hinder pollutant processing by disrupting soil percolation; and (4) impervious surfaces efficiently transport polluta nts into receiving waters. Studies have used impervious surfaces such as buildings, parking lo ts, and roads as a measure of the extent of development within landscapes and have linke d these variables to overall water quality (e.g., McMahon and Cuffney 2000; Roy and Shuster 2009). Relatively low levels of watershed imperviousness can produce negative effects in aquatic systems; for example, Arnold and Gibbons (1996) characterized streams within wate rsheds containing <10% of impervious cover as protected, 10-30% as impacte d, and greater than 30% as degraded. Linking an imperviousness threshold to water quality can be challeng ing, however, because many studies do not differentiate between total and effective imperv ious cover within watersheds (Brabec et al. 2002). Total impervious areas (TIA) include surf aces that may drain to pervious ground while effective impervious areas only include impervi ous cover directly connected to waterways. Using TIA instead of directly connected impervious areas (DCIA) in water quality investigations may therefore obscure the influence of land use changes (Alley and Veenhuis 1983). Reducing DCIA is an important component of low im pact development (Wright and Heaney 2001) because it reflects the potentia l influence of human developm ent on adjacent systems.

PAGE 23

23 Land use analysis studies have estimated pe rcent imperviousness for different land use classes (e.g., Stankowski 1972; Griffin 1980; Alley and Veenhuis 1983). Miami-Dade Department of Environmental Resources Management (DERM) developed percent imperviousness values for land uses classes with in the Biscayne Bay watershed which can be used to calculate area-wei ghted imperviousness for LULC maps (DERM 2004). Quantifying development patterns by calcula ting disturbance indicators, such as watershed imperviousness, therefore has the potential to i ndicate the relative in fluence of land use classes on water quality. Water Quality Analysis Nutrient Enrichment Phosphorus and nitrogen are esse ntial nutrients for metabolic functions in living organisms and in aquatic environments, these nutrients s timulate overall productivi ty. However, excessive inputs of nitrogen and/or phosphorus can lead to over-enrichment, or eutrophication, of surface waters that produce problems such as algal bl ooms, decreased dissolved oxygen concentrations, and increased fish mortality (Carpenter et al 1998; Smith 1998). Eutrophication ranks as the leading pollutant problem affecting the ability of U.S. surface waters to meet designated uses such as recreation, fishing, and irrigation (Howarth et al. 2002) Nutrients originate from both point (e.g., wastewater treatment plants) and nonpoint (e.g., agricultural and urban runoff) sources and watershed management plans often include strategies to mitigate inputs from all significant contributing factors. Water quality monitoring programs provide an opportunity to evaluate temporal variability a nd spatial variability in nutrient inputs within watersheds that enable managers to implement plans targeted to specific concerns. For example, management plans often include measures to control the nu trient limiting algal and aquatic plant growth. Limitation develops because metabolic processes require optimal nutrient ratios and the nutrient that is most scarce will regulate syst em productivity (Hecky and Kilham 1988).

PAGE 24

24 In Biscayne Bay, the primary nutrient limiti ng autotrophic growth is phosphorus (Brand 1988; Kleppel 1996). Average phosphorus concentr ations are low but variable throughout Biscayne Bay with the northern section typically having greater concentrations than the central and southern sections (Brand 1988; Alleman et al 1995; Brand et al. 2002). Biscayne Bay is a natural oligotrophic estuary requ iring minimal inputs of phosphorus and nitrogen to function; nutrient inputs from the watershed therefore have a controlling influence on water quality in the bay (Browder et al. 2005). Due to the difference in phosphorus concentrations in sections of the watershed, Brand (1988) found phytopla nkton levels were five times greater in the north than in the south. Nitrogen may not be the limiting nutri ent in the bay but elev ated nitrate/nitrite concentrations (greater than 4 mg L-1; Cheesman 1989) derived from agricultural and some urbanized areas in the watershed are a concern because artificially high nitrogen concentrations may have subtle ecological effects (Alleman et al 1995), such as the bay being more susceptible to algal blooms. Trends Changes in land use, management practices, and environmental conditions may all lead to detectable differences in constituent concentra tions over time at water quality monitoring sites. Both parametric and nonparametric techniqu es (Hirsch et al. 1982; Hirsch et al. 1991; Letttenmaier et al. 1991) are availa ble to determine temporal changes in specific parameters of interest such as nutrients. Howeve r, long-term water quality datase ts have several characteristics that can complicate trend analysis; there are freque ntly large gaps in the dataset, data are often skewed, censored data are prevalent (values le ss than the minimum detection limit; MDL), and chemical analytical techniques can improve over time producing multiple MDLs for the same parameter. Seasonality (Champley and Doledec 1997; Mcartney et al. 20 03; Qian et al. 2007)

PAGE 25

25 and discharge fluctuations (Alley 1988; Hirsch et al. 1991) are other factor s that can significantly influence trend analysis. USGS Estimate Trend (ESTREND; Schertz et al. 1991) includes both parametric (Tobit regressions) and non-parametric (uncensored/censored seasonal Kendall) methods to determine trends in constituent water quali ty data. Tobit regression uses a maximum likelihood estimation method to determine trends for parameters that contain greater than 5% censored data with multiple MDLs. The seasonal Kendall methods are suitable for parameters with less than 5% censored data (uncensored seasonal Kendall) and greater than 5% censored data at a single detection limit (censored seasonal Kendall). As watershed management plans become more refined to address specific concerns, trend an alysis can provide important information on the relative success of such initiatives. Lietz (2000), for example, analyzed water quality data (1966 to 1994) at a site discharging to Biscayne Bay and found downward trends, indicative of improved water quality, for several parameters including total ammoni a nitrogen and total phosphorus. Loads Different methods are available to quantify nutrients entering a water body depending on the objectives of a particular investigation. Concentration measurements are convenient for comparing field data to water quality criteria, lo ads assess the mass of constituents transported over time and help to quantify the total amount delivered, and yields estimate the mass of constituents delivered per unit area per unit time, which can help to assess best management practices (Christensen 2001). The total amount of nutrients entering Biscayne Bay is of particular concern and previous studies have quantified incoming nutrient loads. Lietz (1999) estimated nutrient loads to Biscayne Bay by analyzing nutrient concentrati ons and freshwater discharges and Caccia and Boyer (2007) developed a nutrient loading budget to the bay by analyzing canals

PAGE 26

26 throughout the watershed. Both studies determined that nitrate/nitrit e-nitrogen loads were elevated in the southern agricultural draina ge areas while ammonia-nitrogen and total phosphorus loads were highest in the north ern and central urba n drainage areas. The U.S. Geological Survey (USGS) devel oped the Load Estimator (LOADEST) model that has been specifically designed to estimate loads in streams and rivers (Runkel et al. 2004). The model is a publicly available FORTRAN progr am that uses linear regressions to estimate daily, monthly, seasonal, or annual loads and users have the ability to customize the model to fit particular objectives. Donato and MacCoy (2005), for example, used the LOADEST model to develop regression equations estimating total phosphorus loads, orthophosphorus loads, and suspended sediment in the lower Boise River. The authors concluded that LOADEST was a useful tool with good spatial and temporal resolution relative to phosphorus and suspended sediment. LOADEST includes three methods to estimat e loads, Maximum Likelihood Estimation (MLE), Adjusted Maximum Likelihood Estimati on (AMLE), and Least Absolute Deviation (LAD). If the calibration dataset is uncensore d, MLE (Cohn et al. 1989) is used for load estimation in LOADEST as follows: L MLE = exp a0+ ajXj M j = 1 gm( m, s2, V ) (1-4) where L MLE is the load estimate using MLE, a0 and aj are maximum likelihood estimates, [ gm (m, s2,V) ] is the bias correction factor, which includes the number of degrees of freedom (m), the residual variance (s2), and a function rela ting to the explanator y variables (V).

PAGE 27

27 The Adjusted Maximum Likelihood Estimation (A MLE) method is used to estimate loads when the calibration dataset includes censored data. The AMLE method (Cohn 1988) is as follows: L MLE = exp a0+ ajXj M j = 1 ( a, b,s2 (1-5) where L AMLE is the load estimate using AMLE, a0 and aj are maximum likelihood estimates adjusted for bias, and H( a, b, s 2 is an approximate bias corre ction factor from Cohn et al. (1992). MLE and AMLE both assume that model resi duals are normally distributed. LOADEST includes the Least Absolute Deviation (LAD) method to estimate loads when the normality assumption is violated (Duan 1983; Powell 1984): L LAD = exp a0+ ajXj M j = 1 exp ( ek)n k= 1n (1-6) where L LAD is the load estimate using LAD, a0 and aj are LAD regression model coefficients, e represents the residual error, and n is the numbe r of uncensored calibration values. The flexibility of LOADEST and its ability to estimate loads when assumptions are violated enhances the utility of this program for water quality analyses. Emergy Index: Pollutant Empower Density The Pollutant Empower Density (PED) index is based on the concept that the effect of chemical stressors (metals, nutrients, etc.) in aquatic environments may be explained by their respective empower densities rela tive to background conditions. Each element has a unit emergy value (UEV) and this value increases as substan ces become more concentrated. Furthermore, for elements and compounds that are rare in nature, more energy is required to concentrate these

PAGE 28

28 materials and this results in higher emergy/ mass ratios (Brown and Ulgiati 2004). Metals, nutrients, and toxins generally have high UEVs and excess con centrations can alter critical ecosystem processes, which can lead to reduced ecosystem function. The LDI index is concerned with empower intensity (emergy per unit area per unit time) but the PED index focuses on empower density (emergy per unit volume per un it time). The PED index is calculated using the flux of pollutants and the background produc tivity of the reference environment: PED = 10 log10 (emPDTotal/emPDRef) (1-7) where PED [unit less] is the Pollutant Empower Density index for an aquatic system, emPDTotal is the total empower density [sej m-3 yr-1] and emPDRef is the background empower density [sej m-3 yr-1]. The total empower density (emPDTotal) is calculated as follows: emPDTotal = emPDRef + emPDi (1-8) where emPDi = empower density of pollutant i [sej m-3 yr-1]. The annual empower density of pollutant i is calculated using the specific emergy of the appropriate nutrient and its annual flow weighted concentration. Regression Analysis Background Regression analysis uses quantitative indepe ndent variables to e xplain variation in quantitative dependent variables. Linear regression models ca n be separated into two broad categories: simple regression models and multiple regression models. In simple linear regression models, the dependent variable (Y) is a direct function of an independent (or explanatory) variable (X). The following is the general equation for simple linear regressions: Y = 0 + 1X (1-9) where 0 is the intercept and 1 is the slope. For multiple linear regression models, multiple independent variables are used to predict the response of the dependent variable. Multiple

PAGE 29

29 regression models enable researchers to explor e phenomena that can be influenced by multiple factors, which offers an advantage over simple regression models. However, in regression analysis, the typical goal is to build a model us ing the fewest variables to explain the greatest variability in the response, and to accurately parameterize regression coefficients for those variables (Graham 2003). The general form of mu ltiple linear regression models is as follows: Y = 0+ 1 x1+ 2 x2...+ i xi+ (1-10) where Y is the dependent variable, s are regression coefficients, Xs are explanatory or independent variables and is the random error term. Several issues exist with the use of multiple regression models, such as ensuring that explanatory variables are independent of each other (Johnston 1972; Mason 1975; Graham 200 3). Variable sele ction (Hocking 1976; Thompson 1978) is another key component of m odel development as researchers often have concerns about the correct variables to use in various models. Finally, validation procedures evaluate the suitability and accuracy of regre ssion models (Snee 1977). Through multiple regression analyses and the in creased availability of GIS software, many studies have correlated landscape characteristics to water quality parameters. For example, Hunsaker and Levine (1995) used regression mode ls to link spatial and terrestrial processes to water quality in Illinois and Texas watersheds. St udies have also explor ed possible explanatory variables relating to land use and water quality at multiple spatial extents, including sub-basins (Mehaffey et al. 2005; Migliaccio et al. 2007), riparian zones (Silva a nd Williams 2001; Schiff and Benoit 2007), and monitoring site proximitie s (Bolstad and Swank 1997; King et al. 2005). Variable Selection Hocking (1976) reviewed numerous proposed variable selection methods for regression models. Predictor variables are often difficult to identify because resear chers require data on a large set of potential variables that satisfy assumptions such as homoscedasticity. Collinear

PAGE 30

30 variables (Johnston 1972; Mason 1975; Graham 2003) can result from incomplete sampling designs and can lead to problems with regressi on coefficients such as high variance (Hocking 1976). Thompson (1978) highlighted some sequential pr ocedures that have been used to select independent variables, includi ng forward ranking, backward ranking, forward selection, and backward elimination. Forward ranking involves assigning the highest rank to the independent variable with the smallest calculated variance ratio (Fs) and ranking additional variables based on comparisons of Fs values at each stage and critical F-va lues at a designated significance level. Independent variables with Fs values that are not significan t are deleted from the regression equation; the forward ranking procedure theref ore ranks independent variables from most important to least important. Conversely, the b ackward ranking procedure ranks independent variables in ascending order of importance. Th e forward and backward procedures may be expected to produce similar ranki ngs but Abt (1967) cautioned that a set of independent variables may form an association or com pound that affects the variance of the dependent variable when one of the independent variables in the group is removed from the model. To avoid the possible effects of these compounds, Thompson (1978) recommended the combined use of both procedures (forward and backward) to select independent variables. The forward selection procedure is similar to the forw ard ranking method: the independent variable that is most highly correlated with the dependent vari able is chosen first and subse quent variables are chosen based on correlation to the dependent variable, given previously selected variables. The backward elimination and backward ranking procedures pr oduce identical rankings a lthough calculated test statistics after the first step are not equal. Si milar to the comparisons between the forward and backward ranking methods, backward elimina tion is favored over fo rward selection if correlations or possible co mpounds among independent vari ables exist (Mantel 1970).

PAGE 31

31 Efroymson (1960) proposed an adaptation of th e forward selection pr ocedure that allows both the inclusion and removal of independent va riables. This stepwise regression procedure enables researchers to identify variables that may no longer be necessary in the model and may eliminate the problem of compounds in the fo rward selection method (Thomas 1978). Studies have used stepwise regressions to select a su b-set of landscape attribut es that had the most influence on historical land use patterns in Illin ois (Iverson 1988), spatial land use variables that explained aquatic organism diversity in stream s (Harding et al. 1998), an d urbanization variables affecting stream ecological functi on (Chadwick et al. 2006). Stepwise regressions ha ve also been used to develop multivariate regression models to predict constituent loads of targeted water quality parameters in response to changes in la nd use variables (e.g., John son et al. 2001; Jones et al. 2001; Paul et al. 2002). In addition, Mehaffey et al. (2005) used pairwise correlations and stepwise regressions to identif y independent variables that we re highly correlated to total nitrogen, total phosphorus, and fecal coliform bact eria in New York watersheds. Two landscape characteristics, percent agriculture development and percent urban development, explained 25 to 75% of the variation in the regression models (M ehaffey et al. 2005). Results such as these have clear management implications and can lead to improved strategies to mitigate threats to vulnerable water resources. Model Validation and Assessment After selecting independent variables, regressi on models need to be validated before being used because the goal of model development is to identify the best possible set of variables for a particular system. However, using the same se t of data for model se lection and inference, without model validation, can lead to unreliable models. Snee (1977) listed the following options to validate regression models: (1) compare the dependent variable and regression coefficients with physical theory; (2) collect new data to check predictio ns; (3) compare model results with

PAGE 32

32 theoretical models and simulated data; (4) reserv e a section of the available data to estimate model prediction accuracy. Collecting new data is the preferred method for validating regression models but this may not be a viable option in certain scenarios. Snee (197 7) notes that splitting the data into an estimation dataset (for model coefficients) and a prediction dataset (to test model accuracy) can simulate the process of collecting new data. Statement of Problem Biscayne Bay requires substantial freshwater inputs to maintain its natural ecosystem processes but water management operations (canals, levees, pump sites, etc.) in south Florida have disrupted historical fr eshwater flows to the bay (F igure 1-2). In addition, urban development and agricultural development in th e watershed have thrived on former wetlands as canals have lowered water tables (Parker et al. 1955), reducing waters hed water storage and creating polluted discharges that degrade sensi tive estuarine habitats (Browder et al. 2005). Numerous studies have analyzed the impact of agriculture on wa ter resources in the Biscayne Bay watershed (e.g., Wang et al. 2003; Zhou et al 2003) but over the last three decades, the predominant form of land use change has been th e conversion of agricultura l and natural areas to residential or urban complexes (Solecki and Walker 2001). Although researchers have addressed increasing population densities (e.g. Finkl a nd Charlier 2003; Renken et al. 2005) and the relative influence of land use and water ma nagement (e.g. Caccia and Boyer 2005) in south Florida, the spatial characteristics of land use change have not been quantified and linked to specific pollutants in the Biscayne Bay watershed. According to Lausch and Thulke (2001), e nvironmental protection requires a landscapebased approach that includes qua ntitative calculations of lands cape structures, functions, and interactions:

PAGE 33

33 Studying the landscape, its current state (str ucture) and its future changes (dynamics) enables understanding of the ecological mechanis ms and processes that drives changes in the landscapes. Thus, the spatio-temporal anal ysis of landscape is a necessary basis for a mechanistic linkage between particular species or human being and the changing characteristics of the landscape. Objectives The overall objective of this research was to evaluate temporal and spatial land use influences on time series nutrient concentrations and loads measured in canals discharging to Biscayne Bay. To achieve this goal, land use indicators were used to quantify and compare human disturbance gradients in th e watershed to historical water quality data. Specific objectives were as follows: 1) Quantify and compare three disturbance indicators (landscape metr ics, LDI index, and percent imperviousness) in the Biscayne Bay watershed by analyzi ng five sub-basins representing agricultural, urban, and mixed land uses; 2) Evaluate nutrient water qualit y data at monitoring sites in the Biscayne Bay watershed by determining concentration trends, estimati ng annual loads, and cal culating a pollutant index; and 3) Evaluate land use-water quality relationships in the Biscayne Bay watershed to determine if disturbance indicators within sub-basins, canal buffers, or site buffers explain more of the variability in nutrient loads at monitoring sites. Significance of Study The South Florida Water Management Distri ct (SFWMD) and the US Army Corps of Engineers manage a complex system of draina ge canals, pumps, levees, and municipal well fields (Figure 1-3). Without these structures the region could not adequately support an expanding population or protect urban complexes and agricultur al fields from seasonal floods. Canals contain gated control stru ctures that release excess water during the wet-season (May to November) and recharge groundwater during the dry season (November to May). Waterconservation areas also help with flood protection, groundwater recharge, and prevention of saltwater intrusion.

PAGE 34

34 Water management practices, agricultural development, and increased urbanization have directly affected surface and groundwater systems in south Florida. For example, the canal conveyance system has altered th e ecology of the region by acceler ating freshwater flow from the Everglades to the Atlantic Ocean (Leach et al 1972); freshwater inputs to Biscayne Bay have also been modified from natural pathways of continuous submarine discharges and overland sheet flow to periodic surface water releases at canal outlets (Langevin 2001). Ecological and hydrological processes within the ba y and aquifer are directly and indirectly linked to population trends and land use dynamics within the region. Ni neteen canals discharge into the bay and water quality in the bay has declined during the 20th century as south Floridas population has increased (Cantillo et al. 2000). Canals bring pollutants from the watershed direct ly to the north, central, and south sections of Biscayne Bay with land use patterns (Table 1-1; Figures 1-4; 1-5; 16) influencing pollutant characteristics (Caccia and Boyer 2005). The he avily urbanized northern bay, which includes Miami Beach and several industrial complexes, st ruggles with sewage discharges, high nutrient loads, turbidity, and heavy meta ls (Alleman et al. 1995).The boundari es of the Biscayne National Park begin in the central bay where problem s such as solid wastes, metals, and fuel/oil pollution are prevalent and extends to the relatively less developed southern bay, which receives discharges from agricultural runoff and toxic contaminants from the Homestead Air Force Base (Caccia and Boyer 2005). The characteri stics of southern bay pollutants are changing however, as developers build new homes in Home stead at a rapid pace on former agricultural and sensitive lands. As agricultural and natural areas are converted to urban uses, understanding the spatial and temporal characteristics of critical pollutant sources will be key to mitigating environmental

PAGE 35

35 contaminants. Watershed managers therefore need to consider both obvi ous and subtle effects of human activities. Obvious effects include hydrological impairment due to increased impervious surfaces while subtle effects include time lags, biological legacies, and cumulative impacts (McDonnell and Pickett 1997). By incor porating concepts from landscape ecology and hydrology within GIS applications, watershed analysis may reveal critical, s ubtle effects that are not reflected in raw water quality data. Canals, fo r example, function as point source discharges to Biscayne Bay and reflect complex interactions between urban and agricultural elements within the watershed. Biophysical data can indicate th e source of contaminants but water quality management and conservation plans require deta iled analyses focused on underlying pollutant processes. The Biscayne Bay watershed in south Florid a has undergone rapid transformation in the 20th century and development patterns as well as agricultural operations will continue to influence land use dynamics. In addition, the Comp rehensive Everglades Restoration Plan will eventually affect the quantity and quality of freshwater discharges to Biscayne Bay (USACE and SFWMD 1999; Browder et al. 2005). Understanding the contex t of environmental change throughout the watershed is therefor e crucial to the long-term hea lth and protection of Biscayne Bay. Scope of Study This study included temporal st atistical analyses on land use and water quality data from sites within the Biscayne Bay watershed (Figure 1-1) to evaluate the im pact of anthropogenic activity. GIS data layers (SFWMD 2009) for the watershed boundary, land use/land cover (LULC) for three different years (1995, 1999, and 2004) (Figures 1-4; 1-5; 1-6), and monitoring site locations were used to calculate anthropoge nic disturbance indicator s at different spatial extents in five sub-basins. Class categories in the LULC layers were based on the SFWMD

PAGE 36

36 modified version of the Florida Land Use a nd Cover Classification System (FLUCCS), which was originally developed by the Florida Departme nt of Transportation (F DOT 1999). Historical (1992 to 2006) nutrient water quality data from m onitoring sites within these five sub-basins were used to evaluate water quality and to identify significant correlations between land use variables and selected parameters (nitrate/nitrite-nitrogen, total ammonia nitrogen and total phosphorus). Water quality datasets were statistically evaluated using trend analysis techniques that are available in the US Geological Surv ey (USGS) add-in pack age ESTREND and SPLUS software. Load estimations were completed using LOADEST (Load Estimator) from USGS. Stepwise multiple regression models were used to determine the relative contribution of specific land use classes to water quality data and to pr edict water quality impacts in the watershed in response to future land use change.

PAGE 37

37 Table 1-1. Land use/land cover data for th e Biscayne Bay watershed (1995, 1999, and 2004). Land use/Land cover class 1995 (%)1999 (%) 2004 (%) Natural land/water 31.9 32.2 30.0 Improved pastures 1.5 0.7 0.5 Low intensity pastures 1.1 0.1 0.1 Medium intensity recreational, open space 4.8 4.9 4.9 Tree crops 5.0 5.6 6.3 Row crops 10.2 8.9 7.3 High intensity agriculture 0.1 0.0 0.0 High intensity recreational 1.7 1.9 2.0 Low density single family residential 5.2 4.6 4.6 Medium density single family residential 16.8 16.9 18.1 High density single family residential 1.8 2.5 3.1 Institutional 1.9 2.5 2.7 Low density multifamily residential 3.3 3.7 4.0 High intensity transportation 3.5 3.5 3.6 Low intensity commercial 3.0 4.2 4.6 Industrial 6.3 6.0 6.6 High intensity commercial 1.1 0.8 0.9 High density multifamily residential 0.8 0.8 0.8

PAGE 38

38 Figure 1-1. Biscayne Bay watershe d in southeastern Florida.

PAGE 39

39 A) B) Figure 1-2. Examples of water management systems in south Florida. A) Canal. B) Flow control structure (S194) operated by the South Florida Water Management District.

PAGE 40

40 Figure 1-3. Canals and drai nage sub-basins in the Biscayne Bay watershed.

PAGE 41

41 Figure 1-4. Watershed land use/land cover map ( 1995). Land use/land cover data from the South Florida Water Management District (SFWMD 1995).

PAGE 42

42 Figure 1-5. Watershed land use/land cover map ( 1999). Land use/land cover data from the South Florida Water Management District (SFWMD 1999).

PAGE 43

43 Figure 1-6. Watershed land use/land cover map ( 2004). Land use/land cover data from the South Florida Water Management District (SFWMD 2004).

PAGE 44

44 CHAPTER 2 EVALUATING LAND USE CHAN GE (1995 TO 2004) USING MULTIPLE DISTURBANCE INDICATORS IN THE BISCAYNE BAY WATERSHED, FLORIDA Introduction Variability, in terms of extent and intens ity of human land uses, creates disturbance gradients that can potentially al ter processes such as nutrient cy cling, energy flows, and pollutant export (Newcombe 1977; Turner 1989; McDonnell et al. 1997; Alberti et al. 2007). Landscape ecology provides a conceptual framework to understand anthropogenic influences by focusing on land use patterns, interactions among different landscape elements, and the effects of changes in the spatial heterogeneity complex over time (R isser et al. 1984; Haines-Young et al. 1994). A fundamental concept in landscape ecology is that patterns influence processes; thus, several studies have provided methods to characterize spa tial heterogeneity (e.g., Forman and Godron 1986; ONeill et al. 1988; Turner and Gardner 1991). Geographic Information System (GIS) software is an important tool in this pro cess because it allows for complex computational analyses in a relative simple and time efficient manner. GIS software also enables the integration of multiple historical datasets (e.g., digital land use data), wh ich create opportunities to study temporal patterns in heteroge neous landscapes. Hundreds of metr ics (or landscape variables) can be obtained from temporal and spatial land use analyses utilizing math ematical operations in GIS. However, specific metrics have been developed for both landscape composition (relative amounts of different elements in the landscape) and configuration (a rrangement of these elements) that can be used to assess and compare landscape pr ocesses (Turner 1989; Li and Wu 2004). Human disturbance gradients ha ve also been quantified usin g landscape indices to assess relative impact (e.g., McMahon and Cuffney 2000; Wang et al. 2008). One such index is the Landscape Development Intensity (LDI) (Brown and Vivas 2005) which considers nonrenewable

PAGE 45

45 energy use among land use and land cover (LULC) classes. The LDI index is based on emergy, a concept developed by systems ecologist H.T. Od um (Odum 1971; Odum 1996) that derived from decades of analysis on the differential ability of various forms of energy to do work (Brown and Ulgiati 2004). Emergy is a useful concept in environmental acc ounting because it measures the amount of energy that is directly and/or in directly associated with both natural and anthropogenic products and services. Emergy, in esse nce, looks back upst ream to record what energy went into the train of transformation processes (Odum 1996) to produce specific landscape features. Thus, LULC classes (e .g., row crops and commercial areas) have characteristic energy transformations that can be described by the change in energy that has occurred due to landscape di sturbance (Odum 1996; Brown and Vivas 2005). Several studies have utilized the LDI index to quantify dist urbance gradients such as Mack (2006), who evaluated Ohio wetlands and concluded the LDI index compared favorably with other assessment tools. In Florida, the LDI index was strongly correlated to a wetland biological integrity index (Reiss 2006; Reiss and Brown 2007) and Lane and Brown (2006) determined that the LDI index explained more of the variation in benthic diatom species within freshwater marshes compared to other landscape metrics (e .g., percent agriculture land use and percent urban land use). As the LDI concept was a pplied under different conditions, important limitations of the index were that human dist urbance intensity was not related to background conditions and LDI coefficients were restricted to a predetermined set of LULC classes. A revised LDI method (Reiss et al. 2009) addressed these issues. In addition to landscape metrics and indice s, another common measurement used to evaluate LULC characteristics and associated impacts is im perviousness. Arnold and Gibbons (1996) described four qualities of imperviousne ss: (1) impervious surfaces contribute to

PAGE 46

46 hydrologic changes that impair water quality; (2) imperviousness is a charact eristic of intensive land use activities; (3) impervious surfaces hi nder pollutant processing by disrupting soil percolation; and (4) impervious surfaces efficien tly transport pollutants into receiving waters. Studies have used impervious surfaces such as buildings, parking lots, an d roads as a measure of the extent of development within landscapes and have linked these variables to overall water quality (e.g., McMahon and Cuffney 2000; Roy and Shuster 2009). Relatively low levels of watershed imperviousness can produ ce negative effects in aquatic systems; for example, Arnold and Gibbons (1996) characterized streams within watersheds containing <10% of impervious cover as protected, 10-30% as impacted, and greater than 30% as degraded. Linking an imperviousness threshold to water quality can be challenging, however, because many studies do not differentiate between total and effective impe rvious cover within watersheds (Brabec et al. 2002). Total impervious areas (TIA) include surf aces that may drain to pervious ground while effective impervious areas only include impervi ous cover directly connected to waterways. Using TIA instead of directly connected impervious areas (DCIA) in water quality investigations may therefore obscure the influence of land use changes (Alley and Veenhuis 1983). Reducing DCIA is an important component of low im pact development (Wright and Heaney 2001) because it reflects the potentia l influence of human devel opment on adjacent systems. Indicators provide information on the condi tion of landscapes (Dale 2001; Bolliger 2007) and multiple indicators addressing different aspects of land use change can help to reveal broader impacts of human disturbance. Although previous studies have compared the efficacy of multiple indicators (e.g., McMahon and Cuffney 2000; Ge rgel et al. 2002), there has been no study comparing landscape metrics, the LDI inde x, and DCIA. Landscape metrics and percent imperviousness have been used extensively to investigate human impacts (e.g., Ritters et al.

PAGE 47

47 1995; Arnold and Gibbons 1996; Alberti et al. 2007) but the LDI index is a relatively new index that can potentially provide unique information to help managers monitor and evaluate the effects of land use change. An ideal location for evaluating these indicato rs together is south Florida because rapid land use transformations in recent decades ha ve impacted its native ecosystems (McPherson and Halley 1996; Solecki and Walker 2001) and both urban and ag ricultural development continues to influence critical areas such as Biscayne Bay, the receiving water body for the Miami metropolitan area. Biscayne Bay is both ecologi cally and economically important to the region because its tropical reefs and mangroves support various species (manatees, dolphins, wading birds, etc.) as well as fishi ng and recreational industries ( BBPI 2001; Browder et al. 2005). Human-dominated landscapes, such as the Biscayne Bay waters hed, are complex mosaics where heterogeneous human activities gradually transf orm biophysical characteristics (Dow 2000) and understanding the context of envi ronmental changes throughout the watershed is crucial to the long-term health and protection of the bay. T hus, the goal of this study was to quantify and compare three disturbance indicators (la ndscape metrics, LDI index, and percent imperviousness) in the Biscayne Bay watershe d by analyzing five s ub-basins representing agricultural, urban, and mixed land uses fo r 1995, 1999, and 2004. Specific objectives were as follows: (1) quantify human disturbance indicator s in the five sub-basins, (2) determine if selected disturbance indicators provide contrasting information, and (3) evaluate how these indicators could potentially influenc e watershed management decisions. Methods Study Area Biscayne Bay, a barrier-island subtropical estuary, includes the federally protected Biscayne National Park and is located along the so utheastern Florida coas tline. Extensive urban

PAGE 48

48 and agricultural development in the watershed (2,500 km2) have thrived on former wetlands as canals have lowered water tables (Parker et al. 1955), reducing waters hed water storage and creating polluted discharges that degrade sensitive estuarine habitats (Browder et al. 2005). The watershed is primarily located in Miami-Dade Co unty, which includes the city of Miami, but the northern section extends into Broward County; the western bo undary of the watershed lies adjacent to the Florida Everglades and the Everglades National Park. The South Florida Water Management District (SFWMD) sub-divided the watershed into several sub-basins based on major canal structures. Five sub-basins (Figure 2-1) were selected based on data availability and these sub-basins represented different type s of LULC within the watershed, including agricult ural, urban, and mixed-land use. C-9 East (118 km2; referred to as C-9 hereafter), C-8 (71 km2), and C-7 (82 km2) sub-basins are located in the northern section of the watershed, which is primarily characterized by urban land uses. In the central section of the watershed, the C-1 (117 km2) sub-basin includes extensive urba n and agricultural land uses and was selected as an example of a mi xed land use sub-basin. The C-103 (113 km2) sub-basin, located in the southern section of the watershed within the South Dade Agricultural Area, is dominated by agricultural land us es such as row and tree crops. Analyses were limited to subsections of both C-1 and C-103 sub-basins to correspond with locati ons of water quality monitoring stations. Land Use Data LULC GIS data layers were obtained from SFWMD for three separate years: 1995 (scale 1: 40000), 1999 (1: 40000) and 2004 (1: 12000). SF WMD created all three layers by photointerpreting aerial photography and digital orth ophotographic quarter qu adrangles (DOQQs). Each layer used a modified form of the Fl orida Land Use and Cover Classification System (FLUCCS; FDOT 1999) as SFWMD FLUCCS codes primarily use community level classes to

PAGE 49

49 identify vegetation. To simplify analysis, land use classes were aggregated into 18 natural, agricultural, and urban classes (T able 2-1). Vector LULC data we re converted to raster format for spatial analysis using a common scale (190 x 190 meters grid cell size). To ensure the accuracy of LULC data for each of the three years obtained (1995, 1999, and 2004), DOQQs corresponding to the timeline of the data layers were retrieved from the Land Boundary Information System (LABINS) of the Florida Department of Environmental Protection. Quality assurance/quality control involve d comparing DOQQs depicting actual LULC within a particular year to assigned LULC classes in corresponding GIS layers. Any GIS cla ssifications that were inconsistent with DOQQs for each of the three data layers were co rrected to reflect actual LULC throughout the watershed. Landscape Metrics FRAGSTATS (McGarigal and Marks 1995), a so ftware package developed to calculate landscape metrics, generates values for several diffe rent categories of metric s that can be useful to understanding LULC changes in watersheds (Table 2-2). Patc h Analyst (Elkie et al. 1999), a modified version of FRAGSTATS designed specifically as an ESRI ArcGIS extension tool, provides an integrated user interface that enable s metrics to be calculated for LULC layers at both landscape and class levels. Landscape-level metrics calculate values with all classes included (e.g., mean patch size within a watershed) while class-level metrics calculate values for specific classes (e.g., mean patch size of row cr ops). Patches are conti guous cells containing single LULC classes grouped together (e.g., row cr op areas and commercial areas) (ONeill et al. 1997) and for each of the three LULC layers (i.e., 1995, 1999, and 2004), 17 landscape metrics and 13 class metrics were calculated for the entire Biscayne Bay watershe d (Table 2-2). Metrics were calculated on a watershed level to identif y important variables throughout the watershed and then applied to the five sele cted sub-basins as an analytical dataset. Area-weighted metrics

PAGE 50

50 (e.g., patch richness density) were preferred to absolute metrics (e.g., patch richness) to compare data from the five sub-basins. Metrics were test ed for normality using the Shapiro-Wilk W test for normality with a p-value < 0.05 (Shapiro and Wilk 1965; Royston 1983). Most metrics deviating from a normal distributi on were either log or square root transformed to improve normality; metrics containing percentage data were arcsin-square root transformed. For large datasets containing multiple variables, it is often easier to analyze this information if the number of vari ables is reduced to a smaller set of components that retains all the important data. Principal Component Analys is (PCA) is a data re duction technique that identifies linear combinations of the original va riables explaining all of the variance in a dataset (Nichols 1977; Bengraine and Marhaba 2003). Th ese linear combinations, or principal components, describe variability in the dataset which are not directly measured. Factor analysis (FA) is another statistical procedure used w ith multi-variable datasets to identify factors contributing to overall variance (McDonald 1985). The difference between PCA and FA is that while PCA attempts to simplify variable inte rpretation through data reduction, FA primarily focuses on identifying significan t, underlying factors. Pair-wise correlation coefficients were cal culated for transformed landscape and class metrics to eliminate redundancy, with only one metr ic in a correlated pair of metrics included in further analysis if Pearson coefficients were greater than 0.90 (Ritters et al. 1995). Significant landscape and class metrics were then identified for the Biscayne Bay watershed using a correlation matrix to conduct PCA and FA in S-Plus 8.0 (Insightful Corporation 2007). In a correlation matrix, the mean of eigenvalues (a measure the variance explained by each principal component) is one and principal components with above average eigenvalues explain more of the overall variance (Burstyn 2004). The number of principal components with eigenvalues

PAGE 51

51 greater than one therefor e determined the number of factors to use in FA. To aid interpretation, FA included a varimax rotation to reveal metrics that had the strongest co rrelations, or loadings, for identified factors across the three di fferent LULC layers (1995, 1999, and 2004). LDI Index The LDI Index was evaluated on a sub-basin level because it measures the intensity of land use activities within a specific area. Data required to calculate LDI index values for Biscayne Bay sub-basins included LULC GIS layers, areas for each LULC class, nonrenewable empower intensity (emergy per time per area) values for LULC classes, and the renewable empower intensity of the background area. The first step in the LDI calculation process was to sum the areas of each LULC class and express these valu es as a percent of the total landscape area. LULC percentages were then multiplied by thei r respective nonrenewable empower intensity values for Florida (Table 2-3). Finally, LDI i ndex values were calculated using Equation 2-1: LDI = 10 log10 (emPITotal /emPIRef) (2-1) where LDI [unit less] is the Landscape Devel opment Intensity index for sub-basins, emPITotal [sej ha-1 yr-1] is the total empower intensity (sum of renewable background empower intensity and nonrenewable empower intensity of land uses), and emPIRef is the renewable empower intensity of the background environment (Florida = 1.97 E15 sej ha-1 yr-1). The total empower intensity (emPITotal) was calculated as follows: emPITotal = emPIRef + (%LUi emPIi ) (2-2) where %LUi is the percent of the total area in LULC class i and emPIi [sej ha-1 yr-1] is the nonrenewable empower inte nsity for LULC class i Imperviousness The Miami-Dade Department of Envir onmental Resources Management (DERM) developed DCIA reference values for land uses cla sses within the county (Tab le 2-3) to evaluate

PAGE 52

52 pollutant loading estimates under alternate scenarios (DERM 2004). DERM calculated these reference values using aerial maps and measur ing the total imperviousness area within typical land uses occurring in basins th roughout the county. DERM DCIA values for various land uses were used to estimate percent imperviousness in each of the five sub-basins and for each LULC layer (1995, 1999, and 2004). The subbasin level was used for imperviousness evaluation to compare DCIA values across the five sub-basins. Results Landscape Metrics Several pairs of metrics were significantly correlated (greater than 0.90) at both landscape and class levels. At the landscape level, PCA wa s performed on 11 of the 17 metrics calculated and three principal components had eigenvalues greater than one. After varimax rotation and FA, these three factors accounted for 76% of the cumula tive variation of the selected parameters for the entire Biscayne Bay watershed. Metrics with the highest loadings for the first factor, interpreted as patch size variab ility, were mean patch size (MPS; 0.92) and patch size standard deviation (PSSD; 0.91). The largest patch index (LPI; -0.94) and landscape shape index (LSI; 0.86) had the highest loadings for the second fact or, patch diversity. Patch complexity was the third and final factor at the landscape level and included patc h size coefficient of variation (PSCoV; 0.84) and area weighted mean sh ape index (AWMSI; 0.77) (Figure 2-2). Among the 18 land use classes (Tab le 2-1), the two most dominant classes in terms of relative area and distribution w ithin the five sub-basins were row crops and medium density single family residential (MSR). Only PCA and FA results for metrics in these two classes are reported to limit analysis to dominant classe s in the sub-basins. For row crops, PCA was performed on 9 of the 13 class-level metric s calculated and two principal components had eigenvalues greater than one. FA revealed these two factors a ccounted for 66% of the spatial

PAGE 53

53 variability in row crops (Figur e 2-3). The first factor, patc h fragmentation, included mean proximity index (MPI; 0.93), PSSD (0.93), AW MSI (0.87), and PSCoV (0.86). Edge density (ED; 0.80) and landscape percentage (%LAND; 0.80) both had high loadings for the second factor, patch distribution. For MSR, PCA was pe rformed on 8 of the 13 class-level metrics and FA identified three factors that were responsib le for 80% of the cumulative variance (Figure 24). MSR factors were interpreted as patch co mplexity (PSCoV, 0.96; AWMSI, 0.88; and class area [CA], 0.84), patch area (LPI, 0.99 and %LAND, 0.88), and patch fragmentation (mean nearest neighbor [MNN], -0.56 and MPS, 0.55). LDI Index and Land Use Percentages LDI values for all five sub-basins ranged fr om 25.4 to 31.0 (Figure 2-5), which reflects a substantial difference in the level of anthropoge nic disturbance as 25.4 is approximately four times greater than 31.0 on the logarithmic LD I scale. In 2004, MSR (37.6%), high intensity transportation (14.4%), and low intensity commer cial (LIC; 10.2%) land use classes dominated the C-8 sub-basin, which produced the highest LD I value (31.0). No other 2004 land use class in the C-8 sub-basin had a land use percentage that was greater than 10%. Between 1995 and 2004, C-9, C-8, and C-7 land uses were primarily MSR (35.9 to 51.7% in C-9, C-8, and C-7), LIC (8.0 to 11.7% in C-8 and C-7), and hi gh intensity transportation (14.4 to 15.5% only in C-8) (Tables 2-4; 2-5; 2-6). From 1995 to 2004, LDI values in the three urban sub-basins (29.6 to 31.0) were greater than the mixed land use (C-1; 27.7 to 28.5) and agricultural (C-103; 25.4 to 26.2) subbasins. LDI values in both C-1 and C-103 subbasins reflect the hi ghest increase in the magnitude of anthropogenic development intensity (F igure 2-5) across the five sub-basins for the study period. In C-1, the mixed land use sub-basin, both ro w crops and MSR land use percentages were 20% or greater for 1995, 1999, and 2004 LULC laye rs. Only one other land use class, high

PAGE 54

54 density single family residential (HSR) in 2004, was greater than 10% between 1995 and 2004 in the C-1 sub-basin. Row crops decreased from 27.6% in 1995 to 21.6% in 2004, MSR increased from 19.7% to 23.4%, and HSR also increased from 5.0% to 12.6%. Four land use classes accounted for greater th an 76% of LULC in C-103, the agricultural sub-basin, from 1995 to 2004: tree crops, row crops, lo w density single family residential (LSR), and MSR (Tables 2-4; 2-5; 2-6). During the st udy period, row crops declined (22.8% to 12.2%) in C-103 while tree crops (26.7% to 32.5%) and the residential cla sses (LSR and MSR) increased (26.8% to 31.4%). Increased residential la nd use in the C-103 sub-basin from 1995 to 2004 corresponded to an increase in LDI values (25.4 to 26.2). Imperviousness For all five sub-basins, percent imperviousne ss values for 2004 were greater than both 1995 and 1999 values (Figure 2-5). DCIA was greates t in the three urban sub-basins and values in C-8 and C-7 for all three LULC layers were greater than 30%. Imperviousness in C-8 and C-7 changed minimally from 1995 to 2004 as values ra nged from 33.2% to 35.8%, the highest in this study. In the mixed sub-basin, C-1 (18.2% to 23.3 %), DCIA was lower than in the three urban sub-basins and C-103 (13.6% to 16.8%), primarily an agricultural sub-basin, had the lowest overall values. The greatest changes in DCIA during the study period occurred in the C-1 (5.0%) and C-103 (3.2%) sub-basins (Figure 2-5). DCIA a nd LDI values for all five sub-basins had a positive linear relationship, with an R2 value of 0.97 (Figure 2-5). Discussion Landscape Metrics Similar to McGarigal and Marks (1995) a nd Li and Wu (2004), many of the landscape metrics were highly correlated. Using PCA and FA, patch size va riability, patch diversity, and patch complexity were identified as significant f actors contributing to over all spatial variability

PAGE 55

55 at the landscape-level (i.e., consid ering all LULC classes) for all sub-basins in the Biscayne Bay watershed. The first factor, patch size variab ility, included both MPS and PSSD and indicated patch size distribution among LULC classes varied significantly in sub-basins throughout the watershed (Figure 2-2). Between 1995 and 2004, MPS and PSSD values suggested that there was great variability in patch sizes in the five study sub-basins. For C-103, the agricultural sub-basin, an increase in PSSD (+0.7 km2) suggests that smaller patche s were being created although several large patches remained in tact. Conversion of agricultural and natural areas to residential or urban complexes has been the predominant fo rm of land use change during the last three decades in south Florida (Solecki and Walker 2001) and this is becoming increasingly evident in C-103. However, not all agricultura l areas in C-103 have been developed for other uses. Row crops in C-103 have been converted as resident ial areas expand but this is only occurring in limited sections of the sub-basi n, thereby leaving large tree cr op patches undeveloped. In C-1, the mixed land use sub-basin, PSSD has actually declined (-0.3 km2) during the study period because the conversion of agricultural areas is occurring in multiple areas throughout the subbasin. The second landscape-level factor, patch dive rsity, included both LPI and LSI metrics and continued to highlight the cont rast between C-103 and C-1 (Figure 2-2). LPI is a percentage measure of class dominance as it assesses the size of the largest single patch relative to total subbasin area and LSI reflects patch shape comple xity and edge density (McGarigal and Marks 1995). In C-103, LPI increased between 1995 and 2004 (7.8 to 23.0) and for the same period in C-1, LPI declined (24.5 to 19.1). As row crops were being converted to di fferent land uses in C103, tree crops became the largest contiguous land use areas because reside ntial development did not reduce tree crops. Row crops have been simila rly converted to reside ntial land uses (MSR

PAGE 56

56 and HSR) in C-1 but LPI values in this sub-basin still reflect large row crop patches. Therefore, as row crops declined in C-1 in favor of MS R and HSR, there was a corresponding decline in LPI values. The largest patch in all three urban sub-basins was MSR but LPI values changed the least in the urbanized sub-basins, with both C-9 and C-8 having some of the lowest LPI values. The lower LPI values for C-9 and C-8 indicated th ese sub-basins had a greater diversity of urban land uses compared to C-7, which was more homogenous and dominated by MSR. C-9 and C103 had similar values for the LSI metric although dominant land us e classes in these sub-basins were different and this may have been due to increasingly fragmented landscapes. Fragmentation leads to increasing amounts of edge (McGarigal and Marks 1995) and in C-9, improved pastures have been drastically reduced (5.5 to 0.76 km2) and splintered as MSR and low intensity multifamily residential land use classes have increased. The third landscape-level factor was patc h complexity and included both PSCoV and AWMSI (Figure 2-2). PSCoV is dependent on both PSSD and MPS and in the five sub-basins, PSCoV was correlated (greater than 0.90) to PSSD. Consequen tly, sub-basin relationships discussed earlier for PSSD in the first factor, patc h size variability, were similar to PSCoV. Both PSCoV and PSSD were included in the analysis because when cons idering the entire Biscayne Bay watershed, these two metrics were not correlated. AWMSI assess es the irregularity of patch shapes, with lower values indicating uniform shapes (McGarigal and Marks 1995). C-1 and C-9 had similar AWMSI values, with C-8 having the lowe st values compared to the other sub-basins. Although C-7 generally had the highest AWMSI va lues, the greatest chan ge occurred in C-103 (2.23 to 3.18). Row crops generally have uniform shapes in the landscape and as row crops declined in C-103 and MSR and LSR increased, this may have potentially led to greater patch shape irregularity in C-103.

PAGE 57

57 Row crops and MSR were the most widely di stributed land use classes throughout the five sub-basins and only factors for these two classe s were considered. Row crops were primarily found in C-1 and C-103 while MSR was a common feat ure in all five sub-basins. Fragmentation and patch distribution factors revealed the spatial configuration of row crops within C-1 and C103 sub-basins (Figure 2-3). MPI, PSSD, and PS CoV values indicated large row crop patches occurring closer together are more prevalent in C-1 and edge densi ties suggested a greater distribution of LULC classes be tween row crops in C-103. Therefore, row crops have smaller areas and are more fragmented in C-103 compar ed to C-1. MSR had three factors controlling spatial variability in the five sub-basins: patch complexity, area, and fragmentation (Figure 2-4). Generally, MSR patches in urban sub-basins were larger, more complex (i.e., considering size, shape, and area) and had greater connectivity than in C-1 and C-103. The urban center of the Biscayne Bay watershed includes C-9, C-8, and C-7 sub-basins whereas C-1 and C-103 are at the fringes of urban development. Fr agmentation is generally low in urban centers because of the concentration of urban land uses (such as MSR) and urban fringes typically exhibit a greater degree of fragmentation as the process of urbanization occurs (Herold et al. 2003; Weng 2007). After PCA and FA, factors for row crops (66%) and MSR (88%) had diffe rent variability and this may be due to the generally stable natu re of urban sub-basins during the study period, compared to land use changes occurring in C-1 and C-103. LDI Index and Imperviousness The three urban sub-basins (C-9, C-8, and C7) had relatively little change in their LDI values compared to C-1 and C-103 (Figure 2-5) Brown and Vivas (2005) described LDI values as representing a combination of several anthropogenic influences occurring in developed landscapes such as pollutants in both ai r and water, physical landscape damage, and environmental modifications (e .g., hydrological impairment leadi ng to increased flooding risk).

PAGE 58

58 Therefore, in the three urban sub-basins, dom inant land use classes such as MSR and low intensity commercial (LIC) did not change substantially from 1995 to 2004 and resulted in LDI values that remained relatively constant. The agricultural sub-basin, C-103, had the lowe st overall LDI values but the conversion of row crops to residential uses (LSR and MSR) especially between 1999 and 2004 increased development intensity at a rate similar to the mixed land use sub-basin, C-1 (Figure 2-5). Westward residential development in C-1, from the coastal areas inland, has led to the gradual decline of row crops and the concurrent expansion of both MSR and HSR land use classes. Interestingly, although resident ial areas increased in both C-103 and C-1 from 1995 to 2004, the types of residential developmen t occurring in these sub-basins were different. LSR and MSR increased in C-103 but in C-1, MSR and HSR land use classes were most responsible for the decline in row crops. LSR has a lower developmen t intensity (i.e., less no nrenewable energy use per unit area) than HSR and as row crops become replaced by residential land uses, the LDI index indicates that C-1 is gradually attaining th e characteristics of the three urban sub-basins. The strong correlation between DCIA and LDI va lues (Figure 2-5) suggests both indicators provided similar information regard ing the intensity of human distur bance in the five sub-basins. The LDI is a continuous index evaluating both urban and agricu ltural land uses (Brown and Vivas 2005) while DCIA primarily reflects th e impact of watershed urbanization. The LDI index, by incorporating anthropogeni c intensity associated with agricultural land uses, has the potential to provide a more c onsistent link between all aspe cts of human disturbance and resultant ecosystem effects. For example, DCIA percentages in agricultural areas are minimal and may not accurately reflect the extent of human involvement required to sustain agricultural production or potential pollutants th at could be generated. The LDI index is therefore useful in

PAGE 59

59 evaluating the impact of a broad range of land us es on adjacent systems, such as the biological community structure in Florida wetlands (Reiss et al. 2009). DCIA is an integrative measure that can be used to evaluate various aspects of urban development contributing to th e increased distribution of im pervious surfaces (Arnold and Gibbons 1996). As a result, DCIA for the five sub-ba sins reflected changes in the distribution of urban land use classes from 1995 to 2004. Predictably, the three urbanized s ub-basins had greater DCIA values than C-1 and C-103 (Figure 2-5). DCIA changed the least in the urbanized subbasins, which would indicate that C-9, C-8, and C-7 have alrea dy been extensively developed and additional modifications to these sub-basins would not necessarily lead to changes in DCIA. Both C-1 and C-103, in contrast, had agricultural, natural, and open land areas at the beginning of the study period and increased urbanization in these sub-basins led to the creation of more impervious surfaces. The three urbanized sub-basins are less likely to be substantially modified by increased DCIA percentages beca use of their already well-devel oped characteristics but this may not apply in C-1 and C-103. The greater ra te of DCIA changes in C-1 and C-103 could potentially affect water resources as new urban development proj ects continue to expand into previously pervious areas. Management Implications The disturbance indicators suggest ed that the three ur ban sub-basins were relatively stable and dominated by complex MSR patches that corr esponded to a greater degree of anthropogenic intensity compared to the mixed land use and agricultural sub-basi ns. Similarly, Lee et al. (2009) determined that in areas where urban land uses re present the largest patch, water quality declines. Water quality issues in the nor thern Biscayne Bay watershed, where the urban sub-basins are located, include sewage discharg es, high nutrient loads, turbidit y, and heavy metals (Alleman et al. 1995) and are unlikely to be altered by land use changes that do not fundamentally shift

PAGE 60

60 overall landscape characteristics. Mitigating impacts from existing land use influences, such as MSR, in these sub-basins through best management practices (BMPs) w ould help to alleviate established water quality issues. Retrofitting of stormwater outfalls and the construction of grassed swales and French drains have been utilize d to intercept the first flush of pollutants (i.e., first inch of runoff) in the Biscayne Bay wate rshed (Alleman et al. 1995). These pollutant-control strategies have improved water qua lity discharges to the bay but increased treatment efficiency could be obtained by using distur bance indicators to target urban complexes with a greater proportion of intensive land use activities lo cated in hydrologically sensitive areas. Unlike the urban sub-basins, critical issues in both C-1 and C-103 were changes in the composition and spatial distribution of residential and agricultural land use classes. In C-1, MSR and HSR land use classes extended further inland and replaced row crops and this has led to changes in the landscape struct ure of C-1. Numerous studies ha ve linked land use with water quality (e.g., Osborne and Wiley 1988; Johnson et al. 1997; Harman-Fetcho et al. 2005), as the proportion and spatial arrangement of LULC within watersheds can have significant influences on water resources (e.g., Hunsaker and Levine 1995; Roth et al. 1996; Johnson et al. 2001). Furthermore, in C-1, the increased developmen t intensity associated with MSR and HSR land use classes, compared to row crops, also affect s the type of possible pollutants. A landfill and possibly a wastewater treatment plant at the furthe st point downstream in C-1 influence pollutant characteristics from this sub-basin (Meed er and Boyer 2001; Cacci a and Boyer 2005) but upstream land use changes could st ill increase pollutant loads di scharged to Biscayne Bay. New residential developments are also a concern in C-103, along with increased agricultural land (tree crops) that could potentially cont ribute additional pollutants.

PAGE 61

61 Watershed managers can therefore use inform ation provided by distur bance indicators to identify and evaluate the mechanisms of land use change and forecast impacts on water quality (Roy and Shuster 2009). Dispersed urban landscape patterns have been linked to degradation of water quality (Lee et al. 2009) a nd it has been shown that de velopment patterns within landscapes typically affect the extent of anthropogenic influences on biophysical processes (Alberti 2005). C-1 and C-103 subbasins have experienced the greatest increase in LDI and DCIA values during the study period but lands cape metrics indicate that compact (HSR) development patterns are more prevalent in C-1 compared to C-103, where low intensity development is relatively scattered. Watershed managers therefore need to consider both obvious and subtle effects of human activ ities. Obvious effect s include hydrological impairment due to increased imperviousness while subtle effects include underlying factors such as cumulative impacts (McDonnell and Pickett 1997). LDI and DCIA values provide an overall view of watershed development but the location of intensive land uses compounds the threat to water resources. Landscape metrics describing spa tial configurations therefore complement LDI and DCIA values by providing information on watershed development that might not be immediately apparent. Watershed management strategies for assessi ng developing sub-basins can include all three disturbance indicators. Similar to C-1 and C-103, LDI values below 30.0 and DCIA values below 25% (Figure 2-5) reflect sub-basins that are not completely urbanized. Identifying important landscape metrics contribu ting to overall spatial distribut ion in these sub-basins helps to reveal factors that can influence water qualit y. Disturbance indicators can therefore be used together to develop watershed policies that addr ess specific relationships between land use and water quality. Implications for muni cipality zoning regulations in urbanizing sub-basins include

PAGE 62

62 promoting MSR and HSR development, as opposed to LSR, at greater distances from aquatic corridors (Schiff and Benoit 2007). Implementing BM Ps in these critical areas would also aid zoning regulations and reduce hydrologic impacts of gradually increasing LDI and DCIA values. Conclusion During the period of analysis (1995 to 2004), all three disturbance indicators revealed different levels of anthropogenic disturbance among urban (C-9, C-8, and C-7), mixed land use (C-1), and agricultural (C-103) sub-basins Landscape metrics provided information on influential land use classes within the five sub-basins and spatial processes occurring over time that have contributed to variabil ity at the landscape and class levels. In contrast, DCIA and LDI values provided similar information on the inte nsity of human disturbance; urban sub-basins were the most disturbed but the greatest chan ges occurred in C-1 and C-103. Residential and agricultural land use classes were most responsi ble for the variability in C-1 and C-103 as the proportion of residential areas in creased and row crops declined. Although residential areas in C1 and C-103 both increased, these sub-basins feat ured different developm ent patterns. Medium density single family residential (MSR) and high density single family residential (HSR) land use classes increased in C-1 while MSR and low density single family residential (LSR) increased in C-103. HSR and LSR land use classes represent different levels of human intensity and indicate C-1 was becomi ng increasingly urbanized through compact development. Disturbance indicators can provide complementary information for watershed management decisions regarding water quality. Strategies to reduce the impacts of existing land use influences such as detention systems are nece ssary for sub-basins that are highly developed and display little variability for disturbance indicators. In urbanizing watersheds, however, disturbance indicators are likel y to reveal the effects of in cremental land use changes and potential impacts on water resources Planning initiatives such as zoning regulations are critical

PAGE 63

63 because the spatial distribution of intensive land uses and impervious surfaces can adversely affect watershed hydrology, especially in urbanizing watersheds. Rapid urbanization is ongoing in south Florida and further research is needed to explore the relati onship between land use characteristics and water quality variables, su ch as constituent loads, to link development patterns to quantifiable effects in the watershed and in Biscayne Bay.

PAGE 64

64 Table 2-1. Description of land use/land cover classes. Land use/Land cover class1 Description Natural land/water For est; wetland; rangeland Improved pastures Improved pastures Low intensity pastures Unimproved pastures; woodland pastures Medium intensity recreational, open space Community recreational facilities; open land; disturbed land Tree crops Groves; orchards; nurseries; vineyards Row crops Field crops High intensity agriculture Feeding operations; specialty farms High intensity recreational Golf courses; parks; zoos Low density single family residential Less than two dwelling units per acre Medium density single family residentia l Two to five dwelling units per acre High density single family residential Gr eater than five dwelling units per acre Institutional Government; educational; religious; medical Low density multifamily reside ntial Multiple dwelling units (less than three stories) High intensity transportation Airports; highways; port facilities Low intensity commercial Mixed use commercial; professional services Industrial Mineral processing; chemical processing; utilities High intensity commercial Retail sales; shopping centers High density multifamily residential Multiple dw elling units (greater than three stories) 1FDOT (1999).

PAGE 65

65Table 2-2. Description of landscape metrics. Category Metric Description Area Class area Sum of patch areas (C)1 Percentage of landscape Percen tage of landscape covered (C) Largest patch index Percentage of landscape comprise d by the largest patch (L or C) Density and Size Mean patch size Mean patch size (L or C) Patch size coefficient of variation Co efficient of variation (L or C) Patch size standard deviation Standard deviation of patch areas (L or C) Patch richness density Variety of pa tch types relative to area (L) Edge Edge density Perimeter of adjacent patches with different land uses relative to area (L or C) Shape Area weighted mean shape index Area weighted mean patch fractal dimension Shape comple xity; increases with irregul ar shapes (L or C) Landscape shape index Isolation and Interspersion Mean nearest neighbor distance Mean distan ce between similar patch types (L or C) Mean proximity index Mean distance between similar pa tch types within a search radius (L or C) Interspersion juxtaposition index Relativ e adjacency of patc h types (L or C) Diversity Shannons diversity index Shannons evenness index Simpson's evenness index Relative patch diversity (L) Modified Simpson's diversity index Modified Simpson's evenness index 1C indicates metric calculated at the cla ss level and L indica tes landscape level.

PAGE 66

66 Table 2-3. Land use/land cover coefficients for Landscape Development Intensity (LDI) index and percent imperviousness. Land use/Land cover class Non-re newable Directly Connected empower intensity Imperviousness (E15 sej ha-1 yr-1)1 % Natural land/water 0.00 0.00 Improved pastures 2.02 0.00 Low intensity pastures 3.38 0.00 Medium intensity recreational, open space 6.06 0.00 Tree crops 7.76 0.00 Row crops 20.30 0.00 High intensity agriculture 50.40 0.00 High intensity recreational 123.00 0.00 Low density single family residential 197.50 30.00 Medium density single family residential 658.33 38.00 High density single family residential 921.67 38.00 Institutional 4,042.20 27.33 Low density multifamily residential 4,213.33 38.00 High intensity transportation 5,020.00 53.00 Low intensity commercial 5,173.40 45.00 Industrial 5,210.60 45.00 High intensity commercial 8,372.40 53.00 High density multifamily residential 12,771.67 38.00 1 LDI Coefficients from Reiss et al. (2009); sej ha-1 yr-1 = solar emjoules (the available solar energy used during energy transformations) per he ctare per year. DCIA coefficients from DERM (2004).

PAGE 67

67 Table 2-4. Land use/land cover data for five study sub-basins in the Biscayne Bay watershed. Land use/Land cover class 1995 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 7.1 5.3 2.0 7.6 5.7 Improved pastures 4.6 2.1 0.1 0.3 0.8 Low intensity pastures 1.2 0.4 0.0 0.3 0.0 Medium intensity recreational, open space 9.5 8.2 5.1 5.9 4.7 Tree crops 0.2 0.2 0.0 6.8 26.7 Row crops 0.1 0.0 0.0 27.6 22.8 High intensity agriculture 0.0 0.0 0.0 0.2 0.4 High intensity recreational 4.1 2.1 2.3 3.5 1.0 Low density single family residential 1.9 5.1 1.3 5.1 14.0 Medium density single family residential 43.3 35.9 51.7 19.7 12.7 High density single family residential 1.2 1.2 2.1 5.0 0.6 Institutional 3.3 4.2 6.4 4.8 1.0 Low density multifamily residential 6.8 3.1 7.6 2.5 2.4 High intensity transportation 5.4 15.5 3.7 3.3 2.1 Low intensity commercial 3.7 8.0 8.9 2.3 3.6 Industrial 3.8 7.7 5.7 3.6 0.6 High intensity commercial 2.1 0.6 1.9 1.2 0.5 High density multifamily residential 1.7 0.5 1.0 0.4 0.3

PAGE 68

68 Table 2-4. Continued. Land use/Land cover class 1999 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 7.0 5.5 1.9 4.8 4.6 Improved pastures 1.7 0.4 0.0 0.0 0.2 Low intensity pastures 0.4 0.2 0.0 0.0 0.0 Medium intensity recreational, open space 9.9 5.9 4.5 6.3 4.5 Tree crops 0.2 0.2 0.0 6.6 28.2 Row crops 0.0 0.0 0.0 23.5 20.0 High intensity agriculture 0.0 0.0 0.0 0.1 0.2 High intensity recreational 4.4 2.1 2.9 3.9 1.1 Low density single family residential 0.2 4.7 0.7 4.2 15.4 Medium density single family residential 43.9 36.7 48.8 21.1 12.7 High density single family residential 2.1 1.8 2.0 9.1 1.6 Institutional 4.6 4.8 6.8 5.6 1.7 Low density multifamily residential 7.4 4.3 8.3 2.9 2.9 High intensity transportation 5.2 14.6 3.4 3.3 1.7 Low intensity commercial 4.7 10.1 11.6 2.6 3.9 Industrial 4.4 7.9 5.9 3.9 0.7 High intensity commercial 2.1 0.4 2.0 1.5 0.6 High density multifamily residential 1.7 0.7 1.2 0.5 0.2

PAGE 69

69 Table 2-4. Continued. Land use/Land cover class 2004 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 5.3 4.8 1.8 3.0 2.7 Improved pastures 0.6 0.5 0.0 0.0 0.1 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 9.4 4.9 4.7 5.6 4.4 Tree crops 0.1 0.1 0.0 5.2 32.5 Row crops 0.0 0.0 0.0 21.6 12.2 High intensity agriculture 0.0 0.0 0.0 0.1 0.1 High intensity recreational 4.6 2.1 2.9 4.0 1.1 Low density single family residential 0.1 4.2 0.6 3.0 15.6 Medium density single family residential 44.6 37.6 48.9 23.4 15.8 High density single family residential 2.6 1.8 1.4 12.6 2.4 Institutional 4.7 5.1 6.6 5.5 1.8 Low density multifamily residential 8.6 4.9 8.6 3.6 3.4 High intensity transportation 5.0 14.4 3.3 3.3 1.8 Low intensity commercial 5.7 10.2 11.7 2.7 4.4 Industrial 5.0 8.2 6.1 4.4 0.7 High intensity commercial 2.1 0.4 2.0 1.6 0.7 High density multifamily residential 1.5 0.7 1.2 0.5 0.2

PAGE 70

70 Figure 2-1. Five study sub-basins in the Biscayne Bay watershed.

PAGE 71

71A) B) Figure 2-2. Landscape-level factors describing spatial variability in the Biscayne Bay watershed. A) Factor 1 (Mean Patch Size [MPS] and Patch Size Standard Deviation [PSSD]) B) Factor 2 (Largest Pa tch Index [LPI] and Landscape Shape Index [LSI]). C) Factor 3 (Patch Size Coefficient of Variation [PSC oV] and Area Weighted Mean Shape Index [AWMSI]). 0.0 0.1 0.2 0.3 0.4 C-9C-8C-7C-1C-103MPS (km2)Basins 1995 1999 2004 0.0 0.4 0.8 1.2 1.6 2.0 C-9C-8C-7C-1C-103PSSD (km2)Basins 1995 1999 2004 0 10 20 30 C-9C-8C-7C-1C-103LPI (%)Basins 1995 1999 2004 0 5 10 15 C-9C-8C-7C-1C-103LSI (no units)Basins 1995 1999 2004

PAGE 72

72C) Figure 2-2. Continued. 0 200 400 600 C-9C-8C-7C-1C-103PSCoV (%)Basins 1995 1999 2004 0 1 2 3 4 C-9C-8C-7C-1C-103AWMSI (no units)Basins 1995 1999 2004

PAGE 73

73A) B) C) D) Figure 2-3. Selected class-level metrics for ro w crops in the Biscayne Bay watershed. A) Mean Proximity Index (MPI). B) Patch Size Standard Deviation (PSSD). C) Pa tch Size Coefficient of Variation (PSCoV). D) Edge Density (ED). 0 50 100 150 200 C-1C-103MPI (no units)Basins 1995 1999 2004 0 2 4 6 8 C-1C-103PSSD (km2)Basins 1995 1999 2004 0 200 400 600 C-1C-103PSCoV (%)Basins 1995 1999 2004 0 1000 2000 3000 C-1C-103ED (m/km2)Basins 1995 1999 2004

PAGE 74

74A) B) C) D) Figure 2-4. Selected class-level metrics for medium density singl e family residential (MSR) land use class in the Biscayne Bay watershed. A). Patch Size Coefficient of Variation (PSCoV). B) Largest Patc h Index (LPI). C) Percent Landscape (%LAND). D) Mean Nearest Neighbor (MNN). 0 100 200 300 C-9C-8C-7C-1C-103PSCoV (%)Basins 1995 1999 2004 0 10 20 30 C-9C-8C-7C-1C-103LPI (%)Basins 1995 1999 2004 0 20 40 60 C-9C-8C-7C-1C-103% LANDBasins 1995 1999 2004 0 200 400 600 800 C-9C-8C-7C-1C-103MNN (m)Basins 1995 1999 2004

PAGE 75

75A) B) C) D) Figure 2-5. Landscape Development In tensity (LDI) index values and Directly Connect ed Impervious Area (DCIA) percentages for five study sub-basins in the Bi scayne Bay watershed. A) LDI i ndex values. B) DCIA percentage s. C) Change in LDI values from 1995 to 2004. D) Relationship between LDI values and DCIA percentages. 25 27 29 31 33 C-9C-8C-7C-1C-103LDIBasins 1995 1999 2004 0 10 20 30 40 C-9C-8C-7C-1C-103DCIA %Basins 1995 1999 2004 0.0 0.2 0.4 0.6 0.8 1.0 C-9C-8C-7C-1C-103Change in LDIBasins R = 0.97 0 10 20 30 40 2527293133DCIA %LDI

PAGE 76

76 CHAPTER 3 NUTRIENT DISCHARGES TO BISCAYNE BAY, FLORIDA (1992 TO 2006): WATER QUALITY TRENDS, LOADS, AND A POLLUTANT INDEX Introduction Phosphorus and nitrogen are es sential nutrients for aquati c productivity but excessive nutrient inputs can lead to over-enrichment, or eutrophication, of surface waters that produce problems such as algal blooms, decreased disso lved oxygen concentrations, and increased fish mortality (Carpenter et al. 1998; Smith 1998). Eutrophication ranks as the leading pollutant problem affecting the ability of U.S. surface waters to meet desi gnated uses such as recreation, fishing, and irrigation (H owarth et al. 2002). As watershed management plans become more refined to address specific concerns, different methods can be used to evaluate nut rient water quality in bot h healthy and impaired systems. Parametric and nonparametric trend an alysis techniques, for example, can provide important information on temporal changes in nutrie nt concentrations (Hirsc h et al. 1982; Hirsch et al. 1991; Letttenmaier et al. 1991). However, long-term water quality datasets have several characteristics that can complicate trend analysis; there are frequently large gaps in the dataset, data are often skewed, censored data are prevalen t (values less than the minimum detection limit; MDL), and chemical analytical techniques can improve over time producing multiple MDLs for the same parameter. Seasonality (Champley and Doledec 1997; McCartney et al. 2003; Qian et al. 2007) and discharge fluctuati ons (Alley 1988; Hirsch et al. 1991) are other factors that can significantly influence trend anal ysis. Although concentration meas urements are convenient for comparing field data to water quality criteria, additional met hods are available to quantify nutrients, such as loads and yields. Loads assess the mass of constituents transported over time and help to quantify the total amount delivered and yield measurements estimate the mass of

PAGE 77

77 constituents delivered per unit area per unit time, which can help to assess best management practices (Christensen 2001). Another aspect of water quality manageme nt is the acknowledgement that chemical stressors that are dispersed in water can have cumulative and deleterious effects in downstream locations far removed from primary pollutant so urces (Carpenter et al 1998; Howarth et al. 2002). A Pollutant Empower Density (PED) index can be used to assess th e potential impact of these pollutants released over ti me in aquatic systems. The PED index addresses environmental effects of dissolved stressors such as nutrients by analyzing their resp ective emergy per unit time per unit volume (empower densities) relative to background conditions. Emergy is the amount of energy that is directly and/or in directly required to provide a gi ven flow or storage of energy or matter (Odum 1996). Therefore, the PED index describes the change in energy that has occurred in aquatic systems due to the influence of pollu tant discharges. Every element (e.g., nitrogen and phosphorus) has an associated emergy per mass ratio, or unit emergy value (UEV), and as dissolved substances become more concentrated UEVs increase because energy is required to concentrate materials. For elements and compou nds that are rare in nature, more energy is required to concentrate these materials and this results in higher UEVs (Brown and Ulgiati 2004). Nutrients generally have high UEVs and excess concentrations can alter critical ecosystem processes, which can lead to reduced ecosystem function. Analyzing water quality using multiple methods enables watershed managers to develop and implement specific plans targeting factors affecting water resources. Trend analysis and nutrient quantification methods have been used extensively in wa ter quality stud ies (e.g., Hirsch et al. 1982; Runkel et al. 2004; Qian et al. 2007) but the PED index is a new, exploratory assessment tool to evaluate the potential imp act of known pollutant di scharges (Mark Brown,

PAGE 78

78 personal communication). The PED i ndex is especially relevant to sensitive aquatic systems experiencing greater anthropogenic influences, such as Biscayne Bay, a natural oligotrophic estuary in southeastern Florida. Biscayne Bay requires minimal inputs of phos phorus and nitrogen to function and thus watershed nutrient inputs have a controlling influence on bay water quality (Browder et al. 2005). The primary nutrient limiting autotrophic gr owth in Biscayne Bay is phosphorus (Brand 1988; Kleppel 1996) and average phosphorus concentrations are low but variable throughout the bay, with the northern section having greater concentrations (0.008 to 0.020 mg L-1) than other areas (<0.008 mg L-1) (Alleman et al. 1995). Due to these phosphorus concentration differences, Brand (1988) found phytoplankton levels were five times greater in the north than in the south. Nitrogen may not be the limiting nutrient in the bay but elevated nitrate/nitrite concentrations (>4 mg L-1; Cheesman 1989) derived from agricultural a nd some urbanized areas in the watershed are a concern because artificially high concen trations may have subtle ecological effects (Alleman et al. 1995), such as making the bay more susceptible to algal blooms. Changes in land use, management practices, and environmental conditions may all lead to detectable differences in nutrients transporte d to Biscayne Bay. Liet z (2000), for example, analyzed water quality data ( 1966 to 1994) at a site discharging to the bay and found downward trends, indicative of improved water quality, fo r several parameters including total ammonia nitrogen and total phosphorus. The total amount of nutrients entering Biscayne Bay is of particular concern and previous studies have quantified incoming nutrient loads. Lietz (1999) estimated nutrient loads to Biscayne Bay by an alyzing nutrient concentrations and freshwater discharges and Caccia and Boyer (2007) develo ped a nutrient loading budget to the bay by analyzing canals throughout the wa tershed. Both studies determined that nitrate/nitrite-nitrogen

PAGE 79

79 loads were elevated in the southern agricultural drainage areas while amm onia-nitrogen and total phosphorus loads were highest in the northern and central urban dr ainage areas. Canal discharges are therefore important factors in fluencing Biscayne Bay water quality and the goal of this study was to evaluate historical nutrient water quality data from 1992 to 2006 at six monitoring sites located near the outlets of canals discharging to the bay. Quantifying the effects of continued urban and agricultural development in the waters hed is important for adaptive management of the bay and thus specific object ives included the following: (1) determine nutrient concentration trends during the study period; (2) estimate annual nutrient lo ads from six canals in the watershed; and (3) use the PED index to assess the proportional impact of nutrient discharges from various canals. Methods Study Area Biscayne Bay is a barrier-island subtropical estuary that is located along the southeastern coastline of Florida and includes the federally protected Biscayne National Park. Designated as an Outstanding Florida Water, Bi scayne Bay requires substantial fr eshwater inputs to maintain its natural ecological balance; however, water management opera tions in south Florida have disrupted historical freshwater flows to the bay. For example, the canal conveyance system has accelerated freshwater flow from the Everglades to the Atlantic Ocean (Leach et al. 1972) and freshwater inputs to Biscayne Bay have been modified from natural pathways of continuous submarine discharges and overland sheet flow to periodic surface water releases at canal outlets (Langevin 2001). Canals function as point source discharges to Biscayne Bay and reflect complex interactions between urban and agricultural elements within the watershed. Nineteen canals discharge into the bay and water quality has declined during the 20th century as south Floridas population has increa sed (Cantillo et al. 2000).

PAGE 80

80 The Biscayne Bay watershed is primarily lo cated in Miami-Dade County, which includes the city of Miami, but the northern section extends into Broward County. The western boundary of the watershed lies adjacent to the Florida Everglades and th e Everglades National Park. The South Florida Water Management District ( SFWMD) and the US Army Corps of Engineers manage a complex system of drainage canals, pu mps, levees, and municipal well fields in the watershed. Without these structures, the region could not adequately pr otect urban complexes and agricultural fields from seas onal floods. Canals contain gated co ntrol structures that release excess water during the wet season (May to Nove mber) and recharge groundwater during the dry season (November to May). Water quality and flow data from six monito ring sites near the outle ts of six separate canals were used in this study (Figure 3-1). Sites SK02, BS04, LR06, BL03, and MW04 were the primary focus of the study and the sixth site, AR03, was used as a reference or baseline. Selected canals were located in sub-basins containing different types of land use/land cover (LULC) such as agricultural, urban, and mixed-land uses SK02, BS04, and LR06 are located on the C-9, C-8, and C-7 canals, respectively; these canals are in the northern section of the watershed, which is primarily characterized by urban land uses. BL 03 is on the C-1 canal, located in the central section of the watershed surrounded by extensiv e mixed (urban and agricultural) land uses. MW04 is on the C-103 canal in the South Dade Agricultural Area, a region dominated by agricultural land uses such as row and tree crops. AR03, located on C-111, is surrounded by wetlands in the extreme southern section of the watershed. Water Quality Data Miami-Dade County Department of Envir onmental Resources Management (DERM) collects monthly grab samples fr om water quality monitoring site s throughout Biscayne Bay and in watershed canals. DERM uses EPA met hods 353.2, 350.1, and 365.1 to analyze water quality

PAGE 81

81 samples for nitrate/ nitrite-nitrogen (NOX-N), total ammonia nitrogen (NH3-N), and total phosphorus (TP) respectively. NOX-N, NH3-N, and TP concentrations from 1992 to 2006 for each of the six monitoring sites used in this study were obtained from DERM. Each of the six monitoring sites had associated flow sites and daily flow data were obtained from SFWMD to estimate nutrient loads. SFWMD uses wireless co mmunications systems to remotely monitor and record flow data through existing structures. Wa ter quality and flow data flagged for violating quality control criteria were excluded from analysis. Trend Analysis USGS Estimate Trend (ESTREND; Schertz et al 1991) is an applica tion extension for SPlus 8.0 (Insightful Corporation 2007) that includ es both parametric (Tobit regressions) and nonparametric (uncensored/censored seasonal Kendall) methods to determine trends in constituent water quality data. Tobit regression uses a Maximum Likelihood Estimation (MLE) method to determine trends for parameters that contain grea ter than 5% censored data with multiple MDLs. The seasonal Kendall methods are suitable for pa rameters with less than 5% censored data (uncensored seasonal Kendall) and greater than 5% censored data at a single detection limit (censored seasonal Kendall). ESTREND was used to analyze nutrient concentration trends at the water quality monitoring sites. For the peri od 1992 to 2006, trend analysis was performed separately at each site on monthly NOX-N, NH3-N, and TP concentrations. To a ccount for seasonality in trend analysis, wet (May 27th to November 7th) a nd dry (November 8th to May 26th) seasons were defined in ESTREND based on an evaluation of historical rainfall in south Florida (Qian et al. 2007). Nutrient data from Miami-Dade DERM were not flow adjusted because SFWMD regulates flow within the six study canals, th ereby deviating from natural discharge patterns characteristic of unaltered sy stems (Hirsch et al. 1991).

PAGE 82

82 Only Tobit regression was used for trend analysis on NOX-N, NH3-N, and TP concentrations because more than 5% of the data at each study site were censored at multiple detection limits. In Tobit regression, censored da ta are not assigned values; instead, values for censored data are predicted based on the distri bution of known values (S chertz et al. 1991). Constituents were log transformed prior to the Tobit procedure and seasonal terms (sine and cosine of yearly cycle) were in cluded for constituents. Studentized residuals below -3 or above 3 for NOX-N, NH3-N, and TP concentrations at each site were considered outliers and excluded from trend analysis. Finally, the Tobit proced ure in ESTREND calculated trend slopes that indicated the rate of change of each nutrient over time and results were evaluated using a significance level of p < 0.1. Nutrient Loads The U.S. Geological Survey (USGS) devel oped the Load Estimator (LOADEST) model that has been specifically designed to estimate loads in streams and rivers (Runkel et al. 2004). The model is a publicly available FORTRAN progr am that uses linear regressions to estimate daily, monthly, seasonal, or annual loads and users have the ability to customize the model to fit particular objectives. LOADEST primarily us es the Adjusted Maximum Likelihood Estimation (AMLE) method to estimate loads but if the calibration dataset is un censored, the MLE method is used. However, both methods assume that model residuals are normally distributed. LOADEST also includes the Leas t Absolute Deviation (LAD) me thod to estimate loads when the normality assumption is violated. The flexib ility of LOADEST and its ability to estimate loads when assumptions are violated enhances the utility of this program for water quality analyses. LOADEST was used to estimate annual nutri ent loads (1992 to 2006) from the six sites using monthly NOX-N, NH3-N, and TP concentrations daily flow data. Nutrient and flow data for

PAGE 83

83 each of the six sites were utilized as LOAD EST calibration datasets to develop separate regression models estimating NOX-N, NH3-N, and TP loads. LOADEST gives users the option of selecting a regression model fo r load estimations or allowing the program to select the best model from a set of pred efined models. The latter option was chosen in this study to identify the best available regression models fo r each constituent at each of the six sites. To select the best models, LOADEST calculated model coefficients fo r several predefined regression models using each calibration dataset and models with the lowest Akaike Information Criterion values (Judge et al. 1988) were selected for load estimations. NOX-N, NH3-N, dissolved inorganic nitrogen (NOX-N plus NH3-N), and TP load estimates from LOADEST were evaluated by calculating Nash-Sutcliffe e fficiency (NSE) coefficients (Nash and Sutcliffe 1970), comparing daily load es timates to actual loads at monitoring sites on days where both nutrient concentrat ions and flow data were avai lable. NSE was calculated using Equation 3-1: NSE 1(3-1) where n is the number of values, Y obs and Y sim are measured and simulated values, respectively, and Y mean is the mean of measured values. NSE coefficients range from to 1 (1 being a perfect model fit) and coefficients above zero indicate acceptable model performance. Negative values indicate that simulated values from a m odel are less efficient than using the mean of the measured values, representing unacceptable model performance (Mor iasi et al. 2007). Pollutant Index The PED index uses empower density (emergy per unit volume per unit time) values to provide a relative indicator of ecological stress because thermodynamic principles guide

PAGE 84

84 ecosystem organization. According to the 4th law of thermodynamics, the maximum power principle (Lotka 1922), systems or ganize to increase efficiency a nd hierarchical relationships involve energy transfers and positive feedback loops (Odum 1996). Emergy signatures for ecosystems are therefore associated with the level of structural organization and can be used as a reference to evaluate ecologi cal integrity (Campb ell 2000). For example, Bastianoni (1998) defined pollution in terms of emergy flows and system organization, noting that pollutants cause an increase in emergy flows and a corresponding d ecrease in ecosystem efficiency. There is a temporal dimension to pollution because eco systems are dynamic networks and increased emergy flows, due to factors such as additional nut rient inputs, inevitably lead to reorganization efforts that attempt to maximize newly availa ble resources. Depending on the range of species that can exploit these new resources, overall st ructural complexity within ecosystems could possibly decline. Coral reefs and seagrass comm unities in Biscayne Bay, for example, are susceptible to increased nutrient co ncentrations (Brand et al. 2002). The PED index is calculated using the flux of pollutants and the background productivity of the reference environment. Campbell et al. (2 000) used nine energy sources to derive emergy signatures for three different types of estuaries. Literature valu es for these nine energy sources that could be applied to Bisca yne Bay were used to calculate the reference empower density (Table 3-1). In emergy analysis all contributing ener gy sources are converte d and expressed in units of solar emergy (the available solar en ergy used during energy transformations; solar emjoules [sej]) to assess the amount of emergy associated with each source [sej yr-1]. However, all emergy values were not added to determin e the background empower density of Biscayne Bay. Adding all the emergy values would lead to an overestimation of background conditions because individual inputs may not be independe nt (e.g., energy inputs from wind and tides)

PAGE 85

85 (Odum 1996; Campbell et al. 2000). To address this potential problem, the following procedure was implemented: (1) all emergy values were calculated; (2) the highest emergy value was identified; (3) energy inputs were evaluated to determine if adding particular emergy values would avoid the problem of overestimation; and (4) the background empower density [sej m-3 yr1] was determined using the highest emergy value by itself or with another emergy value that was derived from a sufficiently different energy sour ce. The PED was calculated using Equation 3-2: PED = 10 log10 (emPDTotal/emPDRef) (3-2) where PED [unit less] is the Pollutant Empower Density index for Biscayne Bay, emPDTotal is the total empower density [sej m-3 yr-1] and emPDRef is the background empower density [sej m-3 yr-1]. The total empower density (emPDTotal) is calculated as follows: emPDTotal = emPDRef + emPDi (3-3) where emPDi = empower density of pollutant i [sej m-3 yr-1]. The annual empower density of pollutant i is calculated using the specific emergy of the appropriate nutrient and its annual flow weighted concentration. For example, the empower density of NOX-N was calculated using Equation 3-4: emPD of NOX-N = (specific emergy of N) (annual NOX-N load/annual discharge volume) (3-4) The specific emergy for nitrogen (7.02E+12 sej g-1) was used to calculate PED values for both forms of nitrogen (NOX-N and NH3-N); the specific emergy for phosphorus (1.08E+11 sej g-1) was used for TP. Annual PED values (1992 to 2006) at each of the six sites were calculated separately for NOX-N, NH3-N, and TP.

PAGE 86

86 Results Trend Analysis Table 3-2 summarizes ESTREND results for NOX-N, NH3-N, and TP at the six study sites using the Tobit procedure and a significance level of p < 0.1. During the study period (1992 to 2006), five sites had signifi cant annual trends for NOX-N. SK02 and AR03 had significant negative (i.e., downward) trends while LR06, BL03, and MW04 had significant positive (i.e., upward) trends. BS04 had no significant trend for NOX-N. For NH3-N, there were four sites with significant trends: BS04, LR06, and AR03 (all ne gative) and BL03 (positive). SK02 and MW04 exhibited no significant trends for NH3-N. Only BS04 and MW04 had si gnificant positive trends for TP during the study period. SK02, LR06, BL03, a nd AR03 all had no significant trends for TP. Nutrient Loads Annual discharges associated with the six sites are shown in Figure 3-2. Average NSE coefficients for nutrient loads at the six sites, when comparing measured loads to LOADEST simulated results, were 0.59 (NOX-N), 0.61 (NH3-N), 0.72 (NOX-N plus NH3-N), and 0.70 (TP) (Table 3-3; Figure 33). Median annual NOX-N loads from 1992 to 2006 indicated MW04 (157,248 kg yr-1) had substantially larger values than the other five sites: BL03 (81,888 kg yr-1), SK02 (79,658 kg yr-1), LR06 (48,907 kg yr-1), BS04 (35,056 kg yr-1), and AR03 (2,434 kg yr-1) (Figure 3-4). Median annual NH3-N loads were greatest for LR06 (105,022 kg yr-1), with SK02 (61,247 kg yr-1), BL03 (46,644 kg yr-1), BS04 (26,204 kg yr-1), AR03 (3,943 kg yr-1), and MW04 (1,168 kg yr-1) having lower loads (Figure 3-4). Combined median annual NOX-N plus NH3-N loads for AR03 (6, 376 kg yr-1) were substantially less than MW04 (158,127 kg yr-1), LR06 (145,854 kg yr-1), SK02 (136,227 kg yr-1), BL03 (136,026 kg yr-1), and BS04 (57,035 kg yr-1) (Figure 3-4). For TP, median annual loads for LR06 (4,738 kg yr-1), SK02 (2,816 kg yr-1), BS04

PAGE 87

87 (2,288 kg yr-1), BL03 (1,196 kg yr-1), and AR03 (211 kg yr-1) all had greater estimated loads than MW04 (155 kg yr-1) (Figure 3-4). Pollutant Index Table 3-1 lists the energy sources and relevant literature values used to calculate the background empower density of Bi scayne Bay (3.17E+11 sej m-3 yr-1). Organic matter (measured as total organic carbon) in the can als had the highest emergy value (5.16E+20 sej yr-1) and was combined with the chemical potential energy in groundwater (1 .83E+19 sej yr-1) to determine the background empower density. Chemi cal inputs from groundwater to Biscayne Bay are temporally different from organic matter i nputs from the canals and therefore emergy values for both of these sources were not expected to lead to an overestimation of background conditions. PED index values are shown in Figure 3-5. For NOX-N, MW04 (average: 18.47) consistently had the highest PED from 1992 to 2006. PED values began to decline at MW04 in 2001 but remained elevated compared to the othe r four sites. The reference site, AR03 (4.61), had the lowest PED for NOX-N during the study period. LR06 (9.96), BL03 (9.87), BS04 (8.91), and SK02 (8.62) all had similar PED values for NOX-N. For NH3-N, LR06 (12.96) had the greatest PED but declined between 2002 and 2006. MW04 (2.34) had lower NH3-N PED values than AR03 (5.49) while BS04 (8.13), BL03 (8.05), and SK02 (7.87) all had similar values. PED values for TP were much lower than those for nitrogen (both NOX-N and NH3-N). LR06 (0.06) had the highest PED values for TP but declin ed from 1992 to 1997 and then stayed relatively constant from 1998 to 2006. BS04 (0.03) had the s econd highest PED for TP and increased from 1999 to 2006. SK02 (0.02), BL03 (0.01), MW04 (0.01) and AR03 (0.01) all had similar PED values for TP (Figure 3-5).

PAGE 88

88 Discussion Trends During the period of analysis (1992 to 2006), the majority of NOX-N, NH3-N, and TP concentrations decreased or exhi bited no change at the six water quality monitoring sites, with only six instances of significantly (p < 0.1) incr easing trends (Table 3-2) Temporal changes in nutrient concentrations are important but sp ecific water quality standards determine the significance of increasing or decreasing trends The U.S. Environmental Protection Agency (USEPA) proposed nutrient criteria guidelines fo r nationwide ecoregions (USEPA 2000) and has allowed individual states to develop and implement specific criteria that reflect local conditions. The Florida Department of Environmental Protec tion (FDEP) is currentl y developing statewide nutrient criteria for surface waters, including str eams/canals in ecoregion XIII (southern Florida coastal plain) where the Biscayne Bay waters hed is located (FDEP 2009) Watershed nutrient criteria following USEPA guidelines is therefor e not yet available but Abbott (2005) described water quality targets for NOX-N (0.05 mg L-1 in Biscayne National Park) and NH3-N (0.05 mg L1 throughout the bay; 0.01 mg L-1 within Biscayne National Park). The Miami-Dade County water quality standard for NH3-N in surface waters is 0.5 mg L-1 and the USEPA nutrient criteria recommendations for ecoregion XII (s outhern coastal plain; directly adjacent to ecoregion XIII) for total nitrogen concen trations is 0.9 mg L-1. The USEPA recommendation for TP in ecoregion XII is 0.04 mg L-1. Surface water quality standards in Miami-Dade County and USEPA ecoregion XII nutrient criteria were used to evalua te concentration trends at the six monitoring sites. Trend analysis results were somewhat diffe rent from other Biscayne Bay water quality reports identifying possible areas of concern in the watershed. Factors that may have contributed to different results include variability in trend analysis methods (e.g., not accounting for

PAGE 89

89 changing minimum detection limits over time) and contrasting study periods. For example, Abbott (2005) evaluated long-term (1991 to 2003) water quality data at 13 sites throughout the watershed, including five of the sites analyzed in this study, and conc luded that nitrogen concentrations were generally increasing and TP concentrations were declining. Trend analysis results at site MW04, located in the C-103 sub-basin, corresponded with data from Abbott (2005) for both NOX-N (increasing trend) and NH3-N (no trend) but TP con centrations had a significant (p < 0.1) increasing trend from 1992 to 2006 (Table 3-2). Factors that may be contributing to gradually increasing TP concentr ations include a wastewater treatment plant located upstream from MW04 and increasing (26.8% to 31.4%) low a nd medium density single family residential land use classes in the C-103 sub-basin be tween 1995 and 2004 (SFWMD 1995; SFWMD 2004). Median TP concentrations at MW04 and AR03 (re ference site) however, were the lowest for all sites and far below the recommended TP criteri a in ecoregion XII (Tab le 3-4). Additional phosphorus inputs in south Bis cayne Bay could produce an eco system shift from benthic autotrophs (i.e., seagrasses and macroalgae) to phytoplankton, as has already occurred in the north (Brand et al. 2002), but low TP concentrat ions at MW04 suggest that an increasing trend will not likely contribute to water qua lity problems in the immediate future. Site BL03, in the C-1 sub-basin, had a signi ficant (p < 0.1) in creasing trend for NOX-N and was the only site with a significan t (p < 0.1) increasing trend for NH3-N concentrations. NOX-N concentrations at BL03 were among the highe st in the study (Table 3-4) and although NH3-N concentrations did not exceed county standards (0.5 mg L-1), multiple factors influence nitrogen concentrations at BL03, including fertilizer applications in upstream agricultural areas. In addition, a cause for concern at BL03 is the influence of the nearby South Dade Landfill. Environmental issues associated with landfills include leachate formation and its potential to

PAGE 90

90 affect both surface and groundwater (El-Fadel et al. 1995; Kjeldsen et al. 2002). Alleman (1990) and Meeder and Boyer (2001) have found elevated ammonia concentrations in canals adjacent to the landfill, including C-1, after evaluating water quality in the area. In addition, McKenzie (1983) monitored wells within th e landfill and determined ammonia concentrations were greatest beneath recent waste deposits during the dry season, indicating active organic matter decomposition. The South Dade Wastewater Trea tment Plant (SDWWTP) is another potential source of ammonia but it uses deep well injectio n to dispose effluents underground and monitoring data suggests that it is not affecting ammonia c oncentrations in surface or groundwater (Alleman 1990). The SDWWTP could be a concern, however, as ammonia was found in shallow monitoring wells after only 11 years of the first effluent injection, contradicting estimates suggesting a 343-year period for upward migration (McNei ll 2000; Brand et al. 2002). Loads Similar to Lietz (1999) and Caccia and Boye r (2007), annual nutrient loads at the six monitoring sites during the st udy period revealed higher NOX-N loads in the southern section of the watershed and higher NH3-N and TP loads in the northern and central areas (Figure 3-4). Downstream water flow in canals transport polluta nts from the watershed directly to the north, central, and south sections of Biscayne Bay with land use patterns influencing pollutant characteristics (Caccia and Boyer 2005). Nutrient loads from the watershed therefore reflect widespread agricultural production in the south as well as urban and residential land use classes in the north. As described in the previous section, the South Dade Landfill and SDWWTP are located in the central watershed and these two fac ilities are potential nutrie nt sources influencing water quality at site BL03. NOX-N loads from site MW04, located on the C103 canal, were the highest for all sites (Figure 3-4), which is similar to the results from other water quality stud ies throughout the entire

PAGE 91

91 watershed (Scheidt and Flora 1983; Cheesman 1989; Alleman et al. 1995). Agricultural production is a dominant influence in the C-103 s ub-basin as row and tree crops accounted for approximately 45% of total land use in 2004 (SFWMD 2004). Local environmental characteristics and land use practices therefore combine to influence nutrient loads from MW04 and the C-103 canal. For example, Orth (1976) analyzed groundwater samples wells within the C-103 sub-basin and found NOX-N concentrations ranging from 3 mg L-1 to 10 mg L-1, which were likely influenced by fertilization rates, inappropriate irrigation, a nd leaching caused by high permeability soils with low water-holding capacity (Li 2000; Li and Zhang 2002). Furthermore, high water tables in the watershed enable ni trogen-enriched groundwater to degrade water quality in the canals and eventua lly Biscayne Bay (Langevin 2000). Although MW04 had the highest loads among all sites, NH3-N loads at this site were the lowest in the study (Figure 3-4). Scheid t and Flora (1983) also reported low NH3-N concentrations in the C-103 canal and Cacci a and Boyer (2007) determined that NOX-N contributed 95% of dissol ved inorganic nitrogen (NOX-N plus NH3-N) inputs from canals in the southern region of the watershed to the bay. MW04 had the lowest median NH3-N concentration during the study period (Table 3-4) among the six s ites analyzed and this suggests that compared to NOX-N, NH3-N concentrations have been histori cally low in the C-103 sub-basin. Site LR06, located on the C-7 canal, had the highest median concentrations and annual loads for both NH3-N and TP (Figure 3-5). The heavily urbanized North Bay, which receives canal discharges from C-9, C-8, and C-7 canals, has historically struggled with sewage discharges, high nutrient loads, turbidity, and heavy metals (Alleman et al. 1995). Land use data from 2004 revealed medium intensity single family residential (48.9%) and low intensity commercial (11.7%) areas dominated the C7 basin (SFWMD 2004). Between 1920 and 1955,

PAGE 92

92 raw wastewater was released in to several northern canals, including C-7 (Wanless 1976), which was identified as a conveyor transporting sewa ge-contaminated discharges to Biscayne Bay (McNulty 1970). Raw wastewater is no longer disc harged into the C-7 canal but Alleman et al. (1995) suggested stormwater r unoff and leaks associated with sewage systems were likely factors contributing to ongoing po llutant problems. SK02 (C-9 can al) and BS04 (C-8 canal) also had elevated NH3-N and TP loads compared to the refere nce site, although loads at these sites were lower than LR06. Pollutant Index Annual nutrient loads at the six monitoring si tes fluctuated greatly but corresponding PED index values were less sensitive, exhibiting a da mped response to nutrient fluxes (Figure 3-5). An important factor contributing to this difference is that the PED index uses the logarithmic decibel scale to represent the inte nsity of discharged pollutants rela tive to reference conditions in aquatic systems. The decibel scale has been pr eviously used in the development of a human disturbance index (Reiss et al. 2009) to evaluate the impact of anthropogenic influences. Brown and Vivas (2005) introduced the Landscape Developm ent Intensity index (LDI) to investigate the effect of human activities on adjacent systems but an important limitation of the index was that disturbance intensity was not related to backgr ound conditions. Reiss et al. (2009) revised the LDI index to address this issue, enabling back ground conditions to regu late the effects of disturbances. Similarly, the PED index assesses the ability of pollutants to affect aquatic system productivity when considering baseline system properties. PED index values for the six monitoring sites in the Biscayne Bay wa tershed suggest that canal discharges from two sites (MW04 and LR06) provide a greater prop ortional impact in the bay compared to the other sites (Figure 3-5) Site MW04 consistently produced the highest annual PED values for NOX-N but PED values for NH3-N and TP at MW04 were the lowest

PAGE 93

93 among all sites (including the referen ce site, AR03), which suggests that NOX-N dominates nutrient inputs from MW04. However, PED values for NOX-N at MW04 began a general decline in 2001 and this could be indicative of improved best management practices or land use changes occurring in the drainage s ub-basin, C-103. From 1999 to 2004, row crops declined (20% to 12.2%) in C-103 while tree crops (28.2% to 32.5%) a nd single family residential land use areas (28.1% to 31.4%) increased (SFWMD 2009). Tr ee crops and residential developments correspond to relatively lower NOX-N canal concentrations than row crops and this may have contributed to lower PED values between 2001 and 2006. Site LR06 had the highest PED index values for NH3-N and TP (Figure 3-5). Similar to MW04, general water quality improvement at LR 06 also occurred during the study period. PED values for NH3-N at LR06 eventually declin ed to a point where they were similar to values from SK02, BS04, and BL03; furthermore, PED values for TP at LR06 declined substantially from 1992 to 1997, before leveling out in subsequent year s. Drainage and sewer improvement projects have helped to lower pollutant discharges from LR06 but additional measures are necessary to control nutrient export to the bay. Brand et al. (2002) conduc ted nutrient bioassays throughout the bay and found that at the mouths of some northern canals, including C-7 where LR06 is located, the ecosystem is actually nitrogen limited; the bay is typically phosphorus limited because calcium carbonate chemically scavenge s phosphorus from water due to groundwater inputs moving through limestone and calcareous sedi ments in shallow surface water. Localized saturation of calcium carbonate in northern Bisc ayne Bay therefore refl ects historical and substantial phosphorous inputs from urbanized sub-basins. Similar to annual nutrient loads, PED values for NH3-N also provide evidence of another locali zed issue in the watershed as the mixed land use site, BL03, consistently had similar values to the more urban sub-basins, SK02 and

PAGE 94

94 BS04. The South Dade Landfill and possibly the SDWWTP are important factors influencing the water quality of adjacent canals. Management Implications Trend analysis, load estimation, and the PED index can be used together to provided a more holistic interpretation of water quality (Figure 3-6), which is necessary for optimizing resources to meet watershed management goals. An extensive period of r ecord (e.g., 10 years) is required to detect statistically significant changes in water quality and thus trend analysis can be used to evaluate existing management strategies to determine if consti tuent trends are improving over time or getting worse. In addition, results fro m trend analysis can help to forecast future conditions when considering hist orical patterns. Both loads and the PED index are useful indicators to determine relative pollutant cont ributions from multiple sites but the PED index provides an ecological assessment of discharges relative to receiving systems. Loads and the PED index can be calculated annually but comp arisons over time would provide additional insight into pollutant dynamics. Trend analysis results suggest that water quality is generally improving throughout the Biscayne Bay watershed and polluta nt loads and the PED index indica te that canal discharges are coupled with land use activities in adjacent drai nage areas. However, development patterns are not the only concern for watershed manage rs because proposed projects under the Comprehensive Everglades Restoration Plan will a ffect the quantity and quality of freshwater flows to the bay. Projects includ e rerouting canal discharges to coastal wetlands and recycling wastewater to supplement freshwat er inflows to the bay; however, proposed criteria for recycled water that would provide protection for the bay ma y not be easily attain able (Browder et al. 2005).

PAGE 95

95 In the Biscayne Bay watershed, therefore, there is an urgent need for assessment tools that can be used to guide management initiatives re garding water quality discharges to the bay. The PED index can be used to evaluate potential ecol ogical impacts associated with discharges from specific water quality monitoring sites. Creating a separate PED index fo r individual pollutants reveals watershed distribution a nd also provides a relative indication of energy-based ecological stress associated with pollutants. The PED inde x can therefore provide a link between watershed processes and subsequent pollutant discharges, whic h can be used to identify watershed locations that may disproportionately affect the ecological health of the bay. The PED index may also be useful in a variety of watersheds because data needed to calculate emergy signature s are available through published re ports pertaining to particular aquatic ecosystems. Emergy values have been calculated for different types of estuaries (Campbell 2000), subtropical springs (Collins and Odum 2000), subtropical lakes (Brown and Bardi 2001) and coral reefs (McC lanahan 1990). The PED index is not limited to a particular type of system or region and can be used to assess the proportional impact of pollutant discharges from multiple sources. Conclusion Nutrient water quality data (1992 to 2006) fr om six water quality m onitoring sites in the Biscayne Bay watershed were evaluated using multip le analytical methods (trends, loads, and the PED index) and although areas of concern were identified, water quality has generally improved during the study period. Trend analysis results indicate that nutrient (NOX-N, NH3-N, and TP) concentrations declined or exhib ited no change at most of the si x water quality monitoring sites. NOX-N and NH3-N concentrations increas ed during the study period at site BL03 and upstream agricultural influences, along with a nearby landf ill and possibly a wastewater treatment plant, are factors contributing to nitr ogen concentrations at BL03. Fo r nutrient loads, the monitoring

PAGE 96

96 site in the southern, agricultural section of the wate rshed (MW04) consiste ntly produced greater NOX-N loads than monitoring sites in the urbani zed northern section of the watershed (SK02, BS04, and LR06), where NH3-N and TP loads were greatest. Compared to nutrient loads, PED index values for the monitoring sites were more consistent during the study period and pollutant discharges from two sites (MW04 and LR06) had th e greatest potential for impact in the bay. The PED index is a new analytic tool to as sess the intensity of discharged pollutants relative to the background producti vity of aquatic systems. All ecosystems rely on fundamental energy relationships and the PED index can be used to identify pollutant discharges that could disrupt energy flows and impair aquatic health. The index can be applied in a broad range of aquatic systems because the volume of incomi ng pollutants can be quantified and background productivity can be calculated. Results from studies investigating additional pollutants beyond nutrients in different types of aquatic systems w ill contribute to the continued development of the PED index.

PAGE 97

97 Table 3-1. Energy sources and conversion factor s used to calculate background productivity (emergy signature) for Biscayne Bay. Energy Source1 Energy (J yr-1)2 Transformity (sej J-1) Emergy (sej yr-1) Sunlight 4.39E+18 1 4.39E+18 Wind 1.16E+17 1496 1.74E+20 Rain, Chemical 5.29E+15 18199 9.62E+19 Tide 8.96E+14 24259 2.17E+19 Estuary Waves 7.69E+15 30550 2.35E+20 Geologic Uplift 2.50E+12 34377 8.59E+16 Ground Water, Chemical 4.47E+14 41000 1.83E+19 River (Canals), Chemical 6.17E+15 48459 2.99E+20 River (Canals), Organic matter3 1.73E+14 2.98E+06 5.16E+20 1 Calculations for each energy source taken fr om Odum (1996) and Cambell (2000). Biscayne Bay area and average depth from Dame et al. ( 2000). Average of insolation from Fend et al. (2003). Wind gradient and diffusion coefficient for Tampa Bay (Odum 1996). Rainfall estimates and wave data from Buchanan and Klein (1976). Tide data from Lee a nd Rooth (1976). Geologic uplift (heat flow per area) from Odum (1996). Groundwater input s from Radell and Katz (1991) and Langevin (2001). Chemical and organic matter input from Alleman et al. (1995) and Lietz (2000). 2 J yr-1 = joules per year; sej J-1 = solar emjoules per joule; sej yr-1 = solar emjoules per year. 3Organic matter in canals refers to total organic carbon concentration.

PAGE 98

98 Table 3-2. Trend analysis results for nutrient con centrations at six water quality monitoring sites in the Biscayne Bay watershed. Constituent Site Observations Trend (%) p value Significance (p < 0.1) NOX-N SK02 142 -5.47 0.00 decreasing BS04 88 -0.63 0.31 none LR06 149 1.97 0.01 increasing BL03 149 4.48 0.00 increasing MW04 138 2.07 0.00 increasing AR03 150 -3.01 0.08 decreasing NH3-N SK02 154 -1.11 0.64 none BS04 149 -2.80 0.09 decreasing LR06 151 -4.75 0.00 decreasing BL03 153 13.20 0.00 increasing MW04 153 0.55 0.73 none AR03 145 -7.72 0.00 decreasing TP SK02 141 1.01 0.27 none BS04 150 2.69 0.00 increasing LR06 153 1.03 0.15 none BL03 155 -0.47 0.74 none MW04 147 4.55 0.00 increasing AR03 151 1.42 0.36 none

PAGE 99

99 Table 3-3. Average Nash-Sutcliffe Efficiency (NSE ) coefficients for six water quality monitoring sites in the Biscayne Bay watershed after comparing LOADEST simulated loads to measured loads (1992 to 2006). Site NOX-N NH3-N NOX-N plus NH3-N TP SK02 0.41 0.76 0.72 0.38 BS04 0.20 0.66 0.46 0.87 LR06 0.73 0.41 0.71 0.92 BL03 0.49 0.38 0.72 0.50 MW04 0.83 0.54 0.84 0.65 AR03 0.88 0.88 0.88 0.88 Average 0.59 0.61 0.72 0.70

PAGE 100

100 Table 3-4. Summary statistics for nutrient concen trations (1992 to 2006) and flow data at six water quality monitoring sites in the Biscayne Bay watershed. Constituent Site Min1 Mean Median Max2 Mean Annual (mg L-1) (mg L-1) (mg L-1) (mg L-1) Discharge (m3 s-1) NOX-N SK02 0.010 0.224 0.210 0.680 8.88 BS04 0.010 0.222 0.200 1.750 3.40 LR06 0.020 0.271 0.280 0.530 3.87 BL03 0.010 0.288 0.245 1.010 6.00 MW04 0.010 2.200 2.250 4.640 1.41 AR03 0.010 0.049 0.040 0.230 1.14 NH3-N SK02 0.008 0.124 0.070 0.540 BS04 0.008 0.111 0.070 0.410 LR06 0.010 0.390 0.370 0.930 BL03 0.008 0.086 0.030 0.400 MW04 0.008 0.025 0.020 0.100 AR03 0.008 0.051 0.040 0.240 TP SK02 0.001 0.010 0.008 0.170 BS04 0.003 0.018 0.015 0.170 LR06 0.004 0.024 0.022 0.170 BL03 0.001 0.007 0.005 0.170 MW04 0.001 0.006 0.004 0.048 AR03 0.001 0.006 0.004 0.170 1Water quality targets: NOX-N (0.05 mg L-1 in Biscayne National Park); NH3-N (0.01 mg L-1 within Biscayne National Park; 0.05 mg L-1 throughout the bay; 0.5 mg L-1 for surface waters in Miami-Dade County); total nitrogen (0.9 mg L-1 for ecoregion XII, southern coastal plain); TP (0.04 mg L-1 for ecoregion XII). 2Minimum detection limits (MDLs) were used as maximum concentrations for censored data. MDLs for censored TP data duri ng the study period included 0.17 mg L-1.

PAGE 101

101 Figure 3-1. Six water quality monitoring sites in the Biscayne Bay watershed.

PAGE 102

102 Figure 3-2. Annual discharge (1992 to 2006) from six water quality monitoring sites in the Biscayne Bay watershed. 0 2 4 6 8 10 12 14 1992 2006Annual Discharge (m3s -1)Year SK02 BS04 LR06 BL03 MW04 AR03

PAGE 103

103A) B) C) D) Figure 3-3. Nash-Sutcliffe Efficien cy (NSE) coefficients for selected water qual ity monitoring sites in the Biscayne Bay waters hed after comparing LOADEST simulated loads to measured loads (1992 to 2006). A) NOX-N loads at LR06. B) NH3-N loads at SK02. C) NOX-N plus NH3-N loads at BL03. D) TP loads at BS04. NSE = 0.73 R = 0.67 0 200 400 600 800 0200400600800Estimated Loads (kg/day)Measured Loads (kg/day) NSE = 0.76 R = 0.75 0 500 1,000 1,500 2,000 05001,0001,5002,000Estimated Loads (kg/day)Measured Loads (kg/day) NSE = 0.72 R = 0.81 0 1000 2000 3000 4000 0100020003000Estimated Loads (kg/day)Measured Loads (kg/day) NSE = 0.87 R = 0.87 0 20 40 60 80 020406080Estimated Loads (kg/day)Measured Loads (kg/day)

PAGE 104

104 A) B) Figure 3-4. Estimated annual nutrient loads (1992 to 2006) at six water quality monitoring sites in the Biscayne Bay watershed. A) NOX-N. B) NH3-N. C) NOX-N plus NH3-N. D) TP. 0 100,000 200,000 300,000 400,000 19922006NOX-N Load (kg/yr)Year SK02 BS04 LR06 BL03 MW04 AR03 0 100,000 200,000 300,000 400,000 1992 2006NH3-N Load (kg/yr)Year SK02 BS04 LR06 BL03 MW04 AR03

PAGE 105

105 C) D) Figure 3-4. Continued. 0 100,000 200,000 300,000 400,000 19922006NOX-N plus NH3-N Load (kg/yr)Year SK02 BS04 LR06 BL03 MW04 AR03 0 2,000 4,000 6,000 8,000 1992 2006TP Load (kg/yr)Year SK02 BS04 LR06 BL03 MW04 AR03

PAGE 106

106A) B) C) Figure 3-5. Pollutant Empower Density (PED) index values (1992 to 2006) at six water qu ality monitoring sites in the Biscayne B ay watershed. A) NOX-N. B) NH3-N. C) TP. 0 5 10 15 20 25 1992 2006NOX-N PED Year SK02 BS04 LR06 BL03 MW04 AR03 0 5 10 15 20 25 1992 2006NH3-N PED Year SK02 BS04 LR06 BL03 MW04 AR03 0.00 0.02 0.04 0.06 0.08 0.10 0.12 19922006TP PEDYear SK02 BS04 LR06 BL03 MW04 AR03

PAGE 107

107 Figure 3-6. Flow chart illustrating the combin ed use of trend analysis, load estimatio n, and the Pollutant Empower Density inde x to evaluate water quality variability in watersheds. Best Management Practices Management Strategies Pollutant Empower Density Index Constituent Loads Trend Analysis Water Quality Variability Ecological Stress from Pollutants Watershed Areas of Concern Concentration Trends Mass of Pollutants Delivered Watershed Pollutant Distribution Water Quality Standards Educational Programs/ New Policies

PAGE 108

108 CHAPTER 4 LAND USE INFLUENCES (1995 TO 2004) A FFECTING NUTRIENT WATER QUALITY IN THE BISCAYNE BAY WATERSHED, FLORIDA Introduction Geographic information system (GIS) software and increasingly available land use data facilitates watershed analysis linking indi cators of landscape c ondition to water quality parameters (e.g., Hunsaker and Levine 1995; Johnson et al. 2001; Kearns et al. 2005). Intensive land use activities affect hydrol ogical, biological, chemical, and geomorphic aspects of aquatic systems (Arnold and Gibbons 1996; Gergel et al. 2002; Alberti 2005). Several indicators can be used to evaluate anthropogenic influences within watersheds including landscape metrics (Forman and Godron 1986; Turner 1990; Li and Wu 2004), disturban ce gradients (McMahon and Cuffney 2000; Brown and Viva s 2005; Wang et al. 2008), and the extent of impervious surfaces (Arnold and Gibbons 1996; Brabec et al. 2002; Roy and Shuster 2009). Landscape metrics quantify spatial confi guration and composition (McGarig al and Marks 1995) and have been used to investigate devel opment patterns in watersheds (K earns et al. 2005; Cifaldi et al. 2004), landscape characteristics influencing sedime nt contamination levels (Paul et al. 2002), and the effect of land use distributio n on water quality (Lee et al. 2009). To evaluate the effects of land use activitie s on adjacent systems, Brown and Vivas (2005) developed the Landscape Development Intensity (L DI) index to quantify disturbance gradients based on nonrenewable energy use and emer gy analysis. Emergy (Odum 1971; Odum 1996) measures the amount of energy that is directly an d/or indirectly associat ed with both natural and anthropogenic products and services. Huma n-dominated landscapes feature greater nonrenewable energy use than natural areas and the LDI index provides a relative indication of the extent of nonrenewable energy use associ ated with landscape disturbances (Odum 1996; Brown and Vivas 2005). Reiss (2005), Mack (2006) and Reiss and Brown (2007) have all used

PAGE 109

109 the LDI index to assess landscape conditions an d Reiss et al. (2009) ut ilized an updated LDI index to evaluate the effect of hum an disturbances on Florida wetlands. Imperviousness is another important environmen tal indicator because it is associated with intensive land use activities and contributes to hydrologic changes that can degrade water quality (Arnold and Gibbons 1996; Brabec et al. 2002; Schiff and Benoit 2007). Watershed imperviousness values can describe either tota l impervious areas (TIA), which include surfaces that may drain to pervious ground, or directly connected impervious ar eas (DCIA) that drain directly to aquatic systems (Brabec et al. 2002). Watershed impervious ness values using TIA therefore include connected (DCIA) and unconne cted surfaces which c ould underestimate the impact of changes in land use activities and subsequent hydrological e ffects (Alley and Veenhuis 1983; Brabec et al. 2009). Roy and Shuster (2009 ) therefore identified DC IA measurements as factors that can be integral to watershed management plans. Numerous studies have used GIS technology and regression analyses to link land use indicators to water quality parameters at multiple spatial extents, including sub-basins (Mehaffey et al. 2005; Migliaccio et al. 2007), riparian zones (Silva and Williams 2001; Schiff and Benoit 2007), and monitoring site proximities (Bol stad and Swank 1997; King et al. 2005). Investigations using regression analyses id entify important explanatory and independent variables (Johnston 1972; Mason 1975; Graham 2003) and conse quently variable selection procedures influence model development and relia bility. The stepwise procedure allows both the inclusion and removal of independent variables during model development to identify the most important variables (Efroymson 1960; Thomas 1978) and has been widely used in regression analyses. Iverson (1988) used stepwi se regressions to select a subset of landscape attributes that had the most influence on historical land use patte rns in Illinois, Harding et al. (1998) identified

PAGE 110

110 land use variables that explained aquatic organism diversity in streams, and Chadwick et al. (2006) investigated urbanization variables a ffecting stream ecological function. Stepwise regressions have also been used to develop multivariate regression models to predict constituent loads of targeted water quality parameters in response to changes in land use variables (e.g, Johnson et al. 2001; Jones et al. 2001; Paul et al. 2002). After se lecting variables, validation procedures (Snee 1977; Moriasi et al. 2007) help to assess the su itability of developed models. Although collecting new data is the preferred method for validating regression models, Snee (1977) notes that splitting the data into an es timation dataset (for model coefficients) and a prediction dataset (to test model accuracy) can simulate the process of collecting new data. Linking land use variables to specific response criteria such as constituent nutrient loads has clear management implications and can lead to improved stra tegies to combat threats to vulnerable water resources. For exam ple, Biscayne Bay is an oligot rophic estuary th at drains the Miami metropolitan area and is sensitive to wate rshed land use activities, including extensive agricultural and urban development. Thus, the goal of this study was to evaluate land use-water quality relationships in the Bi scayne Bay watershed from 1995 to 2004 at eight water quality monitoring sites considering three different spatia l extents: sub-basins, canal buffers, and site buffers. Specific objectives incl uded the following: (1) quantify human disturbance indicators using land use/land cover (LULC) data and GIS spatial analysis; (2) estimate nutrient loads at water quality monitoring sites; (3) develop a nd validate multivariate regression models to identify significant land use vari ables influencing nutrient loads; and (4) determine if disturbance indicators within sub-basins, canal buffers, or site buffers explain more of the variability in nutrient loads at monitoring s ites during the study period.

PAGE 111

111 Methods Study Area Biscayne Bay is a barrier-island subtropical estuary that is located along the southeastern coastline of Florida and includes the federally protected Biscayne National Park. Designated as an Outstanding Florida Water, Bi scayne Bay requires substantial fr eshwater inputs to maintain its natural ecological balance; however, water management opera tions (canals, levees, pump sites, etc.) in south Flor ida have disrupted historical freshwat er flows to the bay. Extensive urban and agricultural development in the watershed (2,500 km2) have thrived on former wetlands as canals have lowered water tables (Parker et al. 1955), reducing waters hed water storage and creating polluted discharges that degrade sensitive estuarine habitats (Browder et al. 2005). The watershed is primarily located in Miami-Dade Co unty, which includes the city of Miami, but the northern section extends into Broward County; the western bo undary of the watershed lies adjacent to the Florida Everglades and the Everglades National Park. Hydrological connectivity may be a significan t factor affecting water quality in the watershed and eight monitoring sites with adequate data were identified and used in this study. Selected sites were located upstream from canal outlets to avoid tidal influences that could potentially affect water quality measurements. In addition, the eight sites we re located in areas of the watershed with contrasting LULCs such as agricultural, urban, and mixed-land uses. Land use and water quality relationships were investig ated at the eight sites considering sub-basins, canal buffers (Tables 4-1; 4-2; 4-3; Figure 4-1), and proximity buffers (Tables 4-4; 4-5; 4-6; Figure 4-2), with the latter two considered at the following distances: 500, 1000, and 1500 meters. Selected sites were associated with a tota l of five canals, with three canals (C-9, C-1, and C-103) having two sites each (one upstream and one downstream) and two other canals (C-8 and

PAGE 112

112 C-7) with only downstream sites. C-8 is a headwater canal and no flow data was associated with the upstream site for C-7 during the study period. The C-9 East (118 km2; referred to as C-9 hereafter), C-8 (71 km2), and C-7 (82 km2) subbasins are located in the norther n section of the watershed, which is primarily characterized by urban land uses. In the central watershed, the C-1 (117 km2) sub-basin includes extensive mixed (urban and agricultural) land uses. The C-103 (113 km2) sub-basin, located in the southern section of the watershed within the South Dade Agricultural Ar ea, is dominated by agricultural land uses such as row and tree crops. Analysis was limited to sub-sections of both C-1 and C-103 sub-basins to correspond with locations of ups tream and downstream monitoring sites. For subbasins with both upstream and downstream site s (C-9, C-1, and C-103), canal buffers covered the distance between the two sites. Canal buffers in sub-basins with only downstream sites (C-8 and C-7) started at the most upstream point of the major canal and continued to the downstream site. Proximity buffers for the eight sites were lo cated upstream of each site to evaluate water quality and LULC influences (Figure 4-2). Land Use Data LULC GIS data layers of the Biscayne Ba y watershed were obta ined from SFWMD for 1995 (scale 1: 40000), 1999 (1: 400 00), and 2004 (1: 12000). SFWMD created all three layers by photo-interpreting aerial photography and digital orthophotographic quarter quadrangles (DOQQs). Each layer used a modified form of the Florida Land Use and Cover Classification System (FLUCCS; FDOT 1999) as SFWMD FLUC CS codes primarily use community level classes to identify vegetation. La nd use classes in this study were aggregated into 18 natural, agricultural, and urban classes to simplify analysis and LULC da ta were converted to raster format for spatial analysis us ing a common scale (190 x 190 meters grid cell size). To ensure LULC data accuracy, DOQQs corresponding to the ti meline of the LULC data were retrieved

PAGE 113

113 from the Land Boundary Information System (LABINS) of the Florida Department of Environmental Protection. DOQQs we re then compared to LULC data and any assigned classes that were inconsistent with actual conditions were corrected to reflect LULC throughout the watershed. Landscape Metrics The most widely used software package to calculate landscape metrics is FRAGSTATS (McGarigal and Marks 1995), whic h generates values for several different categories of metrics that can be useful to understanding changes occu rring in a watershed. Patch Analyst (Elkie et al. 1999), a modified version of FRAGSTATS designed specifically as an ESRI ArcGIS extension tool, provides an integrated user interface that en ables metrics to be calculated for LULC layers at both landscape and class levels Landscape metrics calculate values with all classes included (e.g., mean patch size within a watershed) while class metrics calculate values for specific classes (e.g., mean patch size of row crops). C ontiguous cells from the same land use classes were considered patches in this study. For each of the three LULC layers (i.e., 1995, 1999, and 2004), 17 landscape metrics and 13 class metrics were calculated for the entire Biscayne Bay watershed. Area-weighted metrics (e.g., patch richness density) were preferred to absolute metrics (e.g., patch richness) to compare data from sub-basins and buffers. Metrics were tested for normality using the Shapiro-Wilk W test for normality with a p-value <0.05 (Shapiro and Wilk 1965; Royston 1983). Most metrics deviating from a normal distributi on were either log or square root transformed to improve normality; metrics containing percentage data were arcsin-square root transformed. Pair-wise correlation coefficients were cal culated for transformed landscape and class metrics to eliminate redundancy, with only one metric in a correlated pa ir of metrics selected if Pearson coefficients were greater than 0.90 (Ritters et al. 1995).

PAGE 114

114 Principal component analysis (PCA) and factor analysis (FA) were then used to further reduce the number of landscape and class metric s (e.g., Ritters et al. 1995 ; Cushman et al. 2008). PCA and FA have been used toge ther in land use analysis to i nvestigate landscape structure at different spatial extents (Griffith et al. 2000) and to study development patterns in watersheds (Cifaldi et al. 2004; Kearns et al. 2005). PCA simplifies variable inte rpretation through data reduction (Nichols 1977; Bengraine and Marhaba 2003) and FA identifies significant, underlying factors contributing to overall variance (McDonald 1985). Using a correlation matrix to conduct PCA and FA in S-Plus 8.0 (Insigh tful Corporation 2007), significan t landscape and class metrics were identified for the Biscayne Bay waters hed. Principal components, which are linear combinations of the original metrics, explain all the variance within a dataset and eigenvalues measure the variance explained by each principal component. In a correlation matrix, the mean of eigenvalues is one and principal components w ith above average eigenvalues explain more of the overall variance (Burstyn 2004). The number of principal components with eigenvalues greater than one therefor e determined the number of factors to use in FA. To aid interpretation, FA included a varimax rotation to reveal metrics that had the strongest co rrelations, or loadings, for identified factors across the three different LULC layers (1995, 1999, and 2004). Landscape Development Intensity Index Data required to calculate LD I index values for Biscayne Bay sub-basins included LULC GIS layers, areas for each LULC class, nonrenew able empower intensity (emergy per time per area) values for LULC classes, and the renewable empower inte nsity of the background area. The first step in the LDI calculation process was to sum the areas of each LULC class and express these values as a percent of the tota l landscape area. LULC percentages were then multiplied by their respective nonrenewable empower intensity values for Florida. Equation 4-1 illustrates the revised LDI method (Reiss 2009):

PAGE 115

115 LDI = 10 log10 (emPITotal /emPIRef) (4-1) where LDI [unit less] is the Landscape Devel opment Intensity index for sub-basins, emPITotal [sej ha-1 yr-1] is the total empower intensity (sum of renewable background empower intensity and nonrenewable empower intensity of land uses), and emPIRef is the renewable empower intensity of the background environment (Florida = 1.97 E15 sej ha-1 yr-1). The total empower intensity (emPITotal) was calculated as follows: emPITotal = emPIRef + (%LUi emPIi) (4-2) where %LUi is the percent of the total area in LULC class i and emPIi [sej ha-1 yr-1] is the nonrenewable empower inte nsity for LULC class i For each of the three LULC layers, LDI values were calculated for each sub-basin, canal buffer, and monitoring site buffer. Imperviousness Percent imperviousness is a key environmen tal indicator and multiple studies have estimated percent imperviousness associated w ith different land use classes (e.g., Stankowski 1972; Griffin 1980; Alley and Veenhuis 1983) Miami-Dade County Department of Environmental Resources Management (DERM) developed total impervious area (TIA) and directly connected impervious area (DCIA) reference values fo r various land uses classes to evaluate pollutant loading es timates under alternat e scenarios (DERM 2004) Reference values were calculated using aerial maps and measur ing impervious areas w ithin typical land use classes throughout the county. DERM DCIA values for defined land use classes were used to estimate percent imperviousness for 1995, 1999 and 2004 in each of the five sub-basins, canal buffers, and monitoring site buffers. Water Quality Data Monthly nitrate/nitrite-nitrogen (NOX-N), total ammonia nitrogen (NH3-N), and total phosphorus (TP) concentrations from 1992 to 2006 at the eight water quality monitoring sites

PAGE 116

116 (Table 4-7) were obtained from Miami-Dade County Department of Environmental Resources Management (DERM). Grab samples from site s throughout Biscayne Bay and in watershed canals were analyzed at DERM us ing EPA methods 353.2, 350.1, and 365.1 for NOX-N, NH3-N, and TP, respectively. To estimate nutrient loads, daily flow data (1992 to 2006) from flow sites associated with each of the monitoring sites were obtained from SFWMD (Table 4-7). SFWMD uses wireless communications systems to remotely monitor and record flow data through existing structures. Water quality and flow data flagged for violating quality control criteria were excluded from analysis. Nutrient Loads The U.S. Geological Survey (USGS) devel oped the Load Estimator (LOADEST) model that estimates constituent loads in streams a nd rivers (Runkel et al. 2004). The model is a publicly available FORTRAN program that uses linear regressions to estimate daily, monthly, seasonal, or annual loads and users have the ability to customize the model for specific conditions. The Adjusted Maxi mum Likelihood Estimation (AMLE) method is used if the calibration dataset is censored and the Maximum Likelihood Estima tion (MLE) method is used if the calibration dataset is uncenso red. However, both methods assume that model residuals are normally distributed. The Least Absolute Deviati on method is used to estimate loads when the normality assumption is violated. LOADEST was used to estimate annual nutri ent loads from selected water quality monitoring sites using monthly NOX-N, NH3-N, and TP concentrations from DERM and daily flow data from SFWMD. Loads were estimated fo r the period 1992 to 2006 at eight sites that had available water quality and flow data. To isol ate nutrient loads from specific sub-basins or canals, loads at downstream sites were adjusted to remove nutrient i nputs from upstream sources. Nutrient load estimates from LOADEST were ev aluated by calculating Nash -Sutcliffe efficiency

PAGE 117

117 (NSE) coefficients (Nash and Sutcliffe 1970), comp aring daily load estimates to actual loads at monitoring sites on days where bot h nutrient concentrations and flow data were available. NSE was calculated using the following formula: NSE 1(4-3) where n is the number of values, Y obs and Y sim are measured and simulated values, respectively, and Y mean is the mean of measured values. NSE coefficients range from to 1 (1 being a perfect model fit) and coefficients above zero indicate acceptable model performance. Negative values indicate that simulated values from a m odel are less efficient than using the mean of the measured values, representing unacceptable model performance (Mor iasi et al. 2007). Stepwise Regressions Combined land use variables from 1995 and 1999 were used with LOADEST nutrient loads in stepwise regressions to calibrate models predicting (1) NOX-N, (2) total inorganic nitrogen (NOX-N plus NH3-N), and (3) TP loads. Stepwi se regressions were performed separately for LULC data from sub-basins, canal buffers (500, 1000, and 1500 m), and site buffers (500, 1000, and 1500 m). Three-year average a nnual nutrient load s at each site were used in stepwise regressions to correspond to LU LC data (i.e., 1994 to 1996 loads for 1995 and 1998 to 2000 loads for 1999). Land use variables includ ed landscape metrics, class metrics, LDI values, and DCIA percentages. After conducting PCA and FA on me trics for the entire Biscayne Bay watershed, the metric with the highest load ing for each identified landscape and class-level factor was chosen to represent that factor in stepwise regr essions. Variables for stepwise regression models included the same set of landscape-level metrics, albeit with different land use data corresponding to sub-basins, canal buffers, and site buffers. Class-level metrics included in

PAGE 118

118 stepwise regressions, however, were based on th e area and distribution of each class relative to the total area of sub-basins, canal buffers, a nd site buffers. Stepwise regressions included combined land use data for 1995 and 1999 and thus each class in the sub-basin and canal buffer datasets had 10 values for each variable one valu e for each of the five sub-basins or canals in each year; for the site buffers, each class had 16 values (because of the eight sites) for each variable. To maximize the predictive power of regre ssion equations, class metrics included in stepwise regressions were limited to classes that ha d at least four values greater than 10% of the total areas for sub-basins and canal buffers. Simila rly, class-level criteria for site buffers limited metrics to classes with at least six values gr eater than 10%. Therefore class metrics were only included in stepwise regressions for classes wi th approximately 40% or more of their areal percentages exceeding 10% of the total areas for sub-basins, canal buffers, and site buffers. Two types of stepwise regressions were used to determine which models best predicted nutrient loads considering different spatial extents: (1) forward st epwise regressions only and (2) a combination of both forward and backward st epwise regressions (Tho mas 1978). All variables were included in the initial model for forward stepwise regressions before the final model was determined for the sub-basins, canal buffers, and site buffers. For stepwise regressions in both directions (forward and backward), pair-wise correlation coefficients were calculated for all remaining land use variables and if any two variables were highly correlated (greater than 0.90), only one variable from that pair wa s included in the initial model. Moriasi et al. (2007) reviewed hydrological m odel evaluation techniques and determined that NSE coefficients of at le ast 0.5 represented satisfactory m odel performance, with values greater than 0.75 indicating very good performance. In addition to NSE coefficients, Moriasi et al. (2007) also recommended two qua ntitative statistics to evalua te model performance, percent

PAGE 119

119 bias (PBIAS) and the ratio of the root mean squa re error to the standard deviation of measured data (RSR). PBIAS assesses the average tendency of simulated da ta to exhibit underestimation (positive PBIAS values) or overestimation (neg ative PBIAS values) bi as (Gupta et al. 1999): PBIAS = Yiobs Yisim *100n i=1 Yiobs n i=1 (4-4) where PBIAS is the deviation of simulated values (Y sim) relative to measured values ( Y obs). RSR provides a standardized erro r index of model performan ce (Moriasi et al. 2007): RSR = RMSE STDEVobs = Yiobs Yisim2 n i=1 Yiobs Yimean2 n i=1 (4-5) where RMSE is the root mean square error, STDEVobs is the standard deviation of measured values, Y obs are measured values, Y sim are simulated values, and Y mean is the mean of measured values. Overall, NSE coefficients greater than 0.5, RSR values less than or equal to 0.7, and PBIAS values within 0% indicate satisfactory model performance. Land use-water quality regression models were validated using 2004 land use data and average nutrient loads from 2003 to 2005. PBIAS and RSR values were calculated fo r regression models with NSE coefficients greater than 0.5 to evalua te model efficiency. Results Landscape Metrics After identifying and removing metrics from hi ghly correlated (greater than 0.90) metric pairs, PCA was performed on 11 of the 17 landscape metrics calculated. Three principal components had eigenvalues greater than one and after FA, three factors accounted for 76% of the cumulative variation in landscape metrics fo r the entire Biscayne Bay watershed. Metrics with the highest loadings for the first factor, interpreted as patch size variability, were mean

PAGE 120

120 patch size (MPS; 0.92) and patch size standard deviation (PSSD; 0.91). The largest patch index (LPI; -0.94) and landscape shape index (LSI; 0.8 6) had the highest loadings for the second factor, patch diversity. Patch complexity was the third and final factor at the landscape level and included patch size coefficient of variation (PSCoV; 0.84) and area weighted mean shape index (AWMSI; 0.77). PCA and FA results on class metrics are reported for only the four classes that satisfied the minimum distributi on criteria to be included in st epwise regressions: natural land cover, row crops, medium intensity single fa mily residential (MSR), and low intensity commercial (LIC) (Table 4-8). LDI and Imperviousness Average LDI values for the three urbanized sub-basins, C-7 (30.7), C-8 (30.9), and C-9 (29.9) were greater than the mixed land use (C-1; 28.1) and agricultural (C-103; 25.8) subbasins. The overall average LDI value for the fi ve sub-basins was 29.1, with C-8 (31.0) having the single highest value and C-103 (25.4), the lowest (Table 4-9). DCIA values in the five subbasins exhibited a similar pattern to the LD I index as C-8 (35.2%) and C-103 (13.6%) had the highest and lowest values, respectively. For all canal buffers, C-7, C-8, and C-9 LDI valu es continued to be greater than those for C-1 and C-103. The C-7 (500 and 1000 m) and C-8 (1500 m) canal buffers had maximum LDI values while C-103 had the lowest values for al l three buffer distances (Table 4-9). However, LDI values for the different buffers (500, 1000 and 1500 m) at each canal were similar: C-9 (29.6; 29.9; 29.9), C-8 (30.6; 30.9; 31.0), C-7 (31.1; 31.0; 30.9), C1 (29.1; 28.8; 28.5), and C103 (25.6; 25.9; 26.5). C-103 had the lowest DCIA values for all buffer distances (10.0%; 11.9%; 13.9%) while C-7 had the highest value fo r the 500 m (36.1%) distance and C-8 had the highest values for both the 1000 m (36.0%) and 1500 m (36.9%) buffers.

PAGE 121

121 Average LDI values for monitoring site buffers were lower than both the sub-basins and canal buffers (Table 4-9). MW04, a monitoring s ite in the C-103 sub-basin, had the lowest LDI values in the 500 m (6.9) and 1000 m (6.9) buffe rs. The lowest LDI value in the 1500 m buffer was from BL12 (10.5), located in the C-1 sub-ba sin. SK02 (C-9) had the highest LDI values in the 500 m (33.6), 1000 m (33.0), and 1500 m (31.7) bu ffers. DCIA values for C-103 monitoring sites, MW04 and MW13, were zero for both th e 500 m and 1000 m buffers. BL12 (C-1) also had a zero value for DCIA in the 500 m buffer. LR06 (C-7) had the greatest value in the 500 m (38.0%) buffer while BS04 (C-8) had the highest DCIA values in the 1000 m (37.9%) and 1500 m (37.5%) buffers. Loads and Stepwise Regressions Comparing measured loads to LOADEST simula ted results at the eight monitoring sites produced the following average coefficients: NOX-N (NSE: 0.52; R2: 0.60); NOX-N plus NH3-N (NSE: 0.75; R2: 0.85); and TP (NSE: 0.64; R2: 0.58) (Table 4-10). Estim ated annual loads from LOADEST for each site are listed in Table 4-11. For stepwise regressions, landscape metrics with the highest loadings for each of the three factors were MPS, LPI, and PSCoV; these metric s, along with LDI and DCIA, were selected as landscape variables for sub-basins, canal buffers, and site buffers. Row crops and MSR classes met the distribution criteria to be included in all stepwise regressions for the sub-basins, 1000 m canal buffer, 1500 m canal buffer, and 1000 m site bu ffer. Additional classes were needed for the 500 m canal buffer (natural land cover), 500 m site s buffer (natural land cover; LIC), and 1500 m sites buffer (natural land cover; LIC). Regression models selected us ing the forward stepwise procedure generally had greater NSE coefficients after validation compared to mode ls selected from stepwise procedures in both directions (Tables 4-5; 4-6). Using forward and backward step wise regression, the sub-basins

PAGE 122

122 regression model (NSE: 0.69; R2: 0.43; RSR: 0.55; PBIAS: -50.89%) was the best predictor of NOX-N loads (Table 4-12) and the NSE coefficient (greater than 0.5), RSR value (less than 0.7), and PBIAS value (within 70%) all indicated sa tisfactory model perfor mance. The regression model for the sub-basins (Equation 4-6) was as follows: Log (NOX-N load) = 7.9182 0.0982 (LDI) 0.0115 (LPI) (4-6) There were no positive NSE coefficients for NOX-N plus NH3-N loads in the sub-basins, canal buffers, or site buffers after validation. Howe ver, several regression models had positive NSE coefficients for TP loads and evaluation statistics indicated that TP regr ession models performed better than models for NOX-N loads. Models with NSE coefficients greater than 0.75, RSR values less than 0.50, and PBIAS values less than 25% indicate very good performance (Moriasi et al. 2007) and usi ng forward stepwise regression, the 1000 m canal buffer model (NSE: 0.90; R2: 0.89; RSR: 0.31; PBIAS: 7.92%) for TP loads was the only model to satisfy these criteria (Table 4-13). The regression m odel for the 1000 m canal buffer (Equation 4-7) was as follows: TP load = 190.4019 (MSR LPI) 610.2028 (4-7) Discussion Land Use Variables and Nutrient Loads Evaluating land use-water quality relationships within multiple spatial extents enhances the applicability of human disturba nce indicators to aquatic enviro nments by exploring spatially explicit effects (Sponseller et al. 2001; Kearns et al. 2005). Land use variability is reflected in the biological, chemical, and physical condition of aq uatic environments but the relative impact of human influences is mediated by watershed char acteristics that affect pollutant transport efficiency. Strayer et al. (2003) showed that watershed size was an important factor regulating LULC influences on stream ecosystems, riparian buffers have been suggested as key elements to

PAGE 123

123 protect water quality (Schlosser and Karr 1981; Sponseller et al. 2001; Gergel et al. 2002), and other studies have stressed the influence of entir e watersheds (Hunsaker and Levine 1995; Roth et al. 1996; Migliaccio et al. 2007). In the Biscayne Bay watershed, stepwise multivariate regressions revealed spatially expl icit land use variables that were indicators of nutrient loads at eight canal monitoring sites, such as LDI values and MSR metrics. PCA and FA analyses identified landscape-leve l metrics that were most responsible for watershed LULC variability and these indicators, along with LDI values, were important variables for NOX-N loads. The sub-basins regression mode l, including both LDI and the largest patch index (LPI) values at the landscape-level, was the best predictor of NOX-N loads (Table 412) in the study (Figure 4-3). Brown and Vivas ( 2005) suggested that the LDI index represents aggregate land use influences such as polluta nts (e.g., in air and wate r) and physical landscape alterations that reflect the extent of anthropogenic disturbance. Landscape composition is an important factor because the LDI index meas ures disturbance gradients by evaluating nonrenewable energy use per unit area in LULC classes. During the study period (1995 to 2004), medium intensity single family residential (MSR ) land use percentages in the three urban subbasins (C-9, C-8, and C-7) ra nged from 35.9 to 51.7% and in C-1 (mixed land use sub-basin), MSR and row crops were also widespread (45.1% to 47.3%). In C-103 (agricultural sub-basin), four land use classes were respon sible for greater than 76% of the total land use: tree crops, row crops, low intensity single family residential, and MSR. The urban sub-basins therefore had greater LDI values because extens ive urban land use classes, such as MSR, are characterized by greater non-renewable energy use per unit area than ag ricultural land uses. LOADEST results further indicated that C-1 a nd C-103 contributed greater NOX-N loads to the bay than the urban

PAGE 124

124 sub-basins and this suggests that in the subbasin regression model, ag ricultural land use is a strong predictor of NOX-N loads. In the Biscayne Bay watershed, agricultural activ ity is generally locate d in the central and southern sub-basins while the well-developed urban core is found in northern sub-basins. Increasing urbanization has reduced the relative proportion of agri culture in the watershed but overall, agricultural land use is still much more extensive in central and southern urbanizing subbasins such as C-1 and C-103. Therefore, NOX-N loads are substantially larger in areas away from the urban core, such as the South Dade Ag ricultural Area, and this produces an increasing NOX-N gradient from the north to the south. C onversely, LDI index values are greater in the north and this sharp distinction contributed to LDI values being negatively associated with NOXN loads. LPI values at the landscape-level are also negatively associated with NOX-N loads because this indicator reflects la ndscape connectivity, the percentage of the total landscape area comprised of the largest land use patch (McG arigal and Marks 1995). The types of land use (indicated by LDI) and their relative domi nance (indicated by LPI) both influence NOX-N loads within the sub-basins. In contrast to NOX-N, class-level metrics had strong relationships with TP loads. The regression model for the 1000 m canal buffer was th e best predictor of TP loads and the only variable included in this model was LPI for th e MSR class (Figure 4-4). The MSR class had the largest patch in all three urban sub-basins but agricultural classes were dominant in C-1 (row crops) and C-103 (tree crops); complex MSR patches (i.e., considering size, shape, and area) were therefore not as prominent in C-1 and C-103. Due to the limestone geology and soil types in south Florida, phosphorus does not transport as easily through the watershed as nitrogen (Li 2000; Brand et al. 2002; Li and Zh ang 2002) and as a result, TP lo ads were more closely related

PAGE 125

125 to human disturbance indicators at a smaller sp atial extent (1000 m canal buffer) compared to NOX-N loads (sub-basin level). Th e 1000 m canal buffer highlighted the effects of land use variability on TP loads. For example, MSR LP I values in the 1000 m canal buffers between 1995 and 2004 for the urban canals (MSR LPI: 10.0 to 27.5; TP: 1,550 kg yr-1 to 5,467 kg yr-1) exceeded corresponding values for C-1 and C-103 (MSR LPI: 3.1 to 7.7; TP: 84 kg yr-1to 886 kg yr-1). TP loads are generally higher in the norther n, more urbanized area of the watershed and as urbanizing sub-basins such as C-1 and C-103 become more devel oped, subsequent increased TP loads could pose a threat to Biscayne Bay. Uncertainty Analysis Water quality models help decision makers develop measures addressing water resource policy, management, program evaluation, and re gulation (Beck 1987; Sharpley et al. 2002). Uncertainty inherent in water quality samples, lo ad estimations, and model output can all affect the success of initiatives to improve or protect water resources. For example, uncertainty is introduced into water quality sampling through fl ow data measurement, sampling techniques, preservation protocols, and la boratory analysis. Harmel et al (2006) estimated cumulative probable uncertainties for measured streamflow (%), NO3-N (%), NH4-N (%) and TP (%) under typical scenarios (including moderate to extensive quality control/quality assurance procedures) based on a review of published data. Results suggested that models providing output data within 10 to 31% of thes e measured values would be within average uncertainty ranges (Harmel et al. 2006). After reviewing model evaluation techniques for measured data with typical uncertainty estimates Moriasi et al. (2007) recommended NSE, RSR, and PBIAS quantitative statistics to assess model performance. LOADEST (Runkel et al. 2004) was used to estimate nutrient loads at the eight monitoring sites and these load estimates were then used with human disturbance indicators in additional

PAGE 126

126 regressions to identify important land use va riables influencing wate r quality data in the Biscayne Bay watershed. Errors in estimated nutrient loads from LOADE ST could add to the overall error in subsequent regression models de scribing land use-water quality relationships and this was the primary reason for comparing measur ed loads to simulated estimates. In addition, land use data in each sub-basin were compared to aerial maps to improve the quality of land use variables used in regression models. Monthly wate r quality data and daily flow data associated with each monitoring site were also reviewed ca refully to identify and exclude flagged values, such as data violating qu ality control criteria. Several factors can complicate constituent lo ad estimations including deviations from normality, censored data, and retransformation bias (Runkel et al. 2004) but regression models in LOADEST incorporate features such as bias corrections (Cohn et al. 1992) to increase the validity of simulated data. However, studies ha ve still noted problems with constituent load estimations in LOADEST. Clark (2003) suggest ed that because LOADEST does not account for hysteresis effects (different constituent concentrations at similar discharge rates), measured loads after rapid changes in streamflow may produ ce inaccurate simulations. Donato and MacCoy (2005) discussed large errors in daily load estimates due to unus ually high flows but conceded that simulations over longer time periods (e .g., seasonal or annual) reduced overall model inaccuracies because LOADEST estimates were mo re representative of measured data under typical system conditions. To increase the accuracy of load estimates for 1995, 1999, and 2004 that were used in stepwise regressions, LOADE ST was used to estimate annual nutrient loads over a longer timescale (1992 to 2006). Load compar isons between measured and simulated data suggested LOADEST reasonably modeled the vari ability in nutrient loads across the eight monitoring sites (Table 4-10).

PAGE 127

127 The NSE index was used to evaluate the goodness-of-fit between measured loads and LOADEST data as well as to validate land use-water quality models. Although R2 values were calculated as a comparison to NSE coefficients, they were not used to determine the accuracy of regression models because R2 values are only based on consistency between measured and predicted data, thereby producing high correlati on coefficients for models that vastly overestimate or underestimate nutrient loads (Kra use et al. 2005). However, the NSE index can also overestimate larger values in a time series because differences between measured and predicted data are squared (Legates and McCabe 1999). The NSE index is also sensitive to other factors such as sample size and outliers but desp ite these disadvantages, the NSE index has been applied to numerous models because of its flexib ility as a goodness-of-fit statistic (McCuen et al. 2006). Management Implications Canals draining agricultural or highly urbanized subbasins in the Biscayne Bay watershed typically contain elevated inorganic nitrogen (NOX-N and NH3-N) levels (Alleman et al. 1995). For example, Scheidt and Flora (1983) and Cheesman (1989) described high NOX-N concentrations in the C-103 canal that were related to agricu ltural activity. The sub-basin regression models reflected the in fluence of agriculture in the wa tershed as an important land use variable. Although low LDI index va lues correspond to greater NOX-N loads in this study, this suggests that LDI values should not be used alone as a water quality indicator because natural areas, with much lower NOX-N loads, would also have low LDI values. The LDI index should therefore be used with additional assessment tools to evaluate the overall impact of human disturbances (e.g., Mack 2006; Reiss et al. 2009). In south Florida, plans to reduce nitrate load ing from agriculture sources in the Biscayne Bay watershed require an understanding of the un ique hydrology and soil types in the region. For

PAGE 128

128 example, high permeability soils which do not reta in water leads to nitrogen leaching (Li 2000; Li and Zhang 2002) and elevated water tables cause nitrogen-enriched groundwater to enter Biscayne Bay through canal discharges (La ngevin 2000). Li and Zhang (2002) discussed technologies that can help to reduce the impact of agricultural production in south Florida, such as crop-specific fertilizers, slow release fertilizers, and so il organic amendments (e.g., cover crops and compost). For phosphorus, additional inputs to the bay c oncern watershed manage rs because it is the primary nutrient limiting autotrophic grow th (Brand 1988; Kleppel 1996). Phosphorus concentrations are generally higher in the north ern bay (Alleman et al. 1995; Caccia and Boyer 2005), where phytoplankton levels can be five time s greater than the south (Brand 1988). Model evaluation indicated that the strongest relations hip between TP loads and watershed land uses occurred in the 1000 m canal buffer. These result s suggest that developm ent patterns along the canals were important factors contributing to the variability in TP loads. Watershed management plans focused on development patterns within canal (1000 m) buffers could therefore potentially reduce phosphorus discharges to the bay. Several st rategies could be implemented such as zoning policies developed specifically for new resident ial developments within 1000 m of canals in urbanizing sub-basins such as C-1 and C-103. In the watershed, urban stormwater treatment already includes retention systems in housing de velopments to preventing the first flush of pollutants after storm events from entering can als and detention systems (e.g., grassed swales and French drains) have also been utilized to intercept pollu tants (Alleman et al. 1995). Targeting these management practices within cr itical buffer areas and implementing additional measures, such as increased use of riparian bu ffers, could improve overall treatment efficiency. Development patterns regulate the effects of anthropogenic in fluences on natural resources

PAGE 129

129 (Alberti 2005) and increased urba nization throughout the watershed wi ll likely increase TP loads if specific strategies are not implemented in hydrologically sensitive regions such as 1000 m canal buffers. Conclusion In the Biscayne Bay watershed (1995 to 2004), three human disturbance indicators (landscape metrics, LDI index, and impervious ness) associated with eight water quality monitoring sites at multiple spatial extents (subbasins, canal buffers, and site buffers) were analyzed with annual nutrient loads to determine land use factors that influenced water quality variability. The LDI index and metrics at the landscape level (l argest patch index [LPI]) and class level (LPI for medium density single family residential [MSR] cla ss) were identified as land use variables with the strongest relationships to estimated loads from the monitoring sites. The sub-basin regression model wa s the best predictor of annual NOX-N loads in the watershed and included both LDI and LPI variables, indica ting that the relative di stribution of dominant land use classes influences NOX-N loads. TP loads were more closely related to human disturbance indicators at a sma ller spatial extent (1000 m canal buffer), which is a function of nutrient transport processes in th e watershed. The land use variable included in the 1000 m canal buffer (MSR LPI) model suggests that urban de velopment patterns in this buffer zone are important factors for TP loads discharged from the watershed. The LDI index has been applied under various environmental settings to evaluate human disturbance gradients and results from this study suggest that LDI values can be included as one indicator in an overall assessment of water quality. The LDI index, which quantifies th e intensity of land use activities within watersheds, can be used with landscape metrics that evaluate spatial patterns to link land use development to water quality parameters.

PAGE 130

130 Table 4-1. Land use/land cove r data (500 m canal buffer). Land use/Land cover class 1995 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 18.0 7.6 4.1 8.9 9.4 Improved pastures 1.7 1.0 0.0 0.6 1.2 Low intensity pastures 0.4 0.0 0.0 0.7 0.0 Medium intensity recreational, open space 9.8 4.4 3.7 8.5 6.1 Tree crops 0.6 0.0 0.0 3.6 21.8 Row crops 0.0 0.0 0.0 18.8 34.2 High intensity agriculture 0.0 0.0 0.0 0.3 0.1 High intensity recreational 4.8 1.0 1.1 0.7 0.0 Low density single family residential 2.5 5.8 2.5 4.9 10.2 Medium density single family residential 36.3 41.7 55.3 23.1 6.2 High density single family residential 0.0 0.4 1.1 6.0 0.0 Institutional 1.5 5.4 8.0 5.0 1.4 Low density multifamily residential 7.5 3.2 7.5 4.1 2.5 High intensity transportation 5.4 15.0 1.1 5.2 3.9 Low intensity commercial 3.5 8.6 3.9 3.2 1.7 Industrial 5.0 4.8 3.7 4.6 0.1 High intensity commercial 2.3 0.4 5.0 1.8 0.6 High density multifamily residential 0.8 0.6 3.0 0.0 0.5

PAGE 131

131 Table 4-1. Continued. Land use/Land cover class 1999 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 18.9 8.6 2.3 6.6 6.9 Improved pastures 3.1 0.0 0.0 0.0 0.6 Low intensity pastures 0.0 0.0 0.0 0.1 0.0 Medium intensity recreational, open space 7.5 2.0 2.5 8.6 5.8 Tree crops 0.4 0.0 0.0 3.3 22.4 Row crops 0.0 0.0 0.0 17.0 32.6 High intensity agriculture 0.0 0.0 0.0 0.0 0.0 High intensity recreational 5.6 1.0 1.6 0.9 0.2 Low density single family residential 0.0 6.4 1.8 3.1 11.7 Medium density single family residential 32.0 41.7 50.2 22.5 6.2 High density single family residential 1.2 0.2 1.8 10.8 1.5 Institutional 2.9 6.4 8.9 6.0 1.8 Low density multifamily residential 8.9 6.0 6.4 4.9 4.2 High intensity transportation 3.9 12.2 3.2 5.2 3.5 Low intensity commercial 4.8 9.8 8.7 3.6 1.8 Industrial 7.9 4.8 3.9 4.7 0.3 High intensity commercial 1.2 0.0 5.7 2.3 0.3 High density multifamily residential 1.7 0.8 3.0 0.4 0.0

PAGE 132

132 Table 4-1. Continued. Land use/Land cover class 2004 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 15.4 8.0 2.3 5.5 3.4 Improved pastures 3.1 0.0 0.0 0.0 0.2 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 7.3 2.4 3.2 7.0 6.4 Tree crops 0.4 0.0 0.0 3.0 30.2 Row crops 0.0 0.0 0.0 13.8 17.0 High intensity agriculture 0.0 0.0 0.0 0.0 0.0 High intensity recreational 6.4 0.8 1.8 1.0 0.5 Low density single family residential 0.0 6.2 1.8 1.9 12.7 Medium density single family residential 34.4 42.5 50.2 25.1 11.4 High density single family residential 1.2 0.2 1.1 14.1 4.0 Institutional 2.5 6.6 8.4 5.8 2.1 Low density multifamily residential 10.2 6.0 7.1 6.1 4.7 High intensity transportation 4.1 12.0 3.4 5.1 3.5 Low intensity commercial 4.6 9.6 8.9 3.6 2.8 Industrial 8.5 4.8 3.0 5.4 0.3 High intensity commercial 1.0 0.0 5.5 2.4 0.3 High density multifamily residential 0.8 0.8 3.2 0.3 0.5

PAGE 133

133 Table 4-2. Land use/land cover data (1000 m canal buffer). Land use/Land cover class 1995 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 14.6 4.8 2.9 7.9 7.9 Improved pastures 1.4 1.4 0.0 0.6 0.8 Low intensity pastures 0.6 0.0 0.0 0.4 0.0 Medium intensity recreational, open space 12.2 3.9 5.6 7.1 6.0 Tree crops 0.3 0.0 0.0 4.7 20.6 Row crops 0.0 0.0 0.0 19.5 32.1 High intensity agriculture 0.0 0.0 0.0 0.3 0.1 High intensity recreational 6.0 1.7 1.5 1.3 0.3 Low density single family residential 2.9 5.7 1.6 5.2 10.3 Medium density single family residential 34.5 42.0 53.2 25.3 9.4 High density single family residential 0.2 0.6 2.2 5.3 0.7 Institutional 1.8 5.8 6.9 6.5 1.9 Low density multifamily residential 7.4 3.4 8.4 3.1 2.2 High intensity transportation 5.1 14.4 1.7 4.4 4.2 Low intensity commercial 4.3 8.4 4.2 2.5 2.6 Industrial 5.0 5.9 4.9 3.9 0.1 High intensity commercial 2.3 0.9 3.4 1.7 0.5 High density multifamily residential 1.3 0.8 3.5 0.2 0.3

PAGE 134

134 Table 4-2. Continued. Land use/Land cover class 1999 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 14.4 5.3 2.1 5.3 7.2 Improved pastures 2.4 0.0 0.0 0.0 0.4 Low intensity pastures 0.0 0.0 0.0 0.1 0.0 Medium intensity recreational, open space 10.5 2.9 4.3 7.6 5.1 Tree crops 0.2 0.0 0.0 4.7 22.3 Row crops 0.0 0.0 0.0 16.4 28.8 High intensity agriculture 0.0 0.0 0.0 0.1 0.0 High intensity recreational 7.3 1.0 1.8 1.5 0.4 Low density single family residential 0.0 5.1 1.0 4.2 11.1 Medium density single family residential 30.9 42.1 49.4 24.6 10.0 High density single family residential 1.5 0.1 2.1 10.9 2.1 Institutional 3.8 6.1 8.7 7.2 2.6 Low density multifamily residential 8.7 6.6 7.6 4.2 3.0 High intensity transportation 5.2 12.9 3.0 4.3 3.4 Low intensity commercial 4.9 10.9 8.1 3.1 2.7 Industrial 6.7 5.6 4.9 3.3 0.2 High intensity commercial 1.3 0.2 3.8 2.1 0.7 High density multifamily residential 2.3 1.0 3.2 0.5 0.2

PAGE 135

135 Table 4-2. Continued. Land use/Land cover class 2004 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 11.8 4.5 2.1 4.1 4.7 Improved pastures 2.2 0.0 0.0 0.0 0.1 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 10.0 3.0 4.7 6.5 5.6 Tree crops 0.2 0.0 0.0 3.3 28.2 Row crops 0.0 0.0 0.0 14.2 17.2 High intensity agriculture 0.0 0.0 0.0 0.2 0.0 High intensity recreational 7.5 0.9 2.1 1.5 0.5 Low density single family residential 0.0 5.0 1.1 2.2 11.9 Medium density single family residential 32.6 42.4 49.7 27.7 14.3 High density single family residential 1.5 0.2 1.0 14.0 3.7 Institutional 3.6 6.4 8.4 7.1 2.4 Low density multifamily residential 9.7 7.3 7.8 5.1 3.4 High intensity transportation 5.2 12.8 3.1 4.2 3.6 Low intensity commercial 5.3 10.5 8.4 3.1 3.1 Industrial 7.3 5.6 4.8 4.0 0.2 High intensity commercial 1.2 0.2 3.5 2.2 0.8 High density multifamily residential 1.8 1.1 3.4 0.5 0.4

PAGE 136

136 Table 4-3. Land use/land cover data (1500 m canal buffer). Land use/Land cover class 1995 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 10.6 4.5 2.5 7.2 8.0 Improved pastures 1.8 1.3 0.0 0.6 0.6 Low intensity pastures 1.7 0.0 0.0 0.3 0.0 Medium intensity recreational, open space 10.9 3.6 5.8 6.9 5.2 Tree crops 0.2 0.0 0.1 5.5 20.3 Row crops 0.0 0.0 0.0 21.0 27.8 High intensity agriculture 0.0 0.0 0.0 0.2 0.1 High intensity recreational 7.2 1.9 1.6 2.9 0.6 Low density single family residential 2.9 5.0 1.5 4.9 11.6 Medium density single family residential 37.7 42.6 52.6 25.7 11.1 High density single family residential 0.4 0.6 1.9 4.4 0.7 Institutional 1.6 5.6 5.4 6.0 2.0 Low density multifamily residential 6.7 3.5 8.1 2.5 2.4 High intensity transportation 5.4 14.6 3.6 4.3 4.6 Low intensity commercial 4.6 9.3 5.3 2.2 3.4 Industrial 4.9 6.0 5.8 3.6 0.6 High intensity commercial 2.3 1.0 3.1 1.4 0.7 High density multifamily residential 1.1 0.3 2.6 0.4 0.3

PAGE 137

137 Table 4-3. Continued Land use/Land cover class 1999 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 10.9 4.1 1.9 4.9 7.7 Improved pastures 2.2 0.0 0.0 0.0 0.2 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 9.9 3.4 4.5 7.9 4.3 Tree crops 0.1 0.0 0.0 5.7 21.3 Row crops 0.0 0.0 0.0 17.1 25.3 High intensity agriculture 0.0 0.0 0.0 0.1 0.0 High intensity recreational 7.3 1.3 2.8 3.3 0.5 Low density single family residential 0.0 4.2 1.0 3.6 12.5 Medium density single family residential 35.4 42.5 48.7 26.0 11.5 High density single family residential 2.5 0.3 1.8 9.4 1.9 Institutional 3.9 6.0 7.2 6.6 3.0 Low density multifamily residential 8.2 6.0 8.6 3.2 2.7 High intensity transportation 4.6 14.8 3.1 4.3 3.7 Low intensity commercial 5.7 11.0 9.5 2.7 3.8 Industrial 5.9 5.1 5.1 2.8 0.6 High intensity commercial 1.2 0.3 3.2 1.6 0.8 High density multifamily residential 2.2 1.0 2.4 0.6 0.2

PAGE 138

138 Table 4-3. Continued. Land use/Land cover class 2004 Land use/Land cover (%) C-9 C-8 C-7 C-1 C-103 Natural land/water 9.0 3.2 1.9 3.3 5.2 Improved pastures 1.5 0.0 0.0 0.0 0.1 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 9.7 2.9 4.5 6.5 4.8 Tree crops 0.1 0.0 0.0 4.0 26.9 Row crops 0.0 0.0 0.0 15.0 15.5 High intensity agriculture 0.0 0.0 0.0 0.2 0.0 High intensity recreational 7.4 1.2 2.9 3.4 0.6 Low density single family residential 0.0 4.1 1.1 2.1 12.9 Medium density single family residential 36.6 42.8 48.8 29.0 15.3 High density single family residential 2.7 0.3 1.1 13.0 2.8 Institutional 3.5 6.1 7.0 6.5 2.9 Low density multifamily residential 8.9 7.1 8.9 4.0 3.2 High intensity transportation 4.5 14.5 3.1 4.3 3.9 Low intensity commercial 6.2 11.0 9.7 2.8 4.2 Industrial 6.8 5.1 5.3 3.6 0.6 High intensity commercial 1.3 0.7 3.1 1.7 0.9 High density multifamily residential 1.8 0.9 2.5 0.6 0.2

PAGE 139

139Table 4-4. Land use/land cover data (500 m water quality monitoring site buffer). Land use/Land cover class 1995 Land use/Land cover (%) SK09 SK02 BS04 LR06 BL12 BL03 MW13 MW04 Natural land/water 69.2 0.0 0.0 0.0 0.0 9.1 23.1 0.0 Improved pastures 23.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 0.0 7.7 0.0 0.0 0.0 0.0 0.0 0.0 Tree crops 0.0 0.0 0.0 0.0 0.0 36.4 38.5 61.5 Row crops 0.0 0.0 0.0 0.0 100.0 0.0 38.5 38.5 High intensity agriculture 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High intensity recreational 0.0 7.7 18.2 18.2 0.0 0.0 0.0 0.0 Low density single family residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium density single family residential 0.0 7.7 72.7 72.7 0.0 0.0 0.0 0.0 High density single family residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Institutional 0.0 7.7 0.0 0.0 0.0 0.0 0.0 0.0 Low density multifamily residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High intensity transportation 0.0 23.1 9.1 9.1 0.0 0.0 0.0 0.0 Low intensity commercial 0.0 30.8 0.0 0.0 0.0 0.0 0.0 0.0 Industrial 7.7 0.0 0.0 0.0 0.0 54.5 0.0 0.0 High intensity commercial 0.0 15.4 0.0 0.0 0.0 0.0 0.0 0.0 High density multifamily residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

PAGE 140

140Table 4-4. Continued. Land use/Land cover class 1999 Land use/Land cover (%) SK09 SK02 BS04 LR06 BL12 BL03 MW13 MW04 Natural land/water 46.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Improved pastures 23.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity pastures 7.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 7.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Tree crops 0.0 0.0 0.0 0.0 0.0 27.3 61.5 61.5 Row crops 0.0 0.0 0.0 0.0 100.0 0.0 38.5 38.5 High intensity agriculture 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High intensity recreational 0.0 7.7 9.1 0.0 0.0 0.0 0.0 0.0 Low density single family residential 7.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium density single family residential 0.0 0.0 63.6 53.8 0.0 0.0 0.0 0.0 High density single family residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Institutional 0.0 7.7 18.2 0.0 0.0 0.0 0.0 0.0 Low density multifamily residential 0.0 15.4 9.1 30.8 0.0 0.0 0.0 0.0 High intensity transportation 0.0 23.1 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity commercial 0.0 46.2 0.0 15.4 0.0 0.0 0.0 0.0 Industrial 7.7 0.0 0.0 0.0 0.0 72.7 0.0 0.0 High intensity commercial 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High density multifamily residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

PAGE 141

141Table 4-4. Continued. Land use/Land cover class 2004 Land use/Land cover (%) SK09 SK02 BS04 LR06 BL12 BL03 MW13 MW04 Natural land/water 30.8 0.0 0.0 0.0 0.0 9.1 0.0 0.0 Improved pastures 7.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 15.4 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Tree crops 0.0 0.0 0.0 0.0 0.0 18.2 61.5 100.0 Row crops 0.0 0.0 0.0 0.0 100.0 0.0 38.5 0.0 High intensity agriculture 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High intensity recreational 0.0 7.7 9.1 0.0 0.0 0.0 0.0 0.0 Low density single family residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium density single family residential 7.7 0.0 63.6 53.8 0.0 0.0 0.0 0.0 High density single family residential 23.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Institutional 0.0 7.7 18.2 0.0 0.0 0.0 0.0 0.0 Low density multifamily residential 7.7 15.4 9.1 30.8 0.0 0.0 0.0 0.0 High intensity transportation 0.0 23.1 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity commercial 0.0 46.2 0.0 15.4 0.0 0.0 0.0 0.0 Industrial 7.7 0.0 0.0 0.0 0.0 72.7 0.0 0.0 High intensity commercial 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High density multifamily residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

PAGE 142

142Table 4-5. Land use/land cover data (1000 m water quality monitoring site buffer). Land use/Land cover class 1995 Land use/Land cover (%) SK09 SK02 BS04 LR06 BL12 BL03 MW13 MW04 Natural land/water 30.0 5.0 0.0 2.5 5.0 23.3 12.5 15.0 Improved pastures 32.5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 0.0 2.5 0.0 0.0 0.0 0.0 0.0 0.0 Tree crops 0.0 0.0 0.0 0.0 5.0 58.1 40.0 50.0 Row crops 0.0 0.0 0.0 0.0 87.5 0.0 47.5 35.0 High intensity agriculture 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High intensity recreational 0.0 5.0 2.3 0.0 0.0 0.0 0.0 0.0 Low density single family residential 5.0 0.0 0.0 5.0 2.5 0.0 0.0 0.0 Medium density single family residential 5.0 40.0 86.4 55.0 0.0 0.0 0.0 0.0 High density single family residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Institutional 0.0 2.5 2.3 2.5 0.0 0.0 0.0 0.0 Low density multifamily residential 0.0 0.0 0.0 30.0 0.0 0.0 0.0 0.0 High intensity transportation 2.5 12.5 4.5 0.0 0.0 0.0 0.0 0.0 Low intensity commercial 0.0 17.5 2.3 5.0 0.0 0.0 0.0 0.0 Industrial 25.0 0.0 0.0 0.0 0.0 18.6 0.0 0.0 High intensity commercial 0.0 15.0 2.3 0.0 0.0 0.0 0.0 0.0 High density multifamily residential 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

PAGE 143

143Table 4-5. Continued. Land use/Land cover class 1999 Land use/Land cover (%) SK09 SK02 BS04 LR06 BL12 BL03 MW13 MW04 Natural land/water 7.5 10.0 0.0 0.0 5.0 18.6 0.0 15.0 Improved pastures 32.5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity pastures 5.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 10.0 0.0 0.0 0.0 0.0 7.0 0.0 0.0 Tree crops 0.0 0.0 0.0 0.0 5.0 46.5 57.5 55.0 Row crops 0.0 0.0 0.0 0.0 87.5 0.0 42.5 30.0 High intensity agriculture 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High intensity recreational 0.0 7.5 0.0 0.0 0.0 0.0 0.0 0.0 Low density single family residential 10.0 0.0 0.0 0.0 2.5 0.0 0.0 0.0 Medium density single family residential 17.5 5.0 86.4 57.5 0.0 0.0 0.0 0.0 High density single family residential 2.5 0.0 0.0 2.5 0.0 0.0 0.0 0.0 Institutional 0.0 2.5 4.5 2.5 0.0 0.0 0.0 0.0 Low density multifamily residential 10.0 30.0 4.5 17.5 0.0 0.0 0.0 0.0 High intensity transportation 0.0 15.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity commercial 0.0 27.5 4.5 20.0 0.0 0.0 0.0 0.0 Industrial 5.0 0.0 0.0 0.0 0.0 27.9 0.0 0.0 High intensity commercial 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High density multifamily residential 0.0 2.5 0.0 0.0 0.0 0.0 0.0 0.0

PAGE 144

144Table 4-5. Continued. Land use/Land cover class 2004 Land use/Land cover (%) SK09 SK02 BS04 LR06 BL12 BL03 MW13 MW04 Natural land/water 0.0 10.0 0.0 2.5 5.0 23.3 7.5 5.0 Improved pastures 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 10.0 0.0 0.0 0.0 0.0 7.0 0.0 0.0 Tree crops 0.0 0.0 0.0 0.0 10.0 25.6 52.5 92.5 Row crops 0.0 0.0 0.0 0.0 70.0 16.3 40.0 2.5 High intensity agriculture 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High intensity recreational 0.0 5.0 0.0 0.0 0.0 0.0 0.0 0.0 Low density single family residential 7.5 0.0 0.0 5.0 15.0 0.0 0.0 0.0 Medium density single family residential 25.0 7.5 86.4 55.0 0.0 0.0 0.0 0.0 High density single family residential 35.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Institutional 0.0 2.5 4.5 2.5 0.0 0.0 0.0 0.0 Low density multifamily residential 15.0 30.0 4.5 30.0 0.0 0.0 0.0 0.0 High intensity transportation 0.0 15.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity commercial 0.0 27.5 4.5 5.0 0.0 0.0 0.0 0.0 Industrial 7.5 0.0 0.0 0.0 0.0 27.9 0.0 0.0 High intensity commercial 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High density multifamily residential 0.0 2.5 0.0 0.0 0.0 0.0 0.0 0.0

PAGE 145

145Table 4-6. Land use/land cover data (1500 m water quality monitoring site buffer). Land use/Land cover class 1995 Land use/Land cover (%) SK09 SK02 BS04 LR06 BL12 BL03 MW13 MW04 Natural land/water 20.8 5.9 0.0 1.0 2.0 38.4 8.9 26.7 Improved pastures 26.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity pastures 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 3.0 4.0 0.0 0.0 0.0 4.0 0.0 1.0 Tree crops 0.0 0.0 0.0 0.0 5.0 47.5 33.7 31.7 Row crops 0.0 0.0 0.0 0.0 89.1 0.0 46.5 37.6 High intensity agriculture 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 High intensity recreational 2.0 12.9 1.0 0.0 0.0 0.0 0.0 0.0 Low density single family residential 15.8 0.0 1.0 3.0 3.0 0.0 5.9 0.0 Medium density single family residential 8.9 39.6 85.9 52.5 0.0 0.0 0.0 0.0 High density single family residential 0.0 0.0 0.0 2.0 0.0 0.0 0.0 0.0 Institutional 0.0 2.0 6.1 4.0 0.0 0.0 0.0 0.0 Low density multifamily residential 2.0 0.0 0.0 26.7 0.0 0.0 0.0 0.0 High intensity transportation 2.0 5.9 2.0 0.0 0.0 0.0 5.0 3.0 Low intensity commercial 0.0 18.8 2.0 10.9 0.0 0.0 0.0 0.0 Industrial 18.8 1.0 0.0 0.0 0.0 10.1 0.0 0.0 High intensity commercial 0.0 8.9 2.0 0.0 0.0 0.0 0.0 0.0 High density multifamily residential 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0

PAGE 146

146Table 4-6. Continued. Land use/Land cover class 1999 Land use/Land cover (%) SK09 SK02 BS04 LR06 BL12 BL03 MW13 MW04 Natural land/water 16.8 8.9 0.0 1.0 4.0 27.3 0.0 34.7 Improved pastures 17.8 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity pastures 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 9.9 2.0 0.0 0.0 1.0 13.1 1.0 0.0 Tree crops 0.0 0.0 0.0 0.0 10.9 44.4 42.6 33.7 Row crops 0.0 0.0 0.0 0.0 82.2 0.0 47.5 29.7 High intensity agriculture 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 High intensity recreational 2.0 11.9 0.0 0.0 0.0 0.0 0.0 0.0 Low density single family residential 18.8 0.0 0.0 0.0 1.0 0.0 5.9 0.0 Medium density single family residential 19.8 21.8 83.8 51.5 0.0 0.0 0.0 0.0 High density single family residential 1.0 0.0 1.0 3.0 0.0 0.0 0.0 0.0 Institutional 0.0 4.0 6.1 3.0 0.0 0.0 0.0 0.0 Low density multifamily residential 5.9 20.8 6.1 19.8 0.0 0.0 0.0 0.0 High intensity transportation 1.0 6.9 0.0 0.0 0.0 0.0 3.0 2.0 Low intensity commercial 3.0 21.8 3.0 21.8 0.0 0.0 0.0 0.0 Industrial 2.0 1.0 0.0 0.0 0.0 15.2 0.0 0.0 High intensity commercial 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High density multifamily residential 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0

PAGE 147

147Table 4-6. Continued. Land use/Land cover class 2004 Land use/Land cover (%) SK09 SK02 BS04 LR06 BL12 BL03 MW13 MW04 Natural land/water 6.9 9.9 0.0 1.0 2.0 17.2 6.9 31.7 Improved pastures 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Low intensity pastures 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium intensity recreational, open space 0.0 1.0 0.0 0.0 13.9 13.1 0.0 0.0 Tree crops 10.9 0.0 0.0 0.0 3.0 31.3 40.6 56.4 Row crops 0.0 0.0 0.0 0.0 68.3 8.1 43.6 9.9 High intensity agriculture 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 High intensity recreational 18.8 9.9 1.0 0.0 11.9 0.0 5.9 0.0 Low density single family residential 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Medium density single family residential 22.8 22.8 83.8 51.5 0.0 0.0 0.0 0.0 High density single family residential 20.8 0.0 0.0 2.0 0.0 15.2 0.0 0.0 Institutional 3.0 4.0 3.0 21.8 0.0 0.0 0.0 0.0 Low density multifamily residential 0.0 20.8 6.1 4.0 0.0 0.0 0.0 0.0 High intensity transportation 2.0 7.9 0.0 0.0 0.0 0.0 3.0 2.0 Low intensity commercial 3.0 21.8 0.0 0.0 0.0 0.0 0.0 0.0 Industrial 7.9 1.0 6.1 19.8 0.0 15.2 0.0 0.0 High intensity commercial 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 High density multifamily residential 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0

PAGE 148

148 Table 4-7. Summary statistics for nutrient concentrations (1992 to 2006) and fl ow data at eight water quality monitoring sites in the Biscayne Bay watershed. Constituent Site Min1 Mean Median Max2 Mean Annual (mg L-1) (mg L-1) (mg L-1) (mg L-1) Discharge (m3 s-1) NOX-N SK09 0.010 0.055 0.050 0.200 5.88 SK02 0.010 0.224 0.210 0.680 8.88 BS04 0.010 0.222 0.200 1.750 3.40 LR06 0.020 0.271 0.280 0.530 3.87 BL12 0.010 0.022 0.010 0.480 6.00 BL03 0.010 0.288 0.245 1.010 6.00 MW13 0.010 0.065 0.030 2.410 0.69 MW04 0.010 2.200 2.250 4.640 1.41 NH3-N SK09 0.010 0.204 0.200 0.430 SK02 0.008 0.124 0.070 0.540 BS04 0.008 0.111 0.070 0.410 LR06 0.010 0.390 0.370 0.930 BL12 0.020 0.332 0.340 0.820 BL03 0.008 0.086 0.030 0.400 MW13 0.008 0.230 0.205 0.600 MW04 0.008 0.025 0.020 0.100 TP SK09 0.010 0.006 0.005 0.040 SK02 0.001 0.010 0.008 0.170 BS04 0.003 0.018 0.015 0.170 LR06 0.004 0.024 0.022 0.170 BL12 0.001 0.009 0.005 0.210 BL03 0.001 0.007 0.005 0.170 MW13 0.001 0.007 0.004 0.170 MW04 0.001 0.006 0.004 0.048 1Water quality targets: NOX-N (0.05 mg L-1 in Biscayne National Park); NH3-N (0.01 mg L-1 within Biscayne National Park; 0.05 mg L-1 throughout the bay; 0.5 mg L-1 for surface waters in Miami-Dade County); total nitrogen (0.9 mg L-1 for ecoregion XII, southern coastal plain); TP (0.04 mg L-1 for ecoregion XII). 2Minimum detection limits (MDLs) were used as maximum concentrations for censored data. MDLs for censored TP data duri ng the study period included 0.17 mg L-1.

PAGE 149

149 Table 4-8. Class-level factors and metrics descri bing spatial variability in the Biscayne Bay watershed. Class Factor 11 Factor 2 Factor 3 MPS (0.96) LSI (0.81) ED (0.92) LPI (0.92) PSCoV (0.73) IJI (-0.80) Natural land cover PSSD (0.87) AWMSI (0.65) %LAND (0.76) CA (0.80) MPI (0.93) ED (0.80) N/A PSSD (0.93) %LAND (0.80) Row crops AWMSI (0.87) PSCoV (0.86) PSCoV (0.96) LPI (0.99) MNN (-0.56) MSR AWMSI (0.88) %LAND (0.88) MPS (0.55) CA (0.84) MPI (0.93) %LAND (0.84) N/A LIC AWMSI (0.92) LPI (0.79) PSCoV (0.91) MPS (0.90) CA (0.87) 1Metric loadings for each factor after principal component analysis and factor analysis. Metrics: mean patch size (MPS); largest patch index (LPI); patch size standard de viation (PSSD); class area (CA); mean proximity index (MPI); area weighted mean shape inde x (AWMSI); patch size coefficient of variation (PSCoV); landscape shap e index (LSI); edge density (ED); % landscape (%LAND); interspersion and juxtaposition i ndex (IJI); mean nearest neighbor (MNN).

PAGE 150

150 Table 4-9. Summary statistics for Landscape De velopment Intensity (LDI) index values and Directly Connected Impervious Area (DCI A) percentages (1995 to 2004) considering multiple spatial extents in the Biscayne Bay watershed. Spatial extent LDI DCIA Min Mean Max Min Mean Max Sub-basins 25.4 29.1 31.0 13.6 27.0 35.2 Canals ( 500 m) 25.1 29.2 31.3 10.0 26.6 36.1 Canals (1000 m) 25.5 29.3 31.2 11.9 27.0 36.0 Canals (1500 m) 26.2 29.4 31.2 13.9 27.5 36.9 Stations ( 500 m) 6.9 22.0 33.6 0.0 18.4 38.0 Stations (1000 m) 6.9 21.9 33.0 0.0 18.1 37.9 Stations (1500 m) 10.5 24.1 31.7 0.3 17.7 37.5

PAGE 151

151 Table 4-10. Average Nash-Sutcli ffe Efficiency (NSE) coeffici ents for eight water quality monitoring sites (1992 to 2006) in the Biscayne Bay watershed after comparing LOADEST simulated loads to measured loads. Site NOX-N NOX-N plus NH3-N TP SK09 0.74 0.86 0.57 SK02 0.41 0.72 0.38 BS04 0.20 0.46 0.87 LR06 0.73 0.71 0.92 BL12 0.30 0.86 0.58 BL03 0.49 0.72 0.50 MW13 0.45 0.84 0.61 MW04 0.83 0.84 0.65 Average 0.52 0.75 0.64

PAGE 152

152 Table 4-11. Median annual loads (1992 to 2006) at eight water quality monitoring sites in the Biscayne Bay watershed. Site Basin NOX-N (kg/yr) NOX-N plus NH3-N (kg/yr)TP (kg/yr) SK09 C-9 13,161 56,559 1,099 SK02 C-9 79,658 136,227 2,816 BS04 C-8 35,056 57,035 2,288 LR06 C-7 48,907 145,854 4,738 BL12 C-1 1,572 49,172 562 BL03 C-1 81,888 136,026 1,196 MW13 C-103 615 5,744 63 MW04 C-103 157,248 158,127 155

PAGE 153

153 Table 4-12. Validation results for NOX-N stepwise regression mode ls (forward and backward selection and forward direction only) using three quantitative statistics. Spatial extent Forward and Backward Forward Only NSE RSR1 PBIAS NSE RSR PBIAS Sub-basins 0.69 0.55 -50.89 0.55 0.67 -117.81 Canals ( 500 m) -6.57 0.40 Canals (1000 m) -0.36 0.41 Canals (1500 m) 0.39 0.55 0.67 -109.73 Stations ( 500 m) -3.84 -22.28 Stations (1000 m) -0.52 -0.49 Stations (1500 m) -8512.37 -33.84 1Ratio of the root mean square error to the sta ndard deviation of measur ed data (RSR) values were only calculated for regressi on models with Nash-Sutcliffe Efficiency (NSE) coefficients greater than 0.5. Percentage bias (PBIAS) values were only calcul ated for RSR values less than 0.7.

PAGE 154

154 Table 4-13. Validation results for TP stepwise regression models (forward and backward selection and forward direction only) using three quantitative statistics. Spatial extent Forward and Backward Forward Only NSE RSR1 PBIAS NSE RSR PBIAS Sub-basins 0.79 0.46 -47.37 0.69 0.56 -119.14 Canals ( 500 m) 0.98 0.13 61.34 0.98 0.13 61.34 Canals (1000 m) -4.76 0.90 0.31 7.92 Canals (1500 m) 0.38 0.67 0.57 -45.00 Stations ( 500 m) 0.72 0.53 -37.64 0.57 0.66 33.15 Stations (1000 m) 0.82 0.42 61.71 0.66 0.58 56.27 Stations (1500 m) -0.55 0.47 1Ratio of the root mean square error to the sta ndard deviation of measur ed data (RSR) values were only calculated for regressi on models with Nash-Sutcliffe Efficiency (NSE) coefficients greater than 0.5. Percentage bias (PBIAS) values were only calcul ated for RSR values less than 0.7.

PAGE 155

155 Figure 4-1. Buffers (500, 1000, and 1500 m) for five canals in the Biscayne Bay watershed.

PAGE 156

156 Figure 4-2. Buffers (500, 1000, and 1500 m) for ei ght water quality monitoring sites in the Biscayne Bay watershed.

PAGE 157

157A) B) C) Figure 4-3. Land use vari ables influencing NOX-N loads in the Biscayne Bay watershed. A) Predicted loads (2004) for the sub-basins regression model compared to LOADEST NOX-N loads. B) Landscape Development Intensity Index (LDI) values and LOADEST NOX-N loads (1995 to 2004) in the five study sub-basins. C) Largest Patch Index (LPI ) metric percentages and LOADEST NOX-N loads (1995 to 2004) in the five study sub-basins. R = 0.43 0 50,000 100,000 150,000 050,000100,000150,000Predicted Loads (kg/yr)NOX-N Loads (kg/yr) R = 0.79 20 22 24 26 28 30 32 0100,000200,000300,000LDINOX-N Loads (kg/yr) R = 0.21 0 10 20 30 0100,000200,000300,000LPI (%)NOX-N Loads (kg/yr)

PAGE 158

158A) B) Figure 4-4. Land use variables influencing TP loads in the Biscayne Ba y watershed. A) Predicted TP loads (2004) for the 1000 m canal buffer regression model compared to LOADEST TP loads. B) Largest Patch Index (LPI) metric percentages for medium density single family residential (MSR) class and LOADEST TP loads (1995 to 2004) in the 1000 m canal buffers. R = 0.89 0 2,000 4,000 6,000 02,0004,0006,000Predicted Loads (kg/yr)TP Loads (kg/yr) R = 0.55 0 10 20 30 40 02,0004,0006,000LPI for MSR Class (%)TP Loads (kg/yr)

PAGE 159

159 CHAPTER 5 SUMMARY The predominant form of land use change in south Florida in recent decades has been the conversion of agricultural and natural areas to residential or urban complexes (Solecki and Walker 2001) and land use influences, along with watershed management practices, directly affect water resources such as Biscayne Ba y, which drains the Miami metropolitan area. Biscayne Bay requires minimal inputs of phos phorus and nitrogen to function and thus watershed nutrient inputs have a controlling influence on bay water quality (Browder et al. 2005). Spatial characteristics of land use change have not been quantified and linked to specific pollutants in the watershed and therefore the overall objective for this study was to evaluate temporal and spatial land use influences on ti me series nutrient concentrations and loads measured in canals discharging to Biscayne Ba y. Specific objectives and results are summarized below. Objective 1 Evaluate three disturbance indicators (landscape metr ics, Landscape Development Intensity [LDI] index, and percent imperviousness) in the Biscayne Bay watershed for 1995, 1999, and 2004. Specific objectives: (1) quantify disturbance indicato rs in five subbasins representing agricultural, urban, and mixed land uses; (2) determine if selected disturbance indicators provide contrasting information; and (3) evaluate how these indicators could potentially influe nce watershed management decisions. All three disturbance indicators revealed di fferent levels of anthropogenic disturbance among urban (C-9, C-8, and C-7), mixed land use (C-1), and agricultural (C-103) subbasins. Landscape metrics provided information on influential land use classes within the five sub-basins such as medium density single family residential (MSR) and row crops. LDI and DCIA values both provided similar info rmation regarding the intensity of human disturbance; urban sub-basins were the most disturbed but th e greatest changes occurred in C-1 and C-103. Overall, disturbance indicators suggested that the three urban sub-basins were relatively stable and dominated by complex MSR patches that corresponded to a greater degree of

PAGE 160

160 anthropogenic intensity compared to the mixed land use and agricultural sub-basins, which were urbanizing through the conversi on of row crops to residential uses. LDI values below 30.0 and DCIA values belo w 25% reflect sub-basins that are not completely urbanized. In these developing sub-basins, disturbance indicators can provide complementary information for watershed management decisions rega rding water quality because LDI and DCIA values provide an ove rall view of watershed development but landscape metrics describe spat ial configurations that coul d compound the threat to water resources. Implications for municipality zoning regula tions in urbanizing sub-basins include promoting MSR and HSR development, as oppos ed to LSR, at greater distances from aquatic corridors. Implementing BMPs in th ese critical areas w ould also aid zoning regulations and reduce hydrologic impacts of gradually increasing LDI and DCIA values. Objective 2 Evaluate historical nutrient water quality data from 1992 to 2006 at six monitoring sites located near the outlets of canals discha rging to the bay. Specific objectives: (1) determine nutrient concentra tion trends during the study pe riod; (2) estimate annual nutrient loads from six canals in the watershe d; and (3) use the PED index to assess the proportional impact of nutrient discharges from various canals. The majority of NOX-N, NH3-N, and TP concentrations decr eased or exhibited no change at the six water quality monitoring sites, with only six instances of significantly (p < 0.1) increasing trends. LR06, BL03, and MW04 all had increasing NOX-N trends, BL03 was the only site with an increasing NH3-N trend, and both BS04 and MW04 had increasing TP trends. Only MW04 had median inorganic nitrogen concentrations (2.27 mg /L) that exceeded USEPA Southern Coastal Plain (ecoregion XII) criteria for total nitrogen (0.9 mg/L). The median NH3-N concentration at BL03 (0.03 mg/L) was below Miami-Dade County surface water standard (0.5 mg/L). Median TP concentrations at BS04 (0.015 mg/L) and MW04 (0.004 mg/L) were both below ecoregion XII criteria for TP (0.04 mg/L). Annual nutrient loads at the si x sites revealed higher NOX-N loads in the southern section of the watershed and higher NH3-N and TP loads in the northern and central areas. Annual nutrient loads fluctuated greatly but corresponding PED index values were less sensitive, exhibiting a damped response to nut rient fluxes. PED index values suggest that canal discharges from two sites (MW04 in the C-103 sub-basin and LR06 in the C-7 subbasin) provide a greater proportional impact in the bay compared to other sites. In the Biscayne Bay watershed, there is an ur gent need for assessment tools that can be used to guide management initiatives regard ing water quality discharges to the bay. The PED index is a new analytic tool that can be used to evaluate the potential impact of

PAGE 161

161 canal discharges because it provides a relativ e indication of ecological stress associated with pollutants that could disrupt en ergy flows and impair aquatic health. Trend analysis (concentration trends), load estimation (mass of pollu tants delivered), and the PED index (ecological stress from pollutants) can be used together to provide a more holistic interpretation of water quality, which is necessary for optimizing resources to meet watershed management goals. Objective 3 Evaluate land use-water quality relationships in the Biscayne Bay watershed from 1995 to 2004 at eight water quality monitoring site s considering sub-basins, canal buffers, and site buffers. Specific objectives : (1) quantify human disturbanc e indicators; (2) estimate nutrient loads at monitoring sites; (3) de velop and validate multivariate regression models; and (4) determine if disturbance indi cators within sub-basins, canal buffers, or site buffers explain more of the variability in nutrient loads at monitoring sites. The LDI index and metrics at the landscape le vel (largest patch i ndex [LPI]) and class level (LPI for MSR class) were identified as land use variables with the strongest relationships to estimated lo ads from monitoring sites. The sub-basin regression model wa s the best predictor of annual NOX-N loads in the watershed and included both LDI and LPI variab les. The types of land use (indicated by LDI) and their relative dominan ce (indicated by LPI) influence NOX-N loads. TP loads were more closely related to human disturbance indicators at a smaller spatial extent (1000 m canal buffer), which is a func tion of nutrient trans port processes in the watershed. The land use variable included in the 1000 m canal buffer (MSR LPI) model suggests that urban development patterns in th is buffer zone are important factors for TP loads discharged from the watershed. Additional phosphorus inputs to the bay concern watershed ma nagers because it is the primary nutrient limiting autotrophic growth. Watershed management plans focused on development patterns within canal (1000 m) buffers could therefore potentially reduce phosphorus discharges to the bay. Results from this study suggest that LDI valu es can be included as one indicator in an overall assessment of water quality. The LDI index, which quantifies the intensity of land use activities within watersheds, can be used with landscape metrics that evaluate spatial patterns to link land use developmen t to water quality parameters. Research Synthesis Anthropogenic activity in south Florida, as well as other coastal communities, is continually changing, with dramatic population growth and land use development producing

PAGE 162

162 temporal changes in water quality that can alter ecosystem functionality in receiving waters such as Biscayne Bay. After analyzing land use and nutrient water quali ty data in the Biscayne Bay watershed, research results indicated that the inte nsity of land uses, nutri ent transport processes, and the distribution of land uses close to su rface water conveyance systems were the most influential factors affecting nutrient water quality during the study period. The process used in this study to identify im portant watershed factors influencing water quality variability can be applied to other aquati c ecosystems and watersheds (Figure 5-1). First, identifying possible areas of influence (e.g., sub-ba sins and buffer zones) enables land use data within specific locations to be evaluated relati ve to associated water quality data. Analysis of land use within defined areas includes quantifying the intensity and spatial distribution of land uses. The LDI index and percent imperviousne ss are examples of indicators providing information on the intensity of land uses while landscape metrics reveal the composition and configuration of land uses. Togeth er, these indicators describe la nd use variability within defined areas. Linking land use characteristics to water qualit y data also requires th at land use influences are hydrologically isolated. Methods of isolating land use include using buffers that are upstream of water quality monitoring sites. Trend analys is, load estimation, and the PED index are all examples of analytical methods that describe water quality variability at monitoring sites. Linking data describing land use and water qualit y variability through pr ocedures such as stepwise regressions reveals important factor s influencing water quali ty. Figure 5-1 includes examples of different indicators and methods a nd is intended to be a general guideline for evaluating land use-water quality relationships, not an exhaustiv e description of all possible procedures. Thus, Figure 5-1 pr ovides a starting point for eval uating land use-water quality relationships and would be modifi ed depending on data availability and management objectives.

PAGE 163

163 Aquatic ecosystems and watersheds have specific natural and anth ropogenic influences that combine to create unique systems with ch aracteristic features. Hydrological connectivity, soils, pollutant sources, land use variability, and management pract ices are all factors that can influence nutrient discharges from watersheds. Evaluating particular systems using the overall process previously outlined (Figure 5-1) will likely reveal influen tial factors that contrast with results from the Biscayne Bay watershed. Identify ing and addressing local factors contributing to water quality variability is a key aspect of effective watershed ma nagement strategies that can ultimately protect vulnerable water resources.

PAGE 164

164 Figure 5-1. Flow chart illustrating an overall process that can be used to evalua te land use-water quality relationships in watersheds. Landscape Development Intensity Index Impervious Surfaces Intensity of Land Use Area(s) of Influence Land Use Data Configuration and Composition Land Use Variability Land UseWater Quality Relationships Intensity of Land Use Spatial Distribution Nutrient Transport Processes Pollutant Empower Density Index Constituent Loads Trend Analysis Hydrologic Isolation Area(s) of Influence Water Quality Data Water Quality Variability Landscape Metrics Stepwise Regressions

PAGE 165

165 LIST OF REFERENCES Abt K. 1967. Significant independent variable s in linear models. Metrika 12: 2-15. Abbott E., Alleman R., Barnes T., Bauer M., Be nnett R., Chamberlain R., Coley T., Conboy T., Doering P., Drum D., Goldstein A., Gostel M., Hedgepeth M., Hu G., Hunt M., Kelly S., Madden C., McDonald A., Robbins R., Rudnick D., Stone T., Walker P., and Wan Y. 2005. Management and Restoration of Coas tal Ecosystems. Chapter 12, 2005 South Florida Environmental Report. South Fl orida Water Management District, 110 p. Alberti M. 2005. The effects of urban patterns on ecosystem function. Interna tional Regional Science Review 28: 168-192. Alberti M., Booth D., Hill K., C oburn B., Avolio C., Coe S. and Spirandelli D. 2007. The impact of urban patterns on aquatic ecosystems: An empirical analysis in Puget lowland subbasins. Landscape and Urban Planning 80: 345-361. Alleman R. A. 1990. Surface water quality in the vi cinity of Black Point, Dade County, Florida. Miami Dade Department of Environmental Management, Miami, FL, USA. Technical Report 90-14. Alleman R.W., Mulliken J.D., VanArman J.A., and Burns D.S. 1995. Biscayne Bay surface water improvement and management: technica l supporting document. South Florida Water Management District Planning Depa rtment, West Palm Beach, Fla. Alley W.M. 1988. Using exogenous variables in te sting for monotonic trends in hydrologic time series. Water Resour. Res. 24. Alley W.M. and Veenhuis J.E. 1983. Effective impervious area in urban runoff modeling. Journal of Hydraulic En gineering 109: 313-319. Arnold C.L. and Gibbons C.J. 1996. Impervious surface coverage. Journal of the American Planning Association 62: 16p. Bastianoni S. 1998. A definition of `pollution' based on thermodynamic goal functions. Ecological Modell ing 113: 163-166. BBPI (Biscayne Bay Partnership Initiative). 2001. Biscayne Bay Partnership Initiative: survey team final reports. Biscayne Bay Foundation and FAU/FIU Joint Center for Environment and Urban Problems, Florida, pp. 320. Beck M.B. 1987. Water quality modeling: a review of the analysis of uncertainty. Water Resour. Res. 23. Bengrane K. and Marhaba T.F. 2003. Using principa l component analysis to monitor spatial and temporal changes in water quality. J ournal of Hazardous Materials 100: 179-195.

PAGE 166

166 Bolliger J., Wagner H. and Turner M. 2007. Iden tifying and quantifying landscape patterns in space and time. In: Kienast, F., Wildi, O., and Ghosh, S. (eds). A changing world: challenges for landscape research. Spri nger, Dordrecht, The Netherlands, pp. 177. Bolstad P.V. and Swank W.T. 1997. Cumulative impacts of land use on water quality in a southern Appalachian watershed. Journal of the American Water Resources Association 33: 519-533. Booth D.B. and Jackson C.R. 1997. Urbanization of aquatic systems: degradation thresholds, stormwater detection, and the limits of mitigation. Journal of the American Water Resources Association 33: 1077-1090. Brabec E.A. 2009. Imperviousness and land-use po licy: toward an effective approach to watershed planning. Journal of H ydrologic Engineering 14: 425-433. Brabec E., Schulte S. and Richards P.L. 2002. Imperv ious surfaces and water quality: a review of current literature and its implications fo r watershed planning. Journal of Planning Literature 16: 499-514. Brand L. E. 1988. Assessment of plankton resources and their environmental interactions in Biscayne Bay, Florida. Miami-Dade De partment of Environmental Resources Management, Miami, FL, USA. DERM Technical Report 88-1. Brand L.E. 2002. The transport of te rrestrial nutrients to south Flor ida coastal waters. In: Porter J.W., Porter K.G. (eds). The Everglades, Flor ida Bay, and coral reefs of the Florida Keys: an ecosystem sourcebook. CRC Press, Boca Raton, pp 361-413. Browder J.A., Alleman R., Markley S., Ortner P. and Pitts P.A. 2005. Biscayne Bay conceptual ecological model. Wetlands 25: 854-869. Brown M.T. 1980. Energy Basis for Hierarchie s in Urban and Regional Landscapes. Ph.D. Dissertation. Department of Environmental E ngineering Sciences, University of Florida, Gainesville, 359 pp. Brown M.T. and Bardi E. 2001. Emergy of ecosy stems. Emergy Folio 4 of Handbook of Emergy Evaluation. The Center for Environmental Policy, University of Florida, Gainesville 93 p. Brown M.T. and Ulgiati S. 2004. Energy qualit y, emergy, and transformity: H.T. Odum's contributions to quantifying and understan ding systems. Ecological Modelling 178: 13p. Brown M.T. and Vivas M.B. 2005. Landscape de velopment intensity index. Environmental Monitoring and Assessment 101: 289-309. Brown M.T., Parker N. and Foley A. 1998. Spatia l modeling of landscape development intensity and water quality in the St. Marks River wa tershed. DEP Contract #GW138, Center for Wetlands, University of Flor ida, Gainesville, 136 pp.

PAGE 167

167 Buchanan T. and Klein H. 1976. Effects of wa ter management on fresh-water discharge to Biscayne Bay. In: Biscayne Bay: Past/ Present / Future. Thorhaug, A and Volker, A. (eds). Biscayne Bay Symposia, University of Miami. Burstyn I. 2004. Principal component analysis is a powerful instrument in occupational hygiene inquiries. Ann Occup Hyg 48:655-661. Caccia V.G. and Boyer J.N. 2005. Spatial patterning of water quality in Biscayne Bay, Florida as a function of land use and water manageme nt. Marine Pollution Bulletin 50: 1416-1429. Caccia V.G. and Boyer J.N. 2007. A nutrient load ing budget for Biscayne Bay, Florida. Marine Pollution Bulletin 54: 994-1008. Campbell D.E. 2000. Using energy systems theory to define, measure, and interpret ecological integrity and ecosystem health. Ecosystem Health 6: 24p. Cantillo A.Y., Hale K., Collins E., Pikula L., and Caballero R. 2000. Biscayne Bay: environmental history and annotated bi bliography. NOAA Technical Memorandum. National Status and Trends Program for Mari ne Environmental Quality. Silver Springs, Maryland, USA, 128 pp. Carpenter S.R., Caraco N.F., Correll D.L., Howarth R.W., Sharpley A.N. and Smith V.H. 1998. Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological Applications 8: 559-568. Chadwick M.A., Dobberfuhl D.R., Benke A.C., Huryn A.D., Suberkropp K. and Thiele J.E. 2006. Urbanization affects stream ecosystem function by altering hydrology, chemistry, and biotic richness. Ecological Applications 16: 1796-1807. Champely S. and Doledec S. 1997. How to separate long-term trends from periodic variation in water quality monitoring. Wa ter Research 31: 2849-2857. Cheesman M. 1989. 1986 Intensive canal study: evalua tion of water in the Mowry canal (C-103). Miami-Dade Department of Environmental Resources Management, Miami, FL, USA. DERM Technical Report 89-2. Christensen V.G., Ziegler A.C., Rasmussen P.P. and Geological Survey (U.S.) 2001. Real-time water-quality monitoring and regression anal ysis to estimate nutrient and bacteria concentrations in Kansas streams. U.S. Dept. of the Interior U.S. Geological Survey, Lawrence, Kan. Cifaldi R.L., David Allan J., Duh J.D. and Brow n D.G. 2004. Spatial patterns in land cover of exurbanizing watersheds in southeastern Michigan. Landscape and Urban Planning 66: 107-123.

PAGE 168

168 Clark G.M. 2003. Occurrence and transport of cadmium, lead, and zinc in the Spokane River Basin, Idaho and Washington, water years 1999-2001. National Water-Quality Assessment Program (U.S.) and Geological Survey (U.S.) U.S. Dept. of the Interior Information Services distributor, Boise, Idaho, Denver, CO. Cohn T.A. 1988. Adjusted maximum likelihood estimation of the moments of lognormal populations from type I censored samples. Sc ientific investigati ons report 88-350. U.S. Geological Survey, Reston, Va. Cohn T.A., Gilroy E.J., and Baier W.G. 1992. Estim ating fluvial transport of trace constituents using a regression model with data subject to censoring. Geological Survey (U.S.) and American Statistical Association. Section on St atistics and the Environment. The Survey, Reston, VA. Cohn T.A., Delong L.L., Gilroy E.J., Hirsch R.M. and Wells D.K. 1989. Estimating constituent loads. Water Resources Research 25: 937-942. Collins D. and Odum H.T. 2000. Calculating transformities with the eigenvalue method. In: M.T. Brown (ed). Emergy Synt hesis, Proceeding of the 1st Emergy Research Conference. Center for Environmental Policy. University of Florida, Gainesville FL. pp 265-79. Cushman S.A., McGarigal K. and Neel M.C. 2008. Parsimony in landscape metrics: Strength, universality, and consistency. Ec ological Indicators 8: 13p. Dale V.H. and Beyeler S.C. 2001. Challenges in the development and use of ecological indicators. Ecological Indicators 1: 3-10. Dame R., Alber M., Allen D., Mallin M., Montag ue C., Lewitus A., Chalmers A., Gardner R., Gilman C., Kjerfve B., Pinckney J. and Smith N. 2000. Estuaries of the South Atlantic coast of North America: their geogra phical signatures. Estuaries 23: 793. DERM (Miami-Dade County Department of Environmental Resources Management) and Miami-Dade County Stormwater Utility. 2004. Central Miami-Dade watershed planning stormwater modeling component Phase II. Stor mwater master plan, C-4 Basin, Phase II, Volume 1, CDM, Contract No. E01-DERM01. Donato M.M. and MacCoy D.E. 2005. Phosphorus and suspended sediment load estimates for the lower Boise River, Idaho, 1994-2002. Scient ific investigations report 2004-5235. U.S. Geological Survey, Reston, Va. Dow K. 2000. Social dimensions of gradients in urban ecosystems. Urban Ecosystems 4: 255275. Duan N. 1983. Smearing estimate A nonparametr ic retransformation method. Journal of the American Statistical Association 78: 605.

PAGE 169

169 Efroymson M. A. l960. Multiple regression an alysis. Mathematical Methods for Digital Computers. Ralston A. and Wilf, H. S., (eds.). Wiley, New York. Elkie P., Rempel R., and Carr A. 1999. Patch Anal yst user's manual. TM-002. Ontario Ministry of Natural Resources. Northwest Scie nce & Technology, Thunder Bay, Ontario. ESRI Inc. 2005. ArcGIS 9.1. ESRI Inc., Redlands, California. FDEP (Florida Department of Environmenta l Protection). 2009. Stat e of Florida Numeric Nutrient Criteria Development Plan. Bur eau of Assessment and Restoration Support, Division of Environmental Assessment a nd Restoration, Florida Department of Environmental Protec tion, Tallahassee, FL. FDOT (Florida Department of Transportation). 1999. Florida Land Use and Cover and Forms Classification System: handbook. Third Editi on. Thematic Mapping Section, Surveying and Mapping Office, Tallahassee, FL, USA. Fend T., Hoffschmidt B., Jorgense n G. J., Kster H., Krger D., Pitz-Paal R., Rietbrock P., and Riffelmann K.J., 2003. Comparative assessment of solar concentrator materials. Sol. Energy 74: 149-155. Finkl C.W. and Charlier R.H. 2003. Sustainability of subtropical coastal zones in southeastern florida: challenges for urbanized coastal environments threatened by development, pollution, water supply, and storm hazards. Journal of Coastal Research 19: 934-943. Forman R.T.T. and Godron M. 1986. Landscape Ecology. Wiley, New York. Gergel S.E., Turner M.G., Miller J.R., Melack J.M. and Stanley E.H. 2002. Landscape indicators of human impacts to riverine systems. Aquatic Sciences Research Across Boundaries 64: 118-128. Graham M.H. 2003. Confronting multicollinearity in ecological multiple regression. Ecology 84: 2809-2815. Griffin D.M., Jr., Grizzard T.J., Randall C.W., Hels el D.R. and Hartigan J.P. 1980. Analysis of non-point pollution export from small catchments. Journal (Water Pollution Control Federation) 52: 780-790. Griffith J.A., Martinko E.A. and Price K.P. 2000. La ndscape structure analysis of Kansas at three scales. Landscape and Urban Planning 52: 45-61. Gupta H.V., Sorooshian S. and Yapo P.O. 1999. Status of automatic calibration for hydrologic models: comparison with multilevel expert cal ibration. Journal of Hydrologic Engineering 4: 135-143.

PAGE 170

170 Haines-Young R., Green D.R., and Cousins S. 1994. Landscape ecology and geographical information systems. Haines-Young, R., Green, D.R. and Cousins S. (eds). Bristol: Taylor & Francis, Inc., pp 3-8. Hanna A.J. and Hanna K.A. 1948. Lake Okeechob ee, wellspring of the Everglades. BobbsMerrill, Indianapolis, Indiana, USA. Harding J.S. and Benfield E.F. 1998. Stream biodi versity: the ghost of land use past. Proceedings of the National Academy of Sciences of the United States of America 95: 5p. Harman-Fetcho J.A., Hapeman C.J., McConnell L.L., Potter T.L., Rice C.P., Sadeghi A.M., Smith R.D., Bialek K., Sefton K.A., Schaffer B.A. and Curry R. 2005. Pesticide occurrence in selected South Florida canal s and Biscayne Bay during high agricultural activity. Journal of Agricultural and Food Chemistry 53: 6040-6048. Harmel R.D., Cooper R.J., Slade R.M., Haney R.L. and Arnold J.G. 2006. Cumulative uncertainty in measured streamflow and water quality data for small watersheds. Transactions of the ASAE 49: 13p. Hecky R.E. and Kilham P. 1988. Nutrient limitati on of phytoplankton in fr eshwater and marine environments: a review of recent evidence on the effects of enrichment. Limnology and Oceanography 33: 796-822. Herold M., Goldstein N.C. and Clarke K.C. 2003. The spatiotemporal form of urban growth: measurement, analysis and modeling. Re mote Sensing of Environment 86: 286-302. Hirsch R.M., Alexander R.B. and Smith R.A. 1991. Selection of methods for the detection and estimation of trends in water quality. Water Resour. Res. 27. Hirsch R.M., Slack J.R. and Smith R.A. 1982. T echniques of trend analysis for monthly water quality data. Water Resour. Res. 18. Hocking R.R. 1976. A Biometrics Invited Paper. The analysis and selection of variables in linear regression. Biometrics 32: 1-49. Howarth R., Sharpley A. and Walker D. 2002. Sources of nutrient pollution to coastal waters in the United States: implications for achieving coastal water quality goals. Estuaries and Coasts 25: 656-676. Hunsaker C.T. and Levine D.A. 1995. Hierarchical approaches to the stu dy of water-quality in rivers. Bioscience 45: 193-203. Insightful Corporation Inc. 2007. S-Plus 8.0. Seattle, WA. Iverson L.R. 1988. Land-use changes in Illinois, USA: The influence of landscape attributes on current and historic land use. Landscape Ecology 2: 45-61.

PAGE 171

171 Johnson G.D., Myers W.L. and Patil G.P. 2001. Pred ictability of surface water pollution loading in Pennsylvania using watershe d-based landscape measurements. Journal of the American Water Resources Association 37: 821-835. Johnson L., Richards C., Host G. and Arthur J. 1997. Landscape influences on water chemistry in midwestern stream ecosystems. Freshwater Biology 37:193. Johnston J. 1972. Econometric Methods, (2nd ed.), McGraw-Hill, New York. Jones K.B., Neale A.C., Nash M.S., Van Re mortel R.D., Wickham J.D. 2001. Predicting nutrient and sediment loadings to streams from landscape metrics: a multiple watershed study from the United States mid-At lantic region. Landscape Ecology 16:301. Judge G.G., Hill R.C., Griffiths W.E., Ltke pohl H. and Lee T.C. 1988. Introduction to the theory and practice of econome trics. J. Wiley, New York. Kearns F.R., Kelly N.M., Carter J.L. and Re sh V.H. 2005. A method for the use of landscape metrics in freshwater research and management. Landscape Ecology 20: 113-125. King R.S., Baker M.E., Whigham D.F., Weller D.E ., Jordan T.E., Kazyak P.F. and Hurd M.K. 2005. Spatial considerations for linking waters hed land cover to ecological indicators in streams. Ecological Applications 15: 137-153. Kleppel G. S. 1996. The state of Florida's estuarie s and future needs in estuarine research, 43, 51, Florida Sea Grant College Program, Universi ty of Florida, Gainesville, Florida. Krause P., Boyle D.P. and Bse F. 2005. Comparison of different efficiency criteria for hydrological model assessment Adv. Geosci. 5: 89-97. Lane C. and Brown M. 2006. Energy-based land use predictors of proxima l factors and benthic diatom composition in Florida freshwater marshes. Environmental Monitoring and Assessment 117: 433-450. Langevin C.D. 2001. Simulation of ground-water discharge to Biscayne Bay, southeastern Florida. Place-Based Studies Program (Geologi cal Survey), United States Army Corps of Engineers and Geological Survey (U.S.), U.S. De pt. of the Interior Branch of Information Services distributor, Tallahassee, Fla. Lausch A. and Thulke H. 2001. The analysis of spatio-temporal dynamics of landscape Structures. In: Landscape balance and landscap e assessment. Kronert, R., Steinhardt, U., and Volk, M. (eds). Springer, Berlin, Germany, pp. 113. Leach S.D., Klein H., Hampton E.R. and Geologi cal Survey (U.S.) 1972. Hydrologic effects of water control and management of southeastern Florida. State of Florida Bureau of Geology, Tallahassee, Fla.

PAGE 172

172 Lee S.W., Hwang S.J., Lee S.B., Hwang H. S. and Sung H.C. 2009. Landscape ecological approach to the relationships of land use patterns in watershe ds to water quality characteristics. Landscape and Urban Planning 92: 80-89. Lee T. and Rooth C. 1976. Circulation and excha nge processes in southeast Floridas coastal lagoons. In: Biscayne Bay: Past/ Present / Fu ture. Thorhaug, A and Volker, A. (eds). Biscayne Bay Symposia, University of Miami. Legates D.R. and McCabe G.J., Jr. 1999. Evaluating the use of goodness-of-fit measures in hydrologic and hydroclimatic model va lidation. Water Resour. Res. 35. Lettenmaier D.P., Hooper E.R., Wagoner C. and Fa ris K.B. 1991. Trends in stream quality in the continental United States, 1978. Water Resour. Res. 27. Li H. and Wu J. 2004. Use and misuse of landscape indices. La ndscape Ecology 19: 389-399. Li Y. 2001. Calcareous soils in Miami-Dade County. University of Florida Cooperative Extension Service Institute of Food and Agri culture Sciences EDIS, Gainesville, Fla. Li Y. and Zhang M. 2002. Major problems aff ecting agriculture in Miami-Dade: Natural resources and crop management (fertilizer). In: Miami-Dade County Agricultural Land Retention Study: Final Report, Florida Departme nt of Agricultural and Consumer Services Contract No. 5218. Degner, R. L., Stevens III, T. J. and Morgan K. L. (eds). Florida Agricultural Market Research Center, Institute of Food and Agricultural Sciences, University of Florida, Gainesville. Lietz A.C.1999. Methodology for estimating nutrient lo ads discharged from the east coast canals to Biscayne Bay, Miami-Dade County, Florid a. U.S. Geological Survey, Tallahassee FL. Lietz A.C. 2000. Analysis of water-quality trends at two discharge stat ions one within Big Cypress National Preserve and one near Bisc ayne Bay southern Florida, 1966-94. U.S. Dept. of the Interior, U.S. Geol ogical Survey, Tallahassee, Fla. Lotka A. J. 1922. A contribution to the energetic s of evolution. Proceedings of the National Academy of Sciences 8: 140. Mack J. 2006. Landscape as a predictor of wetland condition: an evaluation of the Landscape Development Index (LDI) with a large reference wetland dataset from Ohio. Environmental Monitoring and Assessment 120: 221-241. Mantel N. 1970. Why stepdown procedures in variable selection. T echnometrics 12: 621-625. Mason R. L., Webster J. T., and Gunst R. F. 1975. Sources of multicollinearity in regression analysis. Comm. in Statist. 4, 277-92.

PAGE 173

173 McCartney M.P., Stratford C., Neal C., Bradford R., Mills S. and Johnson M. 2003. Seasonality and water quality trends in a maturing recreated reed bed. The Science of the total environment 314-316: 233-254. McClanahan T. 1990. Hierarchical Control of Coral Reef Ecosystems. PhD Dissertation, University of Florida. McCuen R.H., Knight Z. and Cutter A.G. 2006. Ev aluation of the Nash--Sutcliffe efficiency index. Journal of Hydrologic Engineering 11: 597-602. McDonald R.P. 1985. Factor an alysis and related methods. Lawrence Erlbaum Associates, Hillsdale, NJ. McDonnell M.J. and Pickett S.T.A. 1997. Introduction: scope and need for an ecology of subtle human effects and populated areas. In McDonnell M.J. and Pickett S.T.A (eds.). Humans as components of ecosystems. Springer, New York, New York, USA, pp. 1. McDonnell M.J., Pickett S.T.A., Groffman P., Bo hlen P., Pouyat R.V., Zipperer W.C., Parmelee R.W., Carreiro M.M. and Medley K. 1997. Ec osystem processes along an urban-to-rural gradient. Urban Ecosystems 1: 21-36. McGarigal K. and Marks B.J. 1995. FRAGSTATS: Spatial Pattern An alysis Program for Quantifying Landscape Structure. Gen. Tec h. Rep. PNW-GTR-351; Pacific Northwest Research Station, USDA-Forest Service, Portland. McKenzie D. J. 1983. Water quality at and ad jacent to the south Da de County solid-waste disposal facility, Florida. Water resources investigations rep. 83-4003. US Geological Survey, Tallahassee, FL. 37 pp. McMahon G. and Cuffney T.F. 2000. Quantifying ur ban intensity in drainage basins for assessing stream ecological conditions. Journal of the American Water Resources Association 36: 1247-1261. McNeill D.F. 2000. A review of upward migration of effluent related to subsurface injection at MiamiDade Water and Sewer South District Pl ant, report prepared for the Sierra Club, Miami Group, 30 pp. McNulty J.K. 1970. Effects of abatement of domestic sewage pollution on the benthos, zooplankton and the fouling organisms of Bi scayne Bay, Florida. University of Miami Press, Coral Gables, FL, 107 pp. McPherson B.F. and Halley R. 1996. The South Florida environment a region under stress. Circular 1134, U.S. Geological Su rvey, Denver, Colorado, USA. 61 pp.

PAGE 174

174 Meeder J. and Boyer J.N. 2001. Total Ammonia Concen trations in soil, sediments, surface water, and groundwater along the western shoreline of Biscayne Bay with the focus on Black Point and a reference mangrove site. Final Re port to the Biscayne National Park Service BISC-N-011000, Florida Internati onal University, Miami, FL. Mehaffey M.H., Nash M.S., Wade T.G., Eber t D.W., Jones K.B. and Rager A. 2005. Linking land cover and water quality in New York C itys water supply watersheds. Environmental Monitoring and Assessment 107: 29-44. Migliaccio K.W., Haggard B.E., Chaubey I. and Matlock M.D. 2007. Linking watershed subbasin characteristics to water quality parameters in War Eagle Creek watershed. Transactions of the ASAE 50: 2007-2016. Moriasi D.N., Arnold J.G., Van Liew M.W., Bi ngner R.L., Harmel R.D. and Veith T.L. 2007. Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Transactions of the ASAE 50: 16p. Nash J.E. and Sutcliffe J.V. 1970. River flow fo recasting through conceptual models. Journal of Hydrology 10: 282. Newcombe K. 1977. Nutrient flow in a major urban settlement Hong-Kong. Human Ecology 5: 179-208. Nichols S. 1977. On the interpretation of princi pal components analysis in ecological contexts. Vegetatio 34: 191-197. Odum H.T. 1971. Environment, power, and so ciety. Wiley-Intersci ence, New York, NY. Odum H.T. 1996. Environmental accounting: EMERGY and environmen tal decision making. Wiley, New York. O'Neill R.V., Hunsaker C.T., Jones K.B., R iitters K.H., Wickham J. D., Schwartz P.M., Goodman I.A., Jackson B.L. and Baillargeon W.S. 1997. Monitoring environmental quality at the landscape scale. Bioscience 47: 513-519. O'Neill R.V., Krummel J.R., Gardner R.H., Sugi hara G., Jackson B., DeAngelis D.L., Milne B.T., Turner M.G., Zygmunt B., Christensen S.W., Dale V.H. and Graham R.L. 1988. Indices of landscape pattern. Landscape Ecology 1: 153-162. Orth P.G., 1976. Nutrient fluctuations in groundwater unde r an agricultural area, Dade County, Florida. Soil and Crop Science Society of Florida, Vol 35 pp. 117-121. Osborne L. and Wiley M. 1988. Empirical relationships between land use/cover patterns and stream water quality in an agricultural wate rshed. Journal of Environmental Management 26: 9.

PAGE 175

175 Parker G. G., Ferguson G.E., and Love S.K. 1955. Water resources of s outhern Florida with special reference to the geology and groundwater of the Miami area. United States Geological Survey, Washington, DC USA. Water Supply Paper 1255. Paul J.F., Comeleo R.L. and Copeland J. 2002. Landscape metrics and estuarine sediment contamination in the Mid-Atlantic and S outhern New England regions. Journal of Environmental Quality 31: 836-845. Powell J.L. 1984. Least absolute deviations es timation for the censored regression model. Journal of Econometrics 25: 303. Qian Y., Migliaccio K.W., Wan Y., Li Y.C. a nd Chin D. 2007. Seasonality of selected surface water constituents in the Indian River Lagoon, Florida. Journal of Environmental Quality 36: 416-425. Radell M.J. and Katz B.G. 1991. Major-ion and sele cted trace-metal chemis try of the Biscayne aquifer, southeast Florida. U.S. Geological Survey WRIR 91-4009, 18 p. Reiss K.C. 2006. Florida Wetland Condition I ndex for depressional forested wetlands. Ecological Indicators 6: 16p. Reiss K. and Brown M. 2007. Evaluation of Florid a Palustrine Wetlands: Application of USEPA Levels 1, 2, and 3 Assessment Methods. EcoHealth 4: 206-218. Reiss K., Brown M. and Lane C. 2009. Characteristic community structure of Floridas subtropical wetlands: the Flor ida wetland condition index for depressional marshes, depressional forested, and flowing water forested wetlands. Wetlands Ecology and Management. Renken R.A. 2005. Impact of anthropogenic deve lopment on coastal ground-water hydrology in southeastern Florida, 1900-2000. U.S. Geological Survey Greater Everglades Priority Ecosystems Science Program. U.S. Dept. of the Interior U.S. Geological Survey, Reston, Va. Risser P.G., Karr J.R. and Forman R.T.T. 1984. Landscape ecology: directions and approaches. Illinois Natural History Survey Special P ubl. No. 2. Illinois Natural History Survey, Champaign, Illinois. Ritters K.H., O'Neill R.V., Hunsaker C.T., Wi ckham J.D., Yankee D.H., Timmins S.P., Jones K.B. and Jackson B.L. 1995. A factor analysis of landscape pattern and structure metrics. Landscape Ecology 10: 23-39. Roth N., Allan J. and Erickson D. 1996. Landscape influences on stream biotic integrity assessed at multiple spatial scales. Landscape Ecology 11: 141-156.

PAGE 176

176 Roy A.H. and Shuster W.D. 2009. Assessing imperv ious surface connectivity and applications for watershed management. JAWRA Journal of the American Water Resources Association 45: 198-209. Royston J.P. 1983. Some techniques for assessing multivariate normality based on the ShapiroWilk W. Journal of the Royal Statistical Society. Series C (Applied Statistics) 32: 121-133. Runkel R.L., Crawford C.G., and Cohn T.A. 2004. Load estimator (LOADEST) a FORTRAN program for estimating constituent loads in st reams and rivers. Techniques and methods bk. 4, Chapter A5. U.S. Dept. of the Interi or U.S. Geological Survey, Reston, Va. Salley G.H. 1986. A history of the Florida suga r industry. G.H. Salley, Miami, Florida, USA. Available from the Florida Sugarcane League. Scheidt D.J. and Flora M.D. 1983. Mowry Canal (C-103)--water quality and discharge into Biscayne Bay, Florida, 1975-1981. National Park Service South Florida Research Center Everglades National Pa rk, Homestead, Fla. Schertz T.L., Alexander R.B. and Ohe D.J. 1991. The computer program estimate trend (ESTREND), a system for the de tection of trends in water-quality data. United States Geological Survey, Water-Resources Invest igations Report 91-4040. Reston, VA. 70pp. Schiff R. and Benoit G. 2007. Effects of impervious cover at multiple spa tial scales on coastal watershed streams. JAWRA Journal of the American Water Resources Association 43: 712-730. Schultz R. 1991. Population growth and migration: Southeast Fl orida in regional context. In: South Florida: the winds of change. Prep ared for the Annual Conference of the Association of American Geographers pp. 43-61, Miami, Florida, USA. SFWMD (South Florida Water Management District). 1995. Land cover land use 1995. SFWMD, West Palm Beach, FL. SFWMD (South Florida Water Management District). 1999. Land cover land use 1995. SFWMD, West Palm Beach, FL. SFWMD (South Florida Water Management District). 2004. Land cover land use 1995. SFWMD, West Palm Beach, FL. SFWMD (South Florida Wate r Management District). 2009. GIS Data Catalog: http://my.sfwmd.gov/gisapps/sfwmdxwebdc/dat aview.asp. SFWMD, West Palm Beach, FL. Shapiro S.S. and Wilk M.B. 1965. An analysis of variance test for normality (complete samples). Biometrika 52: 591-611.

PAGE 177

177 Sharpley A.N., Kleinman P.J.A., McDowell R. W., Gitau M. and Bryant R.B. 2002. Modeling phosphorus transport in agricultural watersheds: processes and possibili ties. Journal of Soil and Water Conservation 57: 425(15). Schreiber K. F. 1990. The history of landscape eco logy in Europe. In: Changing landscapes: an ecological perspective. Zonneveld, I. and Fo rman, R.T.T.T. (eds). Pages 21 in, editors. Springer-Verlag, New York, New York, USA. Sliva L. and Dudley Williams D. 2001. Buffer zone versus whole catchment approaches to studying land use impact on river wate r quality. Water Research 35: 3462-3472. Smith V. H. 1998. Cultural eutrophication of inland, es tuarine, and coastal waters. In: M. L. Pace and P. M. Groffman, editors. Successes, limitations, and frontiers in ecosystem science. Springer-Verlag, New York, New York, USA. Snee R.D. 1977. Validation of regression models : methods and examples. Technometrics 19: 415-428. Solecki W. and Walker R. 2001. Transformation of the South Florida landscape. In Growing Populations, Changing Landscapes : Studies from India, China, and the United States, pp. 237-274, National Academy Press, Washington D.C., USA. Sponseller R.A., Benfield E.F. and Valett H.M. 2001. Relationships between land use, spatial scale and stream macroinvertebrate communities. Freshwater Biology 46: 16p. Stankowski Stephen J. 1972. Populat ion density as an indirect indicator or urban and suburban land-surface modifications. U.S. Geological Survey Profession al Paper 800-B. Washington, DC: U.S. Geological Survey. Strayer D.L., Beighley R.E., Thompson L.C., Bro oks S., Nilsson C., Pinay G. and Naiman R.J. 2003. Effects of land cover on stream ecosystems: roles of empirical models and scaling issues. Ecosystems 6: 17p. Thompson M.L. 1978. Selection of variables in multiple regression: Part I. A review and evaluation. International Statistical Review / Revue Intern ationale de Statistique 46: 1-19. Turner M.G. 1989. Landscape ecology the effect of pattern on process. Annual Review of Ecology and Systematics 20: 171-197. Turner M.G. 1990. Spatial and temporal analysis of landscape patterns. Landscape Ecology 4: 21-30. Turner M.G. 2005. Landscape ecology in North America: past, present, and future. Ecology 86: 8p.

PAGE 178

178 Turner M.G. and Gardner R.H. 1991. Quantitative methods in landscape ecology: the analysis and interpretation of la ndscape heterogeneity. Sp ringer-Verlag, New York. USACE and SFWMD. 1999. Central and southern Florida project comprehensive review study final integrated feasibility report and progr ammatic environmental impact statement. United States Army Corps of Engineers, Jack sonville District, FL, USA and South Florida Water Management District, West Palm Beach, FL, USA. USEPA (United States Environmental Protecti on Agency). Office of Water, United States. Environmental Protection Agency. Office of Science and Technology. 2000. Nutrient criteria technical guidance manual rivers and streams / United States Environmental Protection Agency, Office of Water. Walker R.T., Solecki W.D. and Harwell C. 1997. Land use dynamics and ecological transition: the case of South Florida. Urban Ecosystems 1: 37-47. Wang L., Brenden T., Seelbach P., Cooper A., A llan D., Clark R. and Wiley M. 2008. Landscape based identification of human disturbance grad ients and reference conditions for Michigan streams. Environmental Monito ring and Assessment 144: 483-484. Wanless H. R. 1976. Geologic setting and recent sedi ments of the Biscayne Bay region, Florida. In: Biscayne Bay: Past / Present / Future. A. Thorhaug, and A. Volker, (eds.). Biscayne Bay Symp. University of Miami Sea Grant Pr ogram Spec. Rep. 5. University of Miami, Coral Gables, FL. 1-31. Weng Y.C. 2007. Spatiotemporal changes of land scape pattern in response to urbanization. Landscape and Urban Planning 81: 341-353. Winsberg M. 1991. South Florida Agriculture. In S outh Florida: the winds of change. Prepared for the Annual Conference of the Association of American Geographers, pp. 17-30, Miami, Florida, USA. Wright L. T., and Heaney, J. P. 2001. Design of distributed stormwater control and reuse systems. Stormwater collection systems design handbook. Mays, L.W. (ed.), McGraw-Hill, New York.

PAGE 179

179 BIOGRAPHICAL SKETCH Richard O. Carey was born in Jamaica. As a sm all child, he exhibited a natural tendency to lead, had a keen sense of understanding, and thus was regularly assigned l eadership roles within his age cohort. He excelled both academically and athletically during his preparatory school years, receiving awards for his endeavors. Successful in his high school entrance examination, he was awarded a place at one of Jamaicas most prestigious institutions Wolmers Boys School in the capital city of Kingston. Richard participated in several extra-curricular activities while in high school, including playing for Wolmers football teams. In his quest for higher educa tion after graduating from high school, Richard chose to attend the University of Miami, where he read for a Bachelor of Science degr ee in Biology. Richards curiosity and love for nature then led him to the University of Georgia, where he earned a Master of Science degree in Conservation Ecology and Su stainable Development from the Odum School of Ecology. The University of Floridas Interdisciplinary Ecol ogy Ph.D. program at the School of Natural Resources and Environment allowed him to explore his inte rests further as he investigated anthropogenic influences affecting natural resources. After receiving his Ph.D., he intends to de velop research programs evaluating ecological stressors in urbanized and degr aded areas while providing opportunities for others within his professional capacity.