<%BANNER%>

Projecting Regional Climate Change in Florida via GIS-Based Downscaling of a General Circulation Model

Permanent Link: http://ufdc.ufl.edu/UFE0042070/00001

Material Information

Title: Projecting Regional Climate Change in Florida via GIS-Based Downscaling of a General Circulation Model
Physical Description: 1 online resource (366 p.)
Language: english
Creator: Coenen, Danny
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: ccsm, change, climate, community, conservation, downscaling, ecology, florida, gis, global, interdisciplinary, kriging, model, modeling, precipitation, system, temperature, warming
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: High-resolution projections of temperature and precipitation changes in Florida during January and July through the year 2100 were produced by GIS-based geostatistical downscaling of general circulation model outputs provided by the Community Climate System Model 3.0 (CCSM) for IPCC benchmark scenarios B1, A1B and A2. Calibration methods designed to assess and correct for CCSM biases were evaluated and implemented. Results indicate mean statewide temperature anomalies ranging from +1.32degreeC to +2.64degreeC during January by the end of the century. All scenarios reveal a weakening of the latitudinal climatic gradient during winter. The northwestern panhandle is projected to experience the greatest warming, with anomalies decreasing towards the southeast. July anomalies range from +1.22degreeC to +3.38degreeC with little regional differentiation except for scenario A2, which projects Florida to become more isothermal than is presently the case. Only scenario B1 projects greater warming in January than July. Mean statewide precipitation anomalies for January are near-zero for all scenarios, whereas during July, anomalies range from -21 mm to -42 mm. Panhandle precipitation is expected to remain similar to present conditions or slightly wetter. South Florida is projected to experience highly significant drying, with some areas projected to receive as little as 23.6% of current rainfall under scenario A2. Projected patterns of change strongly suggest increasing temperature stress for temperate and warm-temperate taxa near the southern margin of their distribution, while barriers to northward expansion of subtropical and tropical taxa are reduced due to the decreasing slope of the latitudinal winter temperature gradient. As they expand the northern margin of their range, warm-adapted species are likely to successfully compete with and exploit resources made available by failing temperate and warm-temperate communities. Subtropical and tropical species in south Florida will experience increasing water stress due to sharply reduced summer precipitation, favoring drought-tolerant species in novel ecological communities without present analogs.
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 Danny Coenen.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Crisman, Thomas L.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042070:00001

Permanent Link: http://ufdc.ufl.edu/UFE0042070/00001

Material Information

Title: Projecting Regional Climate Change in Florida via GIS-Based Downscaling of a General Circulation Model
Physical Description: 1 online resource (366 p.)
Language: english
Creator: Coenen, Danny
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: ccsm, change, climate, community, conservation, downscaling, ecology, florida, gis, global, interdisciplinary, kriging, model, modeling, precipitation, system, temperature, warming
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: High-resolution projections of temperature and precipitation changes in Florida during January and July through the year 2100 were produced by GIS-based geostatistical downscaling of general circulation model outputs provided by the Community Climate System Model 3.0 (CCSM) for IPCC benchmark scenarios B1, A1B and A2. Calibration methods designed to assess and correct for CCSM biases were evaluated and implemented. Results indicate mean statewide temperature anomalies ranging from +1.32degreeC to +2.64degreeC during January by the end of the century. All scenarios reveal a weakening of the latitudinal climatic gradient during winter. The northwestern panhandle is projected to experience the greatest warming, with anomalies decreasing towards the southeast. July anomalies range from +1.22degreeC to +3.38degreeC with little regional differentiation except for scenario A2, which projects Florida to become more isothermal than is presently the case. Only scenario B1 projects greater warming in January than July. Mean statewide precipitation anomalies for January are near-zero for all scenarios, whereas during July, anomalies range from -21 mm to -42 mm. Panhandle precipitation is expected to remain similar to present conditions or slightly wetter. South Florida is projected to experience highly significant drying, with some areas projected to receive as little as 23.6% of current rainfall under scenario A2. Projected patterns of change strongly suggest increasing temperature stress for temperate and warm-temperate taxa near the southern margin of their distribution, while barriers to northward expansion of subtropical and tropical taxa are reduced due to the decreasing slope of the latitudinal winter temperature gradient. As they expand the northern margin of their range, warm-adapted species are likely to successfully compete with and exploit resources made available by failing temperate and warm-temperate communities. Subtropical and tropical species in south Florida will experience increasing water stress due to sharply reduced summer precipitation, favoring drought-tolerant species in novel ecological communities without present analogs.
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 Danny Coenen.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Crisman, Thomas L.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042070:00001


This item has the following downloads:


Full Text

PAGE 1

1 PROJECTING REGIONAL CLIMATE CHANGE IN FL ORIDA VIA GIS BASED DOWNSCALING OF A GENERAL CIRCULATION MODEL By DANNY COENEN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Danny Coenen

PAGE 3

3 To Jennifer and Julia, Mama und Papa

PAGE 4

4 ACKNOWLEDGMENTS This project would not have been possible without the support of my committee chair, Dr. Thomas L. Crisman, whose advice helped guide my work from the planning stages of my masters thesis through the final manuscript revisions of my dissertation Dr. Mark Brenner and Dr. William R. Wise s expert knowledge and detailed fee dback were welcome and encouraging. Dr Ed ward J. Phlips willingness to join my committee on short notice was greatly appreciated. Contributions by f ormer committee members Dr. Fernando R. Miralles Wilhelm and Dr. Stephen S. Mulkey are acknowledged. Stanley Latimer and Dr. Paul Zwick were instrumental in my learning of ArcGIS and help ed improve the quality of early models through their advice. The staff responsible for the NCAR GIS Climate Change initiative is recognized for their support, responding swiftly to my frequent inquiries for clarifications and server restarts. The love, patience and encouragement of my wife, Jennifer, and daughter, Julia Skye, are sincerely appreciated. M y sincere thanks and appreciation are extended to my parents, Rosemarie and Franz Peter Coenen whose encouragement, support and love were instrumental to the realization of my academic ambitions. I wish to add a special dedication to my aunt and godmother Inge Mlheims who passed away of cancer shortly before this manuscript w as completed. This research uses data provided by the Community Climate System Model project ( http://www.ccsm.ucar.edu ), supported by the Directorate for Geosciences of the National Science Foundation and the Office o f Biological and Environmental Research of the U.S. Department of Energy. NCAR GIS Initiative provided CCSM data in a GIS format through the GIS Climate Change Scenarios portal ( http://www.gisclimatechange.org ).

PAGE 5

5 TABLE OF CONTENTS ACKNOWLEDGMENTS ...............................................................................................................4 page TABLE OF CONTENTS .................................................................................................................5 LIST OF TABLES ...........................................................................................................................8 LIST OF FIGURES .........................................................................................................................9 ABSTRACT ...................................................................................................................................20 CHAPTER 1 INTRODUCTION ..................................................................................................................22 Background .............................................................................................................................22 Study Overview and Objectives .............................................................................................25 2 CONTEMPORARY GLOBAL CLIMATE CHANGE IN PERSPECTIVE .........................29 Physical Basis and Characteristics of Global Anthropogenic Climate Change .....................29 Climate Change Scenarios ......................................................................................................32 3 INCORPORATING THE REALITY OF CLIMATE CHANGE INTO CONSERVATION AND DEVELOPMENT STRATEGIES ................................................37 Anticipat ed Climate Change Impacts in the 21st Century ......................................................39 Ecological Effects ............................................................................................................39 Socioeconomic Effects ....................................................................................................41 Implications for Conservation and Development ...................................................................43 Climate Change -Integrated Conservation Strategies.......................................................44 Technology Transfer and Market Mechani sms in Support of Biodiversity Conservation ................................................................................................................46 Case Example: Realizing Conservation Objectives via Synergies with Reducing Emissio ns from Deforestation and Degradation (REDD) ...........................................47 Section Summary ....................................................................................................................50 4 METHODS FOR GEOGRAPHIC INFORMATION SYSTEM (GIS) BASED DOWNSCALING OF GENERAL CIRCULATION MODEL OUPUTS TO PRODUCE HIGHRESOLUTION IMPACT MAPS FOR FLORIDA .....................................................52 Literature Precedents for GIS Based and Geostatistical Climate Analyses ...........................53 Modeling Philosophy ..............................................................................................................59 Applicability beyond Florida ..................................................................................................62 Construction of Downscaled Climate Models for Florida ......................................................63

PAGE 6

6 Prerequisites: Input Data and Analysis Software ............................................................63 Step 1: Data Collection and Preparation .........................................................................64 National Clim atic Data Center (NCAR) climate normals ........................................64 Community Climate System Model (CCSM) output data .......................................65 Step 2: Exploratory Analysis of Observed Data..............................................................68 Step 3: Creating the Baseline Reference Model from Observed Data ............................70 Step 4: Analysis of CCSM Output Data ..........................................................................71 Data exploration .......................................................................................................71 Modified Thornton Wilhelmi method to visualize CCSM projected anomalies and test for statistical significance ........................................................................72 Step 5: Creation of Least Root Mean Square Error (RMSE) Interpolations of CCSM Outputs .............................................................................................................74 Step 6: Calibrating CCSM Outputs to the Observed Data: Bias Correction ...................75 Average bias correction ............................................................................................75 Individual bias correction .........................................................................................76 Step 7: Applying Bias Correction to Future Projections .................................................77 Step 8: Deriving Projections for Temperature Minima and Maxima ..............................78 5 CLIMATE CHANGE PROJECTIONS FOR FLORIDA THROUGH 2100 GENERATED BY GIS BASED DOWNSCALING OF A GENERAL CIRCULATION MODEL ..................................................................................................................................93 Literature Descriptions of Late 20th Century Climate of Florida ..........................................93 Surface Air Temperature .................................................................................................95 Precipitation Patterns .......................................................................................................98 Precipitation Regimes ......................................................................................................99 Panhandle and northern peninsula ..........................................................................100 Southern peninsula .................................................................................................100 The Keys .................................................................................................................101 Interpolated Baseline Models Derived From Observed Climate Normals ...........................102 Baseline Models for Observed January Temperatures ..................................................103 January mean temperature baseline model .............................................................103 January minimum temperature baseline model ......................................................105 January maximum temperature baseline model .....................................................107 Baseline Mo dels for Observed July Temperatures........................................................109 July mean temperature baseline model ...................................................................109 July minimum temperature baseline model ............................................................110 July maximum temperature baseline model ...........................................................112 CCSM Bias Assessment for Temperatures ...................................................................113 Baseline Models for Observed January Precipitation ...................................................115 Baseline Models for Observed July Precipitation .........................................................117 CCSM Bias Assessment for Precipitation .....................................................................118 Projections of Climate Change in Florida ............................................................................119 Modified Thornton Wilhelmi Method to Visualize CCSM Projected Temperature Anomalies ..................................................................................................................119 Jan uary ....................................................................................................................120 July .........................................................................................................................122

PAGE 7

7 Projections from Downscaled, Bias Corrected Interpolations ......................................124 January means ........................................................................................................124 January minima and maxima ..................................................................................127 July means ..............................................................................................................128 July minima and maxima ........................................................................................131 Precipitation ...................................................................................................................132 Modified Th ornton Wilhelmi Method to Visualize CCSM Projected Precipitation Anomalies ..................................................................................................................132 January ....................................................................................................................133 July .........................................................................................................................134 Projections from Downscaled, Bias Corrected Interpolations ......................................136 January ....................................................................................................................136 July .........................................................................................................................139 6 LIMITATIONS TO BIO CLIMATIC MODELING ...........................................................324 Scientific Uncertainty ...........................................................................................................324 Model Limitations ................................................................................................................325 Inherent System Properties ...................................................................................................328 7 ECOLOGICAL AND SOCIOECONOMIC IMPACTS ON FLORIDA .............................331 Literature Review of Climate Change Impact Assessments for Florida ..............................331 Magnitude and Patterns of Recent and Future Change .................................................331 Ecological Impacts ........................................................................................................334 Socioeconomic Impacts .................................................................................................341 Implications of Downscaling Results ...................................................................................342 Areas Subject to Greatest Change .................................................................................343 Temperature ............................................................................................................343 Precipitation ............................................................................................................345 Inter Scenario Uncertainties ..........................................................................................347 8 CONCLUSIONS ..................................................................................................................350 Outlook .................................................................................................................................350 Future Directions and Challenges .........................................................................................350 APPENDIX: NAMES AND LOCATIONS OF REFERENCE WEATHER STATIONS .........353 LIST OF REFERENCES .............................................................................................................356 BIOGRAPHICAL SKETCH .......................................................................................................366

PAGE 8

8 LIST OF TABLES Table page 5-1 Projected January mean temperature gradient and mean slope. ......................................197 5-2 Projected January minimum temperature gradient and mean slope. ...............................221 5-3 Projected January maximum temperature gradient and mean slope. ...............................231 5-4 Projected July mean temperature gradient and mean slope. ............................................239 5-5 Projected July minimum temperature gradient and mean slope. .....................................263 5-6 Projected July maximum temperature gradient and mean slope. ....................................273 5-7 Projected January precipitation gradient and mean slope. ...............................................287 5-8 Projected July precipitation gradient and mean slope ......................................................309 A-1 Names and locations of reference weather stations .........................................................353

PAGE 9

9 LIST OF FIGURES Figure page 1-1 Mean global temperature between 1880 and 2009. ...........................................................27 1-2 Results of Coenen (2005), illustrating decadal scale winter temperature variability in Florida since the mid -1970s. ..............................................................................................28 2-1 Projected surface temperature change (relative to 19 80) for benchmark scenarios B1, A1B, A2 and a constant composition scenario............................................35 2-2 CO2 emissions under various scenarios. ............................................................................36 4-1 Location of reference weather stations in Florida, Alabama and Georgia. ........................79 4-2 Results of the Average Nearest Neighbor operation. .........................................................80 4-3 CCSM model grid and grid centroids over the southeastern United States.......................80 4-4 Illustration of differences between CCSM ensemble members. ........................................81 4-5 Histogram of observed 1971-2000 mean January temperature normals. ..........................82 4-6 Voronoi maps illustratin g geostatistical patterns in the data from the 103 reference weather stations ..................................................................................................................83 4-7 Trend analysis of observed 19712000 me an January temperature normals ....................84 4-8 Semivariogram cloud and semivariogram surface of 1971-2000 January temperature normals. ..............................................................................................................................85 4-9 Normal QQ-plot of observed 19712000 mean January temperature normals. .................85 4-10 Histogram of CCSM -generated 1971-2000 mean January temperatures. .........................86 4-11 Normal QQ-plot of CCSM-generated 1971-2000 mean January temperatures.................87 4-12 Voronoi map s illustrating geostatistical patterns in the CCSM data. ................................88 4-13 Trend analysis on the CCSM -generated 19712000 mean January temperatures ............89 4-14 Semivariogram cloud and semivariogram surface of 1971-2000 CCSM outputs. ............90 4-15 Average bias correction workflow summary. ....................................................................91 4-16 Individual b ias c orrection workflow .................................................................................92 5-1 Political map of the State of Florida. ...............................................................................143

PAGE 10

10 5-2 Florida climate classifications. .........................................................................................144 5-3 Beginning of meteorological seasons (month/week) in Florida. .....................................145 5-4 Location of reference meteorological stations in Florida, Alabama and Georgia. ..........146 5-5 Location of the Jasper Tavernier transect and distance from its origin. ..........................147 5-6 Interpolation of observed January mean temperatures 1971-2000. ..............................148 5-7 January mean te mperature profile for 1971-2000............................................................149 5-8 January mean temperature baseline model prediction standard error map. .....................150 5-9 Error statistics and regression function for the January mean temperature baseline model. ...............................................................................................................................151 5-10 QQ -plot for the January mean temperature baseline model. ...........................................151 5-11 Interpolation of observed January minimum temperatures. ............................................152 5-12 January minimum temperature profile for 1971-2000. ....................................................153 5-13 January minimum temperature baseline model prediction standard error map. ..............154 5-14 Error statistics and regression function for the January minimum temperature baseline model..................................................................................................................155 5-15 QQ plot for January minimum temperature baseline model. ..........................................155 5-16 Interpolation of observed January maximum temperatures 1971-2000. ......................156 5-17 January maximum temperature profile for 1971-2000. ...................................................157 5-18 January maximum temperature baseline model prediction standard error map. .............158 5-19 Error statistics and regression function for the January maximum temperature baseline model ..................................................................................................................159 5-20 QQ -plot for January maximum temperature baseline model...........................................159 5-21 Interpolation of observed July mean temperatures 1971-2000. ....................................160 5-22 July mean temperature profile for 1971 -2000..................................................................161 5-23 July mean temperature baseline model prediction standard error map............................162 5-24 Error statistics and regression function for the July mean temperature baseline model. ...............................................................................................................................163

PAGE 11

11 5-25 QQ plot for July mean temperature baseline model. .......................................................163 5-26 Interpolation of observed July minimum temperatures 1971-2000. .............................164 5-27 July minimum temperature profile for 1971 -2000...........................................................165 5-28 July minimum temperature baseline model prediction standard error map. ....................166 5-29 Error statistics and regression function for the Ju ly minimum temperature baseline model. ...............................................................................................................................167 5-30 QQ -plot for July minimum temperature baseline model. ................................................167 5-31 Interpolation of observed July maximum temperatures 1971-2000. ............................168 5-32 July maximum temperature profile for 1971-2000. .........................................................169 5-33 July maximum temperature baseline model prediction standard error map. ...................170 5-34 Error statistics and regression function for the July maximum temperature baseline model. ...............................................................................................................................171 5-35 QQ -plot for July maximum temperature baseline model. ...............................................171 5-36 CCSM bias for January temperatures. .............................................................................172 5-37 CCSM bias for July temperatures. ...................................................................................173 5-38 Interpolation of observed January precipitation 1971-2000. ........................................174 5-39 January precipitation profile for 1971-2000. ...................................................................175 5-40 January precipitation baseline model prediction standard error map. .............................176 5-41 Error statistics and regression function for the Januar y precipitation baseline model. ...177 5-42 QQ -plot for the January precipitation baseline model. ....................................................177 5-43 Interpolation of observed July prec ipitation 1971-2000. ..............................................178 5-44 July precipitation profile for 1971-2000. .........................................................................179 5-45 July precipitation baseline model prediction standard error map. ...................................180 5-46 Error statistics and regression function for the July precipitation baseline model. .........181 5-47 QQ -plot for July precipitation baseline model.................................................................181 5-48 CCSM bias for January precipitation. ..............................................................................182

PAGE 12

12 5-49 C CSM bias for July precipitation ...................................................................................183 5-50 CCSM projected January temperature anomalies and their statistical significance under scenario B1.............................................................................................................184 5-51 CCSM projected January temperature anomalies and their statistical significance under scenario A1B. .........................................................................................................185 5-52 CCSM projected January temperature anomalies and their statistical significance under scenario A2. ...........................................................................................................186 5-53 CCSM projected July temperature anomalies and their statistical significance under scenario B1. ......................................................................................................................187 5-54 CCSM projected July temperature anomalies and their statistical significance under scenario A1B. ...................................................................................................................188 5-55 CCSM -projected July temperature anomalies and their statistical significance under scenario A2 .....................................................................................................................189 5-56 Projected January mean temperature 2020 -2029, Scenario B1. ......................................190 5-57 Projected January mean temperature 2040 -2049, Scenario B1. ......................................191 5-58 Projected January mean temperature 2060 -2069, Scenario B1. ......................................192 5-59 Projected January mean temperature 2090 -2099, Scenario B1. ......................................193 5-60 Anomaly map for January temperatures, comparing 2040-2049 projections under scenario B1 to the 1971-2000 baseline. ...........................................................................194 5-61 Anomaly map for January temperatures, comparing 2090-2099 projections under scenario B1 to the 1971-2000 baseline. ...........................................................................195 5-62 Projected 2040-2049 January mean temperature profile under scenario B1. ..................196 5-63 Projected 20902099 January mean temperature profile under scenario B1. ..................196 5-64 Projected January mean temperature 2020 -2029, Scenario A1B. ...................................198 5-65 Projected January mean temperature 2040 -2049, Scenario A1B. ...................................199 5-66 Projected January mea n temperature 2060-2069, Scenario A1B. ...................................200 5-67 Projected January mean temperature 2090 -2099, Scenario A1B. ...................................201 5-68 Anomaly map for January temperatures, comparing 2040-2049 projections under scenario A1B to the 1971-2000 baseline. ........................................................................202

PAGE 13

13 5-69 Anomaly map for January temperatures, comparing 2090-2099 projections under scenario A1B to the 1971-2000 baseline. ........................................................................203 5-70 Projected 20402049 January mean temperature profile under scenario A1B. ...............204 5-71 Projected 20902099 January mean temperature profile under scenario A1B. ...............204 5-72 Projected January mean temperature 2020 -2029, Scenario A2. ......................................205 5-73 Projected Jan uary mean temperature 2040 -2049, Scenario A2. ......................................206 5-74 Projected January mean temperature 2060 -2069, Scenario A2. ......................................207 5-75 Projected January mean temperature 2090 -2099, Scenario A2. ......................................208 5-76 Anomaly map for January temperatures, comparing 2040-2049 projections under scenario A2 to the 1971-2000 baseline. ...........................................................................209 5-77 Anomaly map for January temperatures, comparing 2090-2099 projections under scenario A2 to the 1971-2000 baseline. ...........................................................................210 5-78 Projected 20402049 January mean temperature profile under scenario A2. ..................211 5-79 Projected 20902099 January mean temperature profile under scenario A2. ..................211 5-80 Projected January minimum temperature 2040 -2049, Scenario B1. ...............................212 5-81 Projected January minimum temperature 2090 -2099, Scenario B1. ...............................213 5-82 Projected January minimum temperature 2040 -2049, Scenario A1B. ............................214 5-83 Projected January minimum temperature 2090 -2099, Scenario A1B. ............................215 5-84 Projected January minimum temperature 2040 -2049, Scenario A2. ...............................216 5-85 Projected January minimum temperature 2090 -2099, Scenario A2. ...............................217 5-86 Projected 2040-2049 January minimum temperature profile under scenario B1. ...........218 5-87 Project ed 20402049 January minimum temperature profile under scenario A1B. ........218 5-88 Projected 20402049 January minimum temperature profile under scenario A2. ...........219 5-89 Projected 2090-2099 January minimum temperature profile under scenario B1. ...........219 5-90 Projected 20902099 January minimum temperature profile under scenario A1B. ........220 5-91 Projected 2090-2099 January minimum temperature profile under scenario A2. ...........220

PAGE 14

14 5-92 Projected January maximum temperature 2040 -2049, Scenario B1................................222 5-93 Projected January maximum temperature 2090 -2099, Scenario B1................................223 5-94 Projected January m aximum temperature 2040-2049, Scenario A1B. ............................224 5-95 Projected January maximum temperature 2090 -2099, Scenario A1B. ............................225 5-96 Projected January maximum temperature 2040 -2049, Scenario A2. ..............................226 5-97 Projected January maximum temperature 2090 -2099, Scenario A2. ..............................227 5-98 Projected 2040-2049 January maximum temperature profile under scenario B1............228 5-99 Projected 2040-2049 January maximum temperature profile under scenario A1B.........228 5-100 Projected 2040-2049 January maximum temperature profile under scenario A2. ..........229 5-101 Projected 2090-2099 January maximum temperature profile under scenario B1............229 5-102 Projected 2090-2099 January maximum temperature profile under scenario A1B. ........230 5-103 Projected 2090-2099 January maximum temperature profile under scenario A2. ..........230 5-104 Projected July mean temperature 2020-2029, Scenario B1. ............................................232 5-105 Projected July mean temperature 2040-2049, Scenario B1. ............................................233 5-106 Projected July mean temperature 2060-2069, Scenario B1. ............................................234 5-107 Projected July mean temperature 2090-2099, Scenario B1. ............................................235 5-108 Anomaly map for July temperatu res, comparing 2040-2049 projections under scenario B1 to the 1971-2000 baseline. ...........................................................................236 5-109 Anomaly map for July temperatures, comparing 2090-2099 projections under scenario B1 to the 1971-2000 baseline. ...........................................................................237 5-110 Projected 2040-2049 July mean temperature pr ofile under scenario B1. ........................238 5-111 Projected 20902099 July mean temperature profile under scenario B1. ........................238 5-112 Projected July mean temperature 2020-2029, Scenario A1B. .........................................240 5-113 Projected July mean temperature 2040-2049, Scenario A1B. .........................................241 5-114 Projected July mean temperature 2060-2069, Scenario A1B. .........................................242 5-115 Projected July mean temper ature 2090-2099, Scenario A1B. .........................................243

PAGE 15

15 5-116 Anomaly map for July temperatures, comparing 2040-2049 projections under scenario A1B to the 1971-2000 baseline. ........................................................................244 5-117 Anomaly map for July temperatures, comparing 2090-2099 projections under scenario A1B to the 1971-2000 baseline. ........................................................................245 5-118 Projected 20402049 July mean temperature profile under scenario A1B. .....................246 5-119 Projected 20902099 July mean temperature profile under scenario A1B. .....................246 5-120 Projected July mean temperature 2020-2029, Scenario A2. ............................................247 5-121 Projected July mean temperature 2040-2049, Scenario A2. ............................................248 5-122 Projected July mean temperature 2060-2069, Scenario A2. ............................................249 5-123 Projected July mean temperature 2090-2099, Scenario A2. ............................................250 5-124 Anomaly map for July temperatures, comparing 2040-2049 projections under scenario A2 t o the 1971-2000 baseline. ...........................................................................251 5-125 Anomaly map for July temperatures, comparing 2090-2099 projections under scenario A2 to the 1971-2000 baseline. ...........................................................................252 5-126 Projected 20402049 July mean temperature profile under scenario A2. ........................253 5-127 Projected 20902099 July mean temperature profile under scenario A2. ........................253 5-128 Projected July minimum temperature 2040 -2049, Scenario B1. .....................................254 5-129 Projected Ju ly minimum temperature 2090 -2099, Scenario B1. .....................................255 5-130 Projected July minimum temperature 2040 -2049, Scenario A1B. ..................................256 5-131 Projected July minimum temperature 2090 -2099, Scenario A1B. ..................................257 5-132 Projected July minimum temperature 2040 -2049, Scenario A2. .....................................258 5-133 Projected July minimum temperature 2090 -2099, Scenario A2. .....................................259 5-134 Projected 2040-2049 July minimum temperature profile under scenario B1. .................260 5-135 Projected 2090-2099 July minimum temperature profile under scenario B1. .................260 5-136 Projected 20402049 July minimum temperature profile under scenario A1B. ..............261 5-137 Projected 20902099 July minimum temperature profile under scenario A1B. ..............261 5-138 Projected 2040-2049 July minimum temperature profile under scenario A2. .................262

PAGE 16

16 5-139 Projected 2090-2099 July minimum temperature profile under scenario A2. .................262 5-140 Projected July maximum temperature 2040 -2049, Scenario B1. ....................................264 5-141 Projected July maximum temperature 2090 -2099, Scenario B1. ....................................265 5-142 Projected July maximum temperature 2040 -2049, Scenario A1B...................................266 5-143 Projected July maximum temperature 2090-2099, Scenario A1B...................................267 5-144 Projected July maximum temperature 2040 -2049, Scenario A2. ....................................268 5-145 Projected July maximum temperature 2090 -2099, Scenario A2. ....................................269 5-146 Projected 20402049 July maximum temperature profile under scenario B1. ................270 5-147 Projected 20902099 July maximum temperature profile under scenario B1. ................270 5-148 Projected 20402049 July maximum temperature profile under scenario A1B. .............271 5-149 Projected 20902099 July maximum temperature profile under scenario A1B. .............271 5-150 Projected 20402049 July maximum temperature profile under scenario A2. ................272 5-151 Projected 20902099 July maximum temperature profile under scenario A2. ................272 5-152 CCSM projected January precipitation anomalies and their statistical significance under scenario B1.............................................................................................................274 5-153 CCSM projected January precipitation anomalies and their statistical significance under scenario A1B. .........................................................................................................275 5-154 CCSM projected January precipitation anomalies and their statistical significance under scenario A2. ...........................................................................................................276 5-155 CCSM projected July precipitation anomalies and their statistical significance under scenario B1. ......................................................................................................................277 5-156 CCSM projected July precipitation anomalies and their statistical significance under scenario A1B. ...................................................................................................................278 5-157 CCSM projected July precipitation anomalies and their statistical significance under scenario A2. .....................................................................................................................279 5-158 Projected January precipitation 2020-2029, Scenario B1. ...............................................280 5-159 Projected January precipitatio n 2040-2049, Scenario B1. ...............................................281 5-160 Projected January precipitation 2060-2069, Scenario B1. ...............................................282

PAGE 17

17 5-161 Projected January precipitation 2090-2099, Scenario B1. ...............................................283 5-162 Anomaly map for January precipitation, comparing 2040-2049 projections under scenario B1 to the 1971-2000 baseline. ...........................................................................284 5-163 Anomaly map for January precipitation, comparing 2090-2099 projections under scenario B1 to the 1971-2000 baseline. ...........................................................................285 5-164 Projected 2040-2049 January precipitation profile under scenario B1............................286 5-165 Projected 2090-2099 January precipitation profile under scenario B1............................286 5-166 Projected Januar y precipitation 2020-2029, Scenario A1B. ............................................288 5-167 Projected January precipitation 2040-2049, Scenario A1B. ............................................289 5-168 Projected January precipitation 2060-2069, Scenario A1B. ............................................290 5-169 Projected January precipitation 2090-2099, Scenario A1B. ............................................291 5-170 Anomaly map for January precipitation, comparing 2040-2049 projections under scenario A1B to the 1971-2000 baseline. ........................................................................292 5-171 Anomaly map for January precipitation, comparing 2090-2099 projections under scenario A1B to the 1971-2000 baseline. ........................................................................293 5-172 Projected 2040-2049 January precipitation profile under scenario A1B. ........................294 5-173 Projected 2090-2099 January precipitation profile under scenario A1B. ........................294 5-174 Projected January precipitation 2020-2029, Scenario A2................................................295 5-175 Projected January precipitation 2040-2049, Scenario A2................................................296 5-176 Projected January precipitation 2060-2069, Scenario A2................................................297 5-177 Projected January precipitation 2090-2099, Scenario A2................................................298 5-178 Anomaly map for January precipitation, comparing 2040-2049 projections under scenario A2 to the 1971-2000 baseline. ...........................................................................299 5-179 Anomaly map for January precipitation, comparing 2090-2099 projections under scenario A2 to the 1971-2000 baseline. ...........................................................................300 5-180 Projected 2040-2049 January precipitation profile under scenario A2. ..........................301 5-181 Projected 2090-2099 January precipitation profile under scenario A2. ..........................301 5-182 Projected July precipitation 2020-2029, Scenario B1......................................................302

PAGE 18

18 5-183 Projected July precipitation 2040-2049, Scenario B1......................................................303 5-184 Projected July precipitation 2060-2069, Scenario B1......................................................304 5-185 Projected July precipitation 2090-2099, Scenario B1......................................................305 5-186 Anomaly map for July precipitation, comparing 2040-2049 projections under scenario B1 to the 1971-2000 baseline. ...........................................................................306 5-187 Anomaly map for July precipitation, comparing 2090-2099 projections under scenario B1 to the 1971-2000 baseline. ...........................................................................307 5-188 Projected 2040-2049 July precipitation profile under scenario B1. ................................308 5-189 Projected 2090-2099 July precipitation profile under scenario B1. ................................308 5-190 Projected July precipitation 2020-2029, Scenario A1B...................................................310 5-191 Projected July precipitation 2040-2049, Scenario A1B...................................................311 5-192 Projected July precipitation 2060-2069, Scenario A1B...................................................312 5-193 Projected July precipitation 2090-2099, Scenario A1B...................................................313 5-194 Anomaly map for July precipitation, comparing 2040-2049 projections under scenario A1B to the 1971-2000 baseline. ........................................................................314 5-195 Anomaly map for July precipitation, comparing 2090-2099 projections under scenario A1B to the 1971-2000 baseline. ........................................................................315 5-196 Projected 2040-2049 July precipitation profile under scenario A1B...............................316 5-197 Projected 2090-2099 July precipitation profile under scenario A1B...............................316 5-198 Projected July precipitation 2020-2029, Scenario A2. ....................................................317 5-199 Projected July preci pitation 2040-2049, Scenario A2. ....................................................318 5-200 Projected July precipitation 2060-2069, Scenario A2. ....................................................319 5-201 Projected July precipitation 2090-2099, Scenario A2. ....................................................320 5-202 Anomaly map for July precipitation, comparing 2040-2049 projections under scenario A2 to the 1971-2000 baseline. ...........................................................................321 5-203 Anomaly map for July precipitation, comparing 2090-2099 projections under scenario A2 to the 1971-2000 baseline. ...........................................................................322 5-204 Projected 2040-2049 July precipitation profile under scenario A2. ................................323

PAGE 19

19 5-205 Projected 2090-2099 July precipitation profile under scenario A2. ................................323 6-1 Percentage o f climate models projecting globally and regionally non analog climates by 2100.............................................................................................................................330 7-1 High resolution detail of mean January temperature pro jection for the Tampa Bay area ..................................................................................................................................348 7-2 High resolution detail of mean January temperature projection for the Tampa Bay area, incorporating anticipated urban growth and 2 m of sea level rise. ..........................349

PAGE 20

20 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy PROJECTING REGIONAL CLIMATE CHANGE IN FL ORIDA VIA GIS BASED DOWNSCALING OF A GENERAL CIRCULATION MODEL By Danny Coenen August 2010 Chair: Thomas L. Crisman Major: Interdisciplinary Ecology High resolution projections of temperature and precipitation change s in Florida during January and July through the year 2100 were produced by GIS based geostatistical downscaling of general circulation model outputs provided by the Community Climate System Model 3.0 (CCSM) for IPCC benchmark scenarios B1, A1B and A2. Calibration methods designed to assess and correct for CCSM biases were evaluated and implemented. Results indicate mean state wide temperature anomalies ranging from +1.32 C to +2.64C during January by the end of the century. All scenarios reveal a weakening of the latitudinal climatic gradient during winter. The northwestern panhandle is projected to experience the greatest warming, with anomalies decreasing towards the south east July anomalies range from +1.22C to +3.38C with little regional differentiation except for scenario A2, which projects Florida to become more isothermal than is presently the case. Only scenario B1 projects greater warming in January than July. Mean statewide p recipitation anomalies for January are near zero for all scenarios whereas during July, anomalies range from -21 mm to 42 mm. P anhandle precipitation is expected to remain similar to present conditions or slightly wetter South Florida

PAGE 21

21 is projected to experience highly significant drying, with some areas projected to receive as little as 23.6% of current rainfall under scenario A2. Projected patterns of change strongly suggest increasing tempe rature stress for temperate and warm temperate taxa near the southern margin of their distribution, while barriers to northward expansion of subtropical and tropical taxa are reduced due to the decreasing slope of the latitudin al winter temperature gradien t. As they expand the northern margin of their range warm adapted species are likely to successfully compete with and exploit resources made available by failing temperate and warm temperate communities Subtropical and tropical species in south Florida w ill experience increasing water stress due to sharply reduced summer precipitation, favoring droughttolerant species in novel ecological communities without present analogs

PAGE 22

22 CHAPTER 1 INTRODUCTION Background According to the 4th assessment report released by the Intergovernmental Panel on Climate Change [IPCC], mean global temperatures rose by 0.740.2C between 1905 and 2005, recently increasing at a rate of 0.2C per decade (IPCC 2007a ). An a nalysis of National Climatic Data Center [NCDC] records (NCDC 2010a) conducted for this study revealed 2006 through 2009 as the 5th, 7th, 10th, and 6th warmest years of the instrumenta l record, respectively, with combined land and ocean anomalies exceeding 1971-2000 averages by 0.27 C to 0.35C (Figure 1-1). The first five months of 2010 ranked as the warmes t January May period since the instrumental record began in 1880 (NCDC, 2010b). By 2100, IPCC e mission s cenarios project carbon dioxide ( CO2) concentrations at 1.9 to 3.5 times of the pre industrial value of 280 ppm as a result of anthropogenic activities (IPCC, 2007a) In combination with other f orcing agents that alter atmospheric retention of outgoing long wave radiation or the amount of incoming shortwave solar radiation, an additional mean global temperature increase of 1.8 to 4.0C relative to the 1980-1999 average is projected by the end of this century (IPCC 2007a ). Despite the relatively minor temperature change to date compared to projections for the next century and beyond numerous studies have documented that many species are already affected and have begun responding via habitat range shifts, altered pheno logy, extirpation and other mechanisms ( McCarty, 2001; Parmesan & Yohe, 2003; Lovejoy & Hannah, 2005; Parmesan, 2006; Pounds et al. 2006; IPCC, 2007b). Without comprehensive adaptation and mitigation efforts, impacts are expected to accelerate and persist worldwide for centuries to come (IPCC, 2007b).

PAGE 23

23 Some governments and non -governmental organizations, recognizing the urgent need for climate change mitiga tion have established guidelines, policies and international treaties to curb greenhouse gas ( GHG ) emissions since the late 1990s. While there has been moderate progress in reducing emissions by some countries party to the Kyoto P rotocol to the United Nat ions Framework Convention on Climate Change [UNFCCC] (UNFCCC, 1997, 2007a) these measures have failed to slow the stillaccelerating growth in global emissions thus far (Canadell et al. 2007; Raupach et al. 2007). Given the limited impact of the Kyoto Protocol and contentious ongoing debate about its post-2012 successor, it appears increasingly unlikely that policy initiatives will succeed at achieving the necessary reduction in GHG emissions in order to stabilize mean global temperatures at 2C above pre industrial conditions ( Bierbaum & Raven 2007; Friedlingstein, 2008). This figure is commonly cited as the approximate threshold for dangerous climate change, beyond which nonlinear positive climate feedbacks and deleterious impacts at all levels of biological organization become inc reasingly more likely (Schellnhuber et al. 2006; Bierbaum & Raven 2007 ; IPCC, 2007b; Lenton et al. 2008). Florida the subject of this study, is arguably among the most vulnerable areas to climate change due to its low topographic relief and potential for inundation of coastal regions by sea level ris e, salt water intrusion, exposure to tropical cyclones, and sensitivity to drought. Further, th ere is a high likelihood of speciesindividualistic responses giving rise to novel ecosystems of unknown structure and function (Williams & Jackson, 2007) a central aspect of which is expected to be n orthward range expansion of tropical taxa currently con strained to the southern regions of the Florida peninsula by physiological limitations to winter cold. These climate forced ecological succession processes have implications for continued provision of ecosystem goods and services, agriculture, biodiversity, and conservation of threatened and endangered species

PAGE 24

24 (Mulkey, 2007; Stanton & Ackerman, 2007). Potential northward spread of tropical disease vectors is a concern for public health. Despite these risks, Florida has been s ubject of few peer reviewed studies, although interdisciplinary workshops and conferences have improved the state of knowledge on past, present and future regional change since 2006 (Florida Center for Environmental Studies [CES] 2006, 2007). In 2005, I completed my masters thesis on recent climatic variability in Florida for which data from 47 lakes in t he Florida peninsula, collected between 1968 and 2004, were examined to quantify interdecadal temperature trends (Coenen, 2005) Statistical comparison of 10year temperature means provided no evidence for warming during summer s whereas winters displayed a complex pattern of interdecadal variability marked by rapid warming of 2.11.0C ( standard deviation (SD) ) between intervals 1 (1975-1984) and 2 (1985-1994), followed by moderate cooling of 1.00.8C between intervals 2 and 3 (1995-2004) (Figure 1-2). A nalysis of time series regression slopes yielded an estimate d 1.00.9C net winter warming between 1973 and 2003, favor ing northward range expansion of tropical and subtropical species paralleled by displacement of temperate species (Coenen, 2005). Other studies have arrived at different conclusions regarding detectability of recent warming in Florida ( Henry, 1998; Marshall et al ., 2004). This is not surprising regional effects of climate change are diverse, and the anticipated magnitude of recent temperature change in Florida is moderate due to its low latitude Floridas climate is also subject to significant interannual variability. In combination with widespread land cover change throughout the 20th century that has affected local climates these factors have complicate d detection of a regional expression of the global climate trajectory to date. As climate change accelerates, this signal is ultimately expected to manifest itself more clearly.

PAGE 25

25 A major issue confounding impact assessments of future climate change in Florid a has, until recently, been that grid resolutions of most general circulation models ( GCMs ) were too coarse to resolve the peninsula as a terrestrial feature, limiting the usefulness of GCM based downscaling studies A new generation of models is beginning to change this. Study Overview and Objectives This study establishes a foundation for assessing future climate change impacts in Florida by presenting results of geographic information system (GIS) -based downscaling models that were applied to Community Climate System Model 3.0 ( CCSM ) outputs for IPCC benchmark scenarios B1 A1B and A2 (IPCC, 2000) A suite of calibrated and bias corrected temperature and precipitation change maps for January and July, the coldest and warmest months for most pa rts of Florida (Henry et al. 1994), were produced These customizable electronic maps illustrate spatial and temporal patterns of future change and highlight key vulnerabilities within the state as has been called for by Mulkey (2007). Global as well as Florida specific interdisciplinary perspectives on 21st century climate change are provided. Following a treatise placing global climate change into a long term perspective in chapter 2, general implications for international conservation and dev elopment efforts are discussed in chapter 3. Chapter 4 introduces m ethods for producing geostatistically downscaled climate projections for Florida results of which are presented in chapter 5. While limitations are inherent in any climate and ecological projection as discussed in chapter 6 they are useful to constrain the range of future scenarios and thereby aid in creation of climate change integrated ecological management strategies based on principles of adaptive management. Chapter 7 discusses ecolo gical and socioeconomic impacts specific to Florida This includes an initial assessment of implications by reinterpreting existing literature in light of the results of this study The vulnerability of Florida to climate change and scarcity of regional

PAGE 26

26 impact assessments to date underlines the significance of this study which is synthesized along with directions for future research in concluding remarks in chapter 8.

PAGE 27

27 Figure 1-1. Mean global temperature between 1880 and 2009 expressed as annual anomalies relative to a 1971-2000 baseline (black dots) and a 10year moving average trend (red line) [ Analysis based on data provided by NCDC (2010a) .] 0.7 C 0.6 C 0.5 C 0.4 C 0.3 C 0.2 C 0.1 C 0.0 C 0.1 C 0.2 C 0.3 C 0.4 C 0.5 C 1870 1890 1910 1930 1950 1970 1990 2010Temperature Anomaly ( C)Year Annual Anomaly 10 year Mov. Avg. Trend

PAGE 28

28 Fig ure 1-2. Results of Coenen (2005), illustrating decadalscale winter temperature variability in Florida since the mid -1970s as reflected in lake surface temperature trend s.

PAGE 29

29 CHAPTER 2 CONTEMPORARY GLOBAL CLIMATE CHANGE IN PERSPECTIVE Physical Basis and Characteristics of Global Anthropogenic Climate Change N umerous studies have documented Earths climate to be highly variable on geologic timescales. In order of decreasing timescale, responsible agents include long term solar intensification (Sagan & Mull en, 1972) tectonic forcing (Zachos et al ., 2001) orbital forcing (Hays et al ., 1976), and shortterm solar variability (IPCC, 2007 a). Vostok ice core data suggest average rates of temperature change of 0.010.017C per century over the past 420,000 years, nearly two orders of magnitude slower than current, anthropogenic climate change (Hoegh G uldberg et al. 2007). R a pid climate excursions have also been identified in the paleoclimate record, attributed to discrete events such as glacial outburst floods providing massive freshwater discharge s into the Atlantic Ocean (Kleiven et al. 2008), major cosmic impact events (Alvarez et al. 1980; Pope et al. 1998), destabilization and outgassing of benthic methane clathrate deposits (Dickens et al. 1995; Katz e t al. 2001; Maslin et al. 2004; Zachos et al. 2005), and, to a lesser extent, volcanism (Robock & Mao, 1995; Zielinski et al. 1996). Fossil records suggest that rapid climate excursions can trigger widespread ecological reorganization, including range shifts and extinctions ( Crowley & North, 1988; Pope et al. 1998; Bowen et al. 2002). While a minority of scientists maintain that humans began influencing climate through agriculture and deforestation as early as 8000 years ago (Ruddiman, 2003, 2005), i t is more widely held that anthropogenic climate change is causally linked to fossil fuel combustion and large scale land cover conversion initiated during the I ndustrial R evolution and rapidly accelerating since the mid 20th century (IPCC, 2007a) Atmosph eric concentrations of CO2 increased from a pre -industrial average of 280 ppm to 387 ppm in 2009 ( Tans 2010). This is the

PAGE 30

30 highest level during the past 650,000 years (Siegenthaler et al. 2005; IPCC, 2007b), and possibly the past 20 million years (Pearson & Palmer, 2000). The rate of change for the period of 2000-2006, 1.93 ppm/year (Canadell et al. 2007), is unprecedented over the past 650,000 years during which it never exceeded 30 ppm/1 000 years (IPCC, 2007a ). The continued growth in emissions is attributable to a combination of global industrial activity, increasing carbon intensity of the global economy, and a decline in the efficiency of the oceanic carbon sink (Canadell et al. 2007; Raupach et al. 2007). Even the global financial and economic crisis that began in 2007 and continues as of 2010 has not slowed this trend. According to National Oceanic and Atmospheric Administration [ NOAA ] data (Tans 2010), CO2 concentrations continued to increase at a mean r ate of 1.93 ppm/y ear between 2007 and 2009. Along with increases in atmospheric methane (NH4), nitrous oxide (N2O) and various industrial halocarbons, the total magnitude of enhanced greenhouse forcing as of 2007 is estimated at +2.64 W/m2. In comparison, solar forcing increased by only 0.12 W/m2, although the level of scientific understanding from which this quantity is derived is described as low (IPCC, 2007a ). Increases in certain types of atmospheric aerosols and other negative forcings partially offse t these effects, resulting in a net change in the planetary energy balance of +1.7 W/m2 (IPCC, 2007a ). Between 1905 and 2005, enhanced greenhouse forcing has increased the global mean temperature by 0.74C0.2C in two distinct phases : Between 1910 and 1945, to which solar forcing contributed significantly, and since 1976, which is almost entirely attributable to human activities (Stott et al. 2001). The three decade -long period following World War II was characterized by a slight cooling trend which w as most pronounced in Europe and North America, and was possibly associated with high regional aerosol emissions prior to implementation of clean air legislation (Stott et al. 2001).

PAGE 31

31 Combined with additional warming projected by the IPCC for the 21st cen tury, total warming between 1900 and 2100 may approach the magnitude of some of the most pronounced climate excursions during the Cenozoic era in what fundamentally constitutes a geological and, for long lived species, an evolutionary instant. Moreover, cl imate change is not limited to temperature change alone. Changing precipitation patterns, increase in frequency of extreme weather events and sea level rise have also been projected. Reduction in carbon sink efficiency and conversion of some carbon sinks to net carbon sources have been projected in models, progressively increasing the fraction of anthropogenic emissions that remain in the atmosphere This suggests that mitigation will be even greater a challenge than commonly assumed (Friedlingstein, 2008) Currently, t h e atmospheric fraction represents 45% of anthropogenic emissions, the rest being sequestered by terrestrial and oceanic carbon sinks (Friedlingstein, 2008). Of particular concern is the potential for the climate system, or critical components thereof, being pushed past critical thresholds known as tipping points that result in positive feedback mechanisms enhancing and accelerating the warming trend (Lenton et al. 2008). Unfortunately, the unique rate and patterns associated with anthropoge nic climate change suggest that the past is not necessarily instructive in quantifying the tipping points of the present. While substantial uncertainty remains, models and expert assessments are beginning to produce quantitative estimates that instill a sense of urgency. Climate components that may approach their estimated critical values within the next 100 years according to Lenton et al. (2008) include arctic summer sea ice, the Greenland ice sheet, the West Antarctic Ice Sheet, the Atlantic thermohaline circulation, the mean state of the El Nio Southern Oscillation ( ENSO ) the Indian Summer Monsoon, the Sahara/Sahel and West African monsoon, the Amazon rainforest and boreal

PAGE 32

32 forests. Critical thresholds for these elements are estimated to range from 0.5 C to 5C of global warming in excess of the 19801999 average, values that mostly lie within the range of IPCC projections for the next century R esponse times vary from element to element, with some expected to complete transitions to alternate stable states within years, while it will take others centuries (Lenton et al. 2008). The process of arctic sea ice disappearing during northern hemisphere summers is likely already underway (Lindsay & Zhang, 2005). Climate Change Scenarios Climate projections are most commonly based on the widely used IPCC -SRES benchmark scenarios B1, A1B and A2 ( IPCC, 2000; Figure 2-1) Scenario based projections are u se d to capture a diverse array of plausible future demographic, socioeconomic, and technological realities Since methods and quantity of fossil fuel consumption along with land use change, are the primary drivers of anthropogenic climate change, future human activity is the major source of uncertainty in climate projections. Events such as the sudde n collapse of the I ron C urtain in 1989 and the global economic crisis of 2007-2010, which unfolded over a matter of weeks to months, powerfully demonstrate that the fickle nature of aggregate human behavior is unpredictable Climate scenarios essentially constrain the range of all possible future development trajectories to a more manageable subset that encompasses most, but not all, plausible ones Aside from the three benchmark scenarios, there are 37 other climate scenarios in four scenario families currently used by the IPCC ( 2000). In modeling, scenarios are translated into quantitative GCM inputs, most notably anticipated rat es of GHG emissions (Figure 2-2A). Scenario A2 assumes a fragmented world focused on regionally heterogeneous development with little convergence between rich and poor nations. It assumes sustained increases in human population and relatively slow percapita economic growth and technological advances. Globally, scenario A2 is projected to produce the greatest warming by 2100 among the

PAGE 33

33 three scenarios (3.4C relative to the 1980-1999 average), significantly diverging from scenario A1B after approximately 2065 (IPCC, 2007a). Scenario A1B is a member of the A1 family of scenarios, which describe a world focused on globalization, with a significant narrowing in the economic disparity between developing and developed nations. E conomic growth and introduction of new technologies are rapid, and glo bal population peaks by midcentury. A1B assumes that these goals are realized by relying on a balanced mix of fossil -based and nonfossil energy resources. Globally, Scenario A1B is associated with intermediate levels of global warming (2.8C relative to the 1980-1999 average) (IPCC, 2007a). Scenario B1 is similar to the A1 family of scenarios, except its developmental focus is a service and information based global economy that places h eavy emphasis on energy efficiency and international cooperation in achieving these goals. It produces the least amount of global warming through 2100 (1.8C relative to the 1980-1999 average) (IPCC, 2007a). None of these three scenarios make specific assumptions about climate change mitigation policies, including the Kyoto P rotocol to the UNFCCC (IPCC, 2007a) and none take the possibility of climate tipping points (i.e. rapid transitions into alternate stable climate states) into account Since 1990, actual emissions have surpassed those assumed under either benchmark scenario, paralleling a more rarely refer enced scenario A1 Fl most closely (Raupach et al. 2007; Figure 2-2B). Scenario A1Fl assumes that the goals of the A1 family of scenarios are reached by satisfying energy demand with fossil fuels, resulting in 4C warming by 2100 relative to the 1980-1999 average (IPCC, 2007a). Given ongoing political resistance to implement meaningful, globally coordinated climate change mitigation policies, the three benchmark scenarios must be considered conservative in the intermediate future.

PAGE 34

34 In addition to the IPCC scenarios, many study specific climate change scenarios exist. These include stabilization scenarios that fix the timing and magnitude of the peak in atmospheric CO2 concentrations (e.g. green and blue dashed lines in Figure 2-2) and various temperature or precipitation -t arget scenarios (e.g. Box et al ., 1999) that specify assumed temperature or precipitation anomalies by some future date. Because study specific scenarios are difficult to compare with the vast body of literature based on IPCC scenarios, they were not furth er investigated for the present study aside from literature reviews.

PAGE 35

35 Figure 2-1. Projected surface temperature change (relative to 1980 ) for benchmark scenarios B1 A1B A2 and a constant composition scenario as continuations of the 20th Century Ex periment. Shading denotes an uncertainty margin of 1 SD Colored numbers indicate the number of model runs used to generate each projection per 100year time interval. [Reprinted from IPCC (2007a), Climate change 2007 T he physical science basis: W orking group I contribution to the fourth assessment report of the IPCC (p 762, Fig.10.4)].

PAGE 36

36 Figure 2-2. CO2 e mission s under various scenarios. A) Projections through 2100; B) A c tual CO2 emissions compared to SRES -scenario assumptions since 1990. [Reprinted from Raupach et al. (2007) Global and regional drivers of accelerating CO2 emissions. Proceedings of the National Academy of Sciences 104, 10288-10293. (p.10289, Fig. 1) Copyright by the National Academy of Sciences. ] A B

PAGE 37

37 CHAPTER 3 INCORPORATING THE RE ALITY OF CLIMATE CHANGE INTO CONSERVATION AND DEVELOPMENT STRATEGIES1Successful integration of conse rvation and economic development into a synergistic framework has emerged as a central objective in interdisciplinary conservation biology. This paradigm developed following realization that approaches focusing on single goals to the exclusion of others ha ve frequently yielded suboptimal results despite significant capital investment. As detailed by Brown (2002) and Holt (2005) lack of consideration for local peoples resource needs and land use practices in design and management of protected areas, in som e cases to the point of total exclusion or forced resettlement, inevitably results in high rates of noncompliance, creating conflicts and undermining conservation objectives. The flip side of the coin is resource exploitation -based development in the absence of conservation planning, which, given high population pressure in much of the developing world, tends to result in rapid environmental degradation and a multitude of socioeconomic impacts ranging from increased income disparity to loss of traditional c ulture and identity. Integrated views of conservation and development seek to ameliorate these issues through participation of local stakeholders and encouraging conservation through sustainable use while minimizing exclusionary ecoprotectionism (Brown 2002; Holt, 2005). Experience has shown that this integration process is not straightforward. Not only have long standing incongruities among theoretical frameworks guiding research and management in natural, social and economic sciences proven difficult to overcome, but, more significantly, approaches successful in one geographic location often fail in others. Complicating factors include, but are not limited to, difficulty quantifying the true economic value of ecosystem goods 1 This chapter is an excerpt from Future directions in conservation and development: Integrating the reality of climate change, published by Coenen et al (2008) in Biod iversity 9 (3 4), 106 113. Reproduced with permission.

PAGE 38

38 and services ( Edwards & Abivar di 1998 ; Wilson & Carpenter 1999; Turner et al ., 2002; Azqueta & Delacmara 2006), difficulty identifying and delineating areas most deserving protection (Brooks et al ., 2006), false assumptions about homogeneity of stakeholder communities (Brown 2002), incomplete models, insufficient data, corruption and lack of strong institutions to enforce regulations (Ferraro, 2001 ; Damania & Hatch 2004), non recognition of local peoples property and use rights (Ferraro 2001; Harris 2005), and barriers to international technology transfers limiting dispersal of resource efficient technology to industrializing nations (Gallagher 2006). For the foreseeable future, conservation projects will continue to be most successful when designed site specific, taking local ecological and socioeconomic complexity into full account, as both have dynamic, emergent properties that cannot be fully predicted using strictly reductionist approaches. Accelerating climatic change underlines the urgent need for sustainable devel opment and successful conservation measures. It also forces changes to their implementation, as conventional approaches may no longer yield positive outcomes under rapidly changing climatic conditions (Harris et al ., 2006). Ecosystem management, because of its intrinsic properties and dynamics, is a complex mix of science and policy practiced by unevenly participating stakeholders within institutions that may in many cases be ill -prepared for the added complexity and uncertainty introduced by climate change Ecology, socioeconomics, technological development and political realities, themselves actors responding dynamically and at times unpredictably to climate change, are becoming increasingly complexly intertwined to shape future conservation outcomes. Adap ting to these developments requires multidisciplinary cooperation and interdisciplinary thinking, backed by appropriate funding for research and monitoring needs if devastating future scenarios are to be avoided.

PAGE 39

39 Anticipated Climate Change Impacts in the 21st Century Delineation of likely ecological and socioeconomic impacts under various scenarios of climate change has emerged as a central research priority. Since climate change exhibits substantial spatial and temporal complexity, analysis at regional sc ales is required for risk assessment informing conservation and mitigation efforts. Yet, GCMs despite their increasing complexity, remain insufficient to characterize small and irregularly shaped landmasses, heterogeneous land cover and small scale circul ation processes adequately. Regional climate models (RCMs) and various downscaling methodologies are beginning to address this information gap. However, while models can assist in evaluating potential consequences of various scenarios of climate change, a number of research articles have highlighted important limitations to the ability to fully predict future climates and ecological systems due to their inherently uncertain or chaotic properties ( Roe & Baker 2007; Beninca et al ., 2008; Lenton et al ., 2008) On local to regional scales those most applicable to conservation management factors indirectly or completely unrelated to global processes such as land use change become increasingly important in shaping climate (Pielke Sr. et al ., 2002; Pearson & Dawson, 2003). These effects are accompanied by substantial existing anthropogenic resilience-depressing stresses (McCarty 2001), including land transformation, unsustainable resource extraction, hydrological manipulation, habitat fragmentation and nutrient enrichment. For these reasons, prescriptive, site -specific bioclimate forecasts are unlikely to ever be produced, highli ghting the importance of accepting and managing for uncertainty as a key element of climate change ecology. Ecological E ffects Climate directly controls or affects many ecological and biological processes, including nutrient cycling, onset and duration of the growing season, timing of reproduction,

PAGE 40

40 environmental sex determination, animal behavior, and body size in some taxa ( Walther et al ., 2002; Parmesan & Galbraith 2004). Sustained temperature and precipitation deviations beyond historical variability f orce climate sensitive species to adjust phenology, exhibit phenotypic plasticity and/or shift habitat ranges to maintain suitable conditions for survival. Numerous studies have established observational evidence that such responses are already occurring in many taxa (Parmesan & Yohe 2003; Parmesan & Galbraith 2004; Parmesan 2006 ; IPCC 2007b). The number of onthe ground observations of ongoing impacts has increased greatly during the past decade, but their spatial distribution is highly heterogeneous. Of nearly 29,000 long term observational data series examined by the IPCC that showed significant biological changes, 98% were collected in Europe, with only two documented impacts in Africa ( IPCC 2007b; Nature Editors 2007). 90% of these observed changes are consistent with expectations for a warming world (IPCC, 2007b). Because climatic tolerance and the speed and ability to migrate vary between species, non synchronous responses are expected to occur as climate change accelerates, altering symbiotic in teractions such that extant communities and trophic webs become progressively uncoupled ( Hannah et al ., 2002b ; Walther et al ., 2002 ; Williams & Jackson 2007). Projected implications include reduced carbon uptake by terrestrial ecosystems, increased extinction risk for up to 30% of plant and animal species depressing local and global biodiversity, and accordingly, a major reorganization of ecosystem stru cture and function, potentially severely disrupting provision of goods and services with cascading socioeconomic effects (IPCC 2007a, b). Evolutionary responses are likely limited to adaptation via phenotypic plasticity for long lived species, whereas those with rapid generation times respond via natural selection, favoring dispersal ability and adaptations beneficial under altered local conditions, e.g. ability to colonize

PAGE 41

41 new habitat types (Pearson & Dawson 2003). Similar processes occurred following the last glacial maximum, albeit over several thousand years, requiring slower rates of dispersal. There is little evidence for changes in absolute climatic tolerances of species via evolutionary mechanisms in the fossil record ( Parmesan 2006 ; Petit et a l ., 2008). Projecting structure and function of future ecosystems is subject to considerable uncertainty, since models of imperfectly known systems incorporate assumptions, parameterizations and inevitably exclude some variables from consideration. Bioclimate envelope models are commonly used to produce initial assessments of potential future species range shifts over large spatial scales, although they have inherent limitations of their own addressed in detail by Pearson & Dawson (2003). Bioclimate envelope shifts of dozens to hundreds of kilometers are not uncommon in projections, potentially resulting in species of conservation concern adjusting their habitats b eyond borders of protected areas established to preserve them (Hannah et al ., 2002a). Successful habitat adjustment is further constrained by anthropogenic barriers, as well as resource availability and community dynamics in the new geographic range (Walther et al ., 2002). These combined effects may ultimately produce unique communities consisting of species not co existing currently. Williams & Jackson (2007) suggest that novel climate regimes are expected to develop throughout the tropics and subtropics, including areas considered biodiversity hotspots. Attempting to project future ecologi cal organization in climates without present analogs is fraught with additional uncertainty due to the extrapolative nature of such projections (Williams & Jackson 2007). Socioeconomic E ffects The total cost of unmitigated climate change as direct human impacts and indirect impacts resulting from the disruption of ecosystem services has been estimated to reach up to 20% of the

PAGE 42

42 global gross domestic product by 2100, implying that the cost of inaction vastly exceeds that of comprehensive mitigation (Stern 2007). Direct socioeconomic impacts are caused by limitations of human physiology to withstand weather extremes, and displacement as a result of sea level rise, glacial outburst floods and other climate linked processes. If projections are correct, tragic events such as the European heat waves of 2003 and 2006 will recur at higher frequencies and intensities (IPCC 2007a ). Implications for tropical cyclone frequency and intensity are still subject to considerable debate, with published research yielding con flicting results about the relative contributions of increasing sea surface temperatures and wind shear in the Atlantic Ocean (IPCC 2007a ; Vecchi & Soden 2007 ; Saunders & Lea 2008). There is a need to prepare for potentially tens of millions of climate change refugees expected to be displaced from their homes as a result of climate change related impacts (Bierbaum & Raven 2007) most significantly flooding of densely settled coastal areas and inundation of lowelevation island nations due to sea level rise. Indirect effects are imposed by changes in ecosystem structure and function, and will be particularly hard -felt by people dependent on ecosystem goods and services for provision of food, water and other essential resources. Spread of disease vectors affecting humans and their food crops is an added concern. Impacts on food production will be regionally and temporally heterogeneous, as multiple variables ranging from carbon fertilization to altered water budgets interact to boost or depress yields. In general, low latitude agrosystems are projected to be among the first to experience adverse impacts (IPCC 2007b), potentially forcing rural people to increase use or conversion of land to compensate for loss of income and sustenance. The 4th assessment r eport of IPCC Working Group II describes scenarios including altered volume of glacial melt that is a major source of drinking water in many parts of the world, major losses of

PAGE 43

43 coastal ecosystems and infrastructure due to increased erosion, flooding and sa lt water intrusion increased drought and flash flood risk in currently semi arid to arid areas, and a variety of human health impacts (IPCC 2007b). Generally, these consequences will be more acute in developing countries due to lower adaptive capacity, w ith Africa likely the most adversely affected continent. The Sahel and other transitional ecosystems have been subjected to climatic change for decades, mostly via altered precipitation regimes associated with progressive desertification. Giles (2007) not ed that local adaptation to these processes remains poorly studied, but evidence suggests that many communities have proven surprisingly resilient. Changing agricultural practices and a focus on collective production, allowing for diversification and risk sharing, have occurred. While it is impossible to generalize from these isolated studies, they suggest that adaptation is possible even in the absence of a modern technological base. The question remains as to how much additional change can be absorbed by local ingenuity without further deterioration of already marginal living conditions. Implications for Conservation and Development Current strategies for minimizing climate change impacts on biota and human populations fall into the broad categories of mitigation (reducing and sequestering emissions) and adaptation (infrastructure upgrades, climate change integrated conservation strategies) ( Hannah et al ., 2002a; Bierbaum & Raven 2007). Both go handin -hand within the general framework of conservation and development and present opportunities for synergies. Ultimately, success of future biodiversity conservation endeavors hinge on minimizing the magnitude of additional climate cha nge developing and applying climate changeintegrated conservation and development strategies and taking advantage of funds available through carbon markets to reach conservation objectives

PAGE 44

44 Climate C hange -I ntegrated C onservation Strategies According to Folke (2006), a major shift in perspective is currently taking place with an emphasis on complex adaptive systems characterized by nonlinear relations, path dependency, thresholds, regime shifts, and multiple basins of attraction. [] Conservation thinkin g needs to move away from steady state solutions to accept that change is the rule rather than the exception (p. 687) Climate change integrated conservation strategies build upon these principles, with resilience management and adaptive management at multiple spatial and temporal scales being central elements. As pointed out by Harris et al. (2006), not all classic conservation biology wisdom will still apply as global warming accelerates, requiring a comprehensive re evaluation of traditional conservatio n practices. The inappropriateness of continuing to utilize historical benchmarks as restoration targets, and species conservation efforts in static preserves that may soon experience abiotic conditions falling outside the target organisms tolerance, are just two examples. Identifying and embracing concepts and methodologies in need of change will be a key challenge for conservationists. Opportunities for synergistic management exist that seek to achieve traditional conservation and climate change -proofing objectives simultaneously by alleviating stresses, maximizing functional redundancy, and increasing connectivity between conservation areas. Increasing local genetic diversity by facilitating reproduction among specimens from distal portions of their natu ral range could be utilized to maximize the adaptive potential of protected species (Harris et al ., 2006). The rapid rate of projected climate change may exceed the ability of some species to keep up with shifting climate zones (Petit et al ., 2008). This h as given rise to a debate among conservation scientists about the merits of creation of migration corridors and assisted migration (also known as assisted colonization) (Hunter 2007; McL achlan et al ., 2007) Where economic and geopolitical realities preclude creating new protected lands along leading edges of shifting

PAGE 45

45 bioclimatic envelopes, and for species whose bioclimatic envelopes are projected to collapse entirely, attempts may focus on delaying responses to preserve the status quo in protected areas f or as long as possible via strategic human intervention to boost resilience that could conceivably incorporate disease control, removal of invasive species and assisted propagation as well as techniques learned from management of non native specimens in zo ological and botanical gardens. Generally, vulnerability likely increases in systems subject to other stresses depressing overall resilience and adaptive capacity (IPCC 2007b), a condition sometimes referred to as general stress syndrome (Western 2006). This may be particularly applicable to coastal systems, such as salt marshes, mangrove swamps and dune communities as they tend to be squeezed between the figurative rock and a hard place, comprised of rising seas and ever expanding human settlements th at fringe many of the worlds coastlines with few opportunities for migration or adaptation. Similar scenarios apply to montane and polar environments. As previously discussed, models are important tools to evaluate potential impacts under scenarios of cl imate change and to assist in formulating risk assessments to inform management, but they cannot predict the precise timing and consequences of climatic and biotic threshold events, many of which will come as a surprise, unanticipated until they occur. F or this reason, robust adaptive management frameworks need to be implemented that are capable of responding to unanticipated events and adapting to new types of ecological communities lacking current analogs. This necessitates development of temporally nes ted management schemes that retain current 3 to 5 year planning intervals while also incorporating long term, multi decadal visions under a variety of impact scenarios (Hannah et al ., 2002a ). Monitoring will contribute greatly to success, providing data necessary to critically evaluate the adequacy of the planning framework

PAGE 46

46 in regular intervals, while also providing opportunities for model validation and refinement. Experiments, both in controlled laboratory environments and in the field, augment the toolbox of adaptive management by providing data points on how representative sample plots, functional groups or individual species may respond to altered environmental conditions in the future. Spatially hierarchical organization is also important. Ideally, management should be global in extent, local in implementation, and regionally coordinated. The regional nexus is an indispensible component, as traditional top -down and bottom-up approaches tend to lose focus at the far end of their scalar spectra, creating issues of noncompliance and freeridership, respectively. Regional coordination is also bestsuited for landscape scale management of the matrix between conservation lands and other pristi ne areas, which is crucial to allow species to migrate as they adjust their ranges ( Hannah et al ., 2002a, b ). Finally, the regional nexus allows for economies of scale in monitoring and modeling by sharing resources. Technology T ransfer and M arket M echani sms in S upport of B iodiversity Conservation Even the best possible conservation management and climate change adaptation efforts cannot succeed without funding and successful mitigation to limit accumulation of atmospheric GHGs to manageable levels. The r emainder of this chapter therefore transcends disciplinary boundaries to assume a more holistic perspective by highlighting a selection of evolving incentives promoting technology transfers, carbon trading mechanisms and conservation funding opportunities, illustrated by pending implementation of the concept of REDD (Reducing Emissions from Deforestation and Degradation). In countries with high adaptive capacity, technological innovation is a central part of climate change mitigation efforts. In Germany, la rge scale deployment of renewable energy generation has contributed to reducing national CO2 emissions by approximately 18.4% between 1990 and 2005 (UNFCCC, 2007b). Carbon capture and storage (CCS) from large point sources

PAGE 47

47 is a potential, yet technologically complex fix that is advocated as an attractive contribution to climate change mitigation, although it substantially decreases the efficiency of power plants utilizing this technology (IPCC 2005). This may have conservation implications by increasing demand for fuels, including mining of fossil sources and biomass. There are numerous other technological options currently available and in development that could contribute to stabilizing atmospheric GHGs concentrations if widely adopted. This process is slowed in part by a lack of accounting for the true cost of carbon emissions, essentially subsidizing fossil fuel intensive industries by allowing them to externalize those costs. More significant for developing countries, economic and capacity limitations, as well as barriers to technology transfers (Gallagher 2006), pose severe limitations to deployment of lowcarbon technologies in parts of the world where emissions are rising most rapidly. Removing these barriers and providing incentives via instruments like the Climate Change Adaptation Fund (Zahabu et al ., 2007) to assist in capacity building are shortterm necessities to reduce new construction of long lived, inefficient infrastructure such as coal fired power plants. Case Example: Realizing C onserv ation Objectives via S ynergies with Reducing Emissions from Deforestation and Degradation ( REDD) Since anthropogenic GHGs are well mixed in the atmosphere, mitigation has global applicability, meaning that sequestration does not have to occur proximal to emission sources. To date, credit for biomass sequestration is limited to afforestation and reforestation projects under the Kyoto clean development mechanism (CDM) whereas efforts to reduce emissions via implementation of improved forest management and reduced deforestation are not currently credited (Zahabu et al ., 2007). Efforts are underway to change this by incorporating REDD into the follow up to the Kyoto P ro tocol Many developing nations have extensive forest cover experiencing rapid rates of deforestation and degradation to accommodate rapidly growing

PAGE 48

48 populations and global resource demand. REDD promises to provide funds to reverse these trends by assigning a direct, creditable market value to carbon sequestration services provided by forests. Globally, $1.2 billion to $10 billion have been cited as becoming available for forest protection (Miles & Kapos 2008). Zahabu et al (2007) describe a scenario for im plementation of REDD in Tanzania, whereby up to $119 per average rural household ($630 million total) could be gained annually by halting deforestation. Reducing deforestation also slows regional climatic changes that would otherwise occur via alteration o f albedo and hydrologic cycles. Deforestation in Amazonia, for instance, creates warmer, drier climates that decrease evaporative cooling and increase both susceptibility to fires and forest dieback, initiating a positive feedback loop (Bonan 2008). Witho ut intervention, critical thresholds may be exceeded to initiate conversion to a lternate biomes such as savanna s, which would irrevocably result in mass extinctions of forest fauna and flora (Lenton et al ., 2008). Most REDD proposals target trading among nations rather than companies or individuals (Miles & Kapos 2008). To what extent funds would be used and re distributed would be left to each government. However, because implementation would occur at the site scale, landowning individuals and communities will be in a strong position to contribute to sequestration efforts, allowing them to negotiate with governments to maximize economic payments for maintaining or expanding land cover compatible with sequestration if empowered by property or use rights, f ulfilling the criteria of integrated conservation and development discussed in the introduction to this chapter This will be most effective for collective efforts, whereby intracommunal competition is reduced, risks are shared, and better negotiating positions achieved, as governments will be keen to minimize transaction and monitoring costs by avoiding contracts with individual smallholders (Grieg Gran et al ., 2005). While governance challenges will have to

PAGE 49

49 be overcome in many countries to build administr ative and enforcement capacity, a model framework already exists in Tanzania through village forest reserves, which experience less degradation from illegal resource extraction than traditional public reserves (Zahabu et al ., 2007). As suggested by Koenig (2008), forest tracts may alternatively be leased from the government by rural communities, which receive payments for avoided deforestation or degradation, based on the principle of additionality relative to a pre determined baseline, in return. This inco me replaces lost logging revenues and allows purchase of resources traditionally extracted from forests if such extraction would compromise sequestration objectives. Many forest uses, including limited harvest of timber and biomass, are compatible with climate change mitigation and can be integrated with sustainable stand management. Lease payments would cover government administrative costs and allow for monitoring to ensure that sequestration objectives are, in fact, met. Such an implementation scheme wou ld provide the type of direct market integration for non -consumptive or lowimpact use that has been so difficult to achieve. Re and afforestation projects established via the CDM framework complement primary forests protected under REDD by creating seco nd growth forests and plantations that, while less diverse than comparable old growth plots, increase habitat connectivity and decrease fragmentation (Stokstad 2008). Therefore, development of carbon trading schemes that incorporate biomass sequestration and appropriate monitoring and enforcement mechanisms are the single most important steps towards integrating rural communities in the climate change adaptation and mitigation process. Conservation objectives may not always coincide with forest plots assig ned the highest priority through REDD, which is strictly based on carbon stocks and does not take other environmental services into account. Many conservation projects do not target forests at all.

PAGE 50

50 However, REDD funds may release conservation funding curre ntly tied up in high carbon, old growth forests to be re -assigned to boost protection of nonforested areas that may be at increased risk for conversion and degradation as ranching and agricultural crop production is displaced from forests elsewhere (Miles & Kapos, 2008), or from direct impacts of climate change. It is in this sense that the market integration REDD promises to generate opportunities that reach far beyond forests alone, but allow climate change integrated conservation strategies to be more w idely employed to boost resilience of many of the most vulnerable ecosystems to avert or delay the most devastating impacts of climate change. Conservation organizations should be prepared for the consequences of REDD implementation and identify at risk a reas to boost investment in while following through by developing capacity for adaptive, regionallycoordinated management with a focus on resilience management and long term monitoring. Section Summary Climate change presents a formidable challeng e to biodiversity conservation, requiring substantial revisions to existing methodologies while also creating new opportunities for success. A central challenge for conservation managers is to ensure a continued supply of ecosystem services while facing su bstantial structural adjustments. Additional objectives include minimization of biodiversity losses, facilitation of migration by increasing habitat connectivity or partaking in assisted migration where appropriate, minimization of non climate related stre ss to increase resilience, and fostering adaptation by increasing local genetic diversity, among others. Realization of these objectives requires climate change integrated conservation strategies based on principles of adaptive management within a long ter m visioning framework, supported by extensive monitoring and modeling efforts. Of crucial importance is development of market mechanisms that account for the true cost of carbon emissions, while simultaneously creating funds for mitigation and adaptation e fforts that are in many cases directly compatible with

PAGE 51

51 conservation objectives or, alternatively, free up funding for conservation targets unsuitable for carbon sequestration. International cooperation to facilitate technology transfer and reduce freerider ship, empowerment of local communities to allow conservation friendly development via payments for carbon sequestration, and regional coordination to monitor local compliance and ensure matrix management are important shortterm goals to form a basis for s uccess. There is no doubt that implementing these measures will be difficult, but that does not diminish their necessity. Climate change is a present reality, and no conservation organization can afford to ignore this fact if long term success is to be ach ieved.

PAGE 52

52 CHAPTER 4 METHOD S FOR GEOGRAPHIC INFORMATI ON SYSTEM (GIS) BASED DOWNSCALING OF GENERAL CIRCULATION MODEL OUPUTS TO PRODUCE HI GHRESOLUTION IMPACT MA PS FOR FLORIDA Interpolation of climate variables is a common component of climatological and ecological studies, and many methods have been utilized to do so (Moral, 2010). After a brief, non technical introduction to the most common deterministic and geostatistical techniques studies based on interpolation of observational climate records and climate projections via GCM downscaling are reviewed and methods used for the present study introduced. Inverse distance weighted ( IDW ) models are deterministic, exact interpolator s that estimate data values for unsampled location s by calculating a weighte d average based on proximity to sampled points with known data values (Nalder & Wein, 1998; ESRI, 2003; Chuanyan et al ., 2005). The exponent of the weight vs. distance power function is determined by user input (Chuanyan et al ., 2005) or can be suggested b y the ArcGIS software (ESRI, 2003). Any point coinciding with the location of a sampling station carries a weight of 100%, i.e. the calculated value o f the map surface at a sample point always match es the observed value at that sample point. The total rang e of v alues of the map surface is constrained by the highest and lowest sample point value s in the dataset respectively (ESRI, 2003). Therefore, this method is prone to underestimating the true range of values in the study area. It is also highly sensitive to uneven sample point distribution and tend s to yield rough surfaces (ESRI, 2003; Chuanyan et al ., 2005) especially in the proximity of data outliers Spline models are also exact deterministic interpolators, but are not constrained by the ran ge of values of its sample points. Instead, splines minimize curvature of the surface, resembling a flexible rubber membrane anchored at each sampling point. The user specifies the

PAGE 53

53 tenseness of the membrane to yield smoother or more undulating surfaces ( ES RI, 2002 ; Chuanyan et al ., 2005). K riging is a geostatistical interpolator incorporating spatial and statistical relationships between sample points by constructing a semivariagram of the data (ESRI, 2002; ESRI 2003). It assigns weights such that bias an d variance of estimates is minimized, providing the best linear unbiased estimate for unsampled locations Provided sampling densit ies are sufficiently high, kriging usually yields better estimates than more simple methods (Moral, 2010). Unlike deterministic methods, kriging produces prediction standard error statistics, allowing prediction errors to be quantified for all points of the interpolated surface (Nalder & Wein, 1998; Moral, 2010). Additional information, most commonly topography, can be incorporated by using co kr iging or kriging extension methods that incorporate multivariate regression functions to improve results further Normality of data is not a requirement for kriging (Moral, 2010), although non normal data compromises val idity of certain kriging -derived products such as quantile and pr obability maps. Detailed information of the equations underlying kriging and it s various semivariogram models are provided by Nalder & Wein (1998), Boer et al (2001), ESRI (2003), and in a n on technical way by Collin s & Bolstad (1996). L iterature Precedents for GIS -Based and Geostatistical Climate Analyses Boer et al (2001) compared four kriging and three thin plate spline models to interpolate maximum temperatures and mean precipitation in Jalisco State, Mexico to support crop model ing. Specific methods explored include bivariate thin plate spline, partial thin plate spline, trivariate thin plate spline, ordinary kriging, ordinary co kriging, regression kriging and trivariate regression -kriging. M odel quality assessed by mean square error ( MSE ) and maximum prediction error (MPE) statistics varied by month For temperature projections, trivariate regression -kriging and trivariate thin plate splines were preferred among method s incorporating

PAGE 54

54 elevation as ancillary data. Among models that did not incorporate elevation data, ordinary kriging produced lower MSE values relative to bivariate thin plate splines in all but one case, while the two approaches were nearly evenly matched in terms of MPE with a slight advantage to ordinary kriging. For precipitation data, which the authors note are considerably more difficult to interpret due to higher data variability ordinary kriging was preferred over spline models by a narrow margin Chua nyan et al (2005) compared linear regressions of temperature against topography to splines, IDW and exponential ordinary kriging to map surface temperature patter n s in the Qilian Mountains of China, where forest area has been in sharp decline and models are required to develop restoration strategies. Their temperature projections were intended as inputs to vegetation distribution models on landscape scales to guide restoration site selection. Assessing models via comparison of prediction error statistics including mean error ( ME, degree of bias) mean absolute error ( MAE magnitude of errors in predictions) and root mean square error ( RMSE, identification of data outliers by comparison to MAE values ) the authors described splines as performing consistently poor for their study area Overall, linear regression of temperature against topography produced the smallest prediction errors, whereas ordinary kriging produced smaller prediction errors during the growing season, and was selected as the preferred model by virtu e of being more accurate at the ecologically most meaningful times Despite being selected for their good performance, t he authors stated that their kriging models may have been suboptimal because of the limited number of data points available within their study area to base their interpolations on. Nalder & Wein (1998) evaluated a method described as gradientplus inverse distance squared ( GIDS ) to interpolate climate normals in northwestern Canada for study ing boreal forest

PAGE 55

55 floor dynamics They describe d their study area as homogenous and of low relief, with a sparse network of climate stations. They compar ed their GIDS method to various geostatistical and deterministic approaches, including detrended kriging, universal kriging, co kriging with elevation, ordinary kriging and IDW among others U sing the same selection criteria as Chuanyan et al (2005), they strove to identify a single preferential approach for the entire study area for any time of the year. The type of variogram model for each krige was s elected subjectively by visual examination of the semivariogram cloud, whereas lag size and anisotropy were selected based on a systematic increment based brute force approach. They found detrended kriging having the lowest ME values for temperature and universal kriging for precipitation. GIDS produced the lowest MAE and RMSE in either case. T hey f ound ordinary kriging to be least suitable for temperature interpolation in their study area, identifying the GIDS method and detrended kriging as the best statistically equally well performing models For precipitation modeling, GIDS ordinary kriging, co-kriging, nearest neighbor, and IDW models were of equal statistical significance. One limitation of Nalder & Weins (1998) kriging models was use of static se arch neighborhoods, rather than adjusting the search neighborhood to optimize prediction error statistics. This was a necessity imposed by the small number of sampling stations available, which precluded trainingtest subset validation and forced them to r ely on cross validation as their sole independent means for model assessment. They note, in agreement with Boer et al (2001) that precipitation is substantially more difficult to model geostatistically than temperature. Ustrnul & Czekierda (2005) evaluat ed various kriging methods using contemporary GIS software to produce annual, seasonal and monthly air temperature maps for Poland based on a dataset of mean monthly temperatures collected from 168 stations within the country and 55 in

PAGE 56

56 the surrounding coun tries proximal to the Polish border. They selected residual (detrended) kriging as the best modeling approach for their study area and applied it to generate, among other applications, anomaly maps for warmest and coldest months between 1951 and 2000, and temperature probability maps based on the observational record. They emphasize d that caution is required when modeling microscale climate patterns, particularly in mountainous areas where predictor variables such as slope, aspect, land use and soil type are relatively more important than on meso and macroscales. Modeling on very small spatial scales therefore requires development of new, independent models based on regionspecific parameters. Moral (2010) provided an excellent summary of various geost atistical interpolation methods for precipitation modeling and applie d them to the Extremadura region of Spain which is describe d as heterogeneous in terms of elevation and land use. While sample station locations were distributed suboptimally, he was able to assess the skill of a variety of univariate and multivariate kriging approaches at predicting values at unsampled locations. He noted that univariate models yielded considerably cruder estimates due to low sampling densities Since no additional sampl ing stations could be identified, the models were complemented with digital elevation model ( DEM ) data, which were correlated to precipitation under the assumption that precipitation increases linearly with increasing elevation These multivariate models outperformed univariate ones Regression kriging and simple kriging with varying means were the best models, whereas ordinary kriging yielded the best results among univariate models. A study on GIS based spatial interpolation of precipitation in a highly h eterogeneous study area in upper Pakistan (Ashiq et al ., 2010) parallels some design elements of my study despite obvious differences in topography between the Himalaya n fringe and Florida. They assessed GIS based deterministic and geostatistical techniques to downscale a regional climate model

PAGE 57

57 with a spatial resolution of 0.4 degrees latitude by 0.4 degrees longitude to yield 250 m x 250 m mean monthly precipitation surfaces for a 1961-1990 baseline period. They based selection of the preferred model on crossvalidation RMSE statistics and coefficients of determination of regressions on modeled against observed mean precipitation at weather station locations. Geostatistical models outperformed det erministic models in almost all cases a common thread across most of the studies cited above. Ordinary kriging with a rational quadratic variogram model was selected as the best interpolator. E xtending this model by incorporating elevation data via co -kriging had seasonally inconsistent effects on RMSE. The final model was nevertheless co -kriging-based due to better performance during the more ecologically and socioeconomically relevant monsoon season. Unlike the previously cited studies, Ashiq et al (2010) also projected future precipitation at the end of the 21st century under two climate change scenarios using the same interpolation model they used for observed data. Students ttests were applied to assess statistical significance of modeled anomalies They support observations by Nalder & Wein (1998) and Boer et al (2001) that precipitation is substantially more challenging to model due to its inherently variable spatial and temporal distribution that GCMs and even some dynamical RCMs cannot process due to their limited resolution Of note is the authors statement that systematic errors in the RCM or GCM driving a geostatistical downscaling model cannot be compensated for by choice of interpolation technique. I claim, to the contrary, that the individual bias correction approach introduced in my study can compensate for such errors at least in part, provided they are consistent over time. Two white papers released by the National Center for Atmospheric Sciences [ NCAR ] specifically described how CCSM output data can be used in climatological analysis. Hoar &

PAGE 58

58 N ychka (2008) described a hybrid interpolation technique combining spline-based downscaling of CCSM outputs with regression analysis. A spline was applied to downscale CCSM outputs to match the spatial resolution of a climate mapping system known as PRISM (P arameter E l evation R egressions on I ndependent Slopes M odel), itself a constantly updated, interpolated model of climatic variable s that was used as ground truth Conversely, PRISM data were aggregated ( i.e. averaged) over each CCSM grid cell creating a n upscal ed PRISM surface that was then splined using the same model parameters as above to return it to its original spatial resolution. This produced a ground truth dataset of original PRISM data and a modeled dataset of aggregated and downscaled PRISM data, which were compared via linear regressions individual to each PRISM grid cell. G iven that the PRISM grid consists of 872,505 cells for the contiguous United States (a spatial resolution of 4.5 x 4.5 km), and linear regression models were unique for each season or calendar month mapped, in excess of 10 million regression s had to be produced in statistical software external to ArcGIS to cover the entire United States. Final downscaled es timates were calculated using simple slopeintercept equation solutions to each linear regression. Regression slopes and sum of squared error (SSE) statistics were used to produce spatially differentiated reliability estimates. Alternatives to thin plate s pline interpolation of CCSM data were not evaluated, nor were future projections. Thornton & Wilhelmi (2008) described procedures for production of temperature anomaly maps based on unmodified CCSM output data in ArcGIS which neither downscale nor correct for biases, but illustrate general anomaly patterns and trends of climatic change over large areas that can be evaluated for statistical significance by ttests with pool ed variances Their process is used in modified form for my study and described in more detail later in this chapter

PAGE 59

59 The main lesson from these diverse case studies was succinctly stated by Nalder & Wein (1998), who despite attempting to find a onemethod fits all model for their study area, prefaced t heir paper with the statement that no single method is optimal for all regions (p. 211) a sentiment that Moral (2010) shares Nevertheless, most studies share important conclusions. These include: Geostatistical models usually outperform deterministic models. Model performance is typically gauged by prediction error statistics, most commonly MAE or RMSE generate d by cross validation or training test sample validation Sparse or clustered distributions of sampling stations reduce the utility of univariate models, particularly kriging. Multivariate models usually outperform univariate models in mountainous regions, whereas their benefit in regions of low topographic relief is questionable. In those cases, more complex models do not necessarily yield better results, but require additional work and possibly additional software to perform. Ordinary kriging or detrended kriging methods are frequently preferred among univariate interpolators, in part due to their quality as best linear estimators and availability of prediction error statistics for unsampled locations that other methods cannot produce. Precipitation is considerably more challenging to interpolate due to its greater variability and discontinuous nature Modeling Philosophy The main objective of this study was to determin e how temperatures and precipitation in Florida will change through the end of the 21st century under three scenarios of climate change. Goals included performing this analysis at low cost using widely available desktop PC software allowing for rapid development of customizable derivative products for human health, socioeconomics, infrastructure, energy, ecology, hydrology, and conservation applications. For these reasons, ArcGIS was selected to produce calibrated, bias corrected projections via geostatistical downscaling of existing GCM outputs.

PAGE 60

60 Many of the challenges faced by the studies cited above do not apply to Florida, which presents its own unique set of challenges for climate variable interpolations due to its narrow peninsular shape In Florida, a dense network of meteorological stations recording climatic data on an ongoing basis exists and t opographic relief is low Therefore, univariate ordinary kriging and/or detrended kriging (using global or local polynomial trend surfaces) can be expected to perform well in Florida. My study incorporates many of the elements and methodologies of the reviewed studies and improves upon them with an independently developed calibration and bias correction process includin g two distinct interpolation steps. The result is a compromise between the convenience and expedience of univariate models while integrating some locationspecific information as reflected in how the properties of each sampling station affect their data values. Following extensive literature review and exploratory evaluation of several downscaling models a final development process was decided upon that can be broken down into eight steps: 1. Collection and preparation of two sets of climate data: Observed 19712000 climate normals from meteorological stations (ground truth) GCM -produced (modeled) climate data for 1971-2099 2. Exploratory and structural analysis of observed climate data 3. Creation of interpolated baseline models fitted to observed climate normals selected by prediction error statistics 4. Exploratory and structural analysis of GCM output data, including production of uncalibrated anomaly maps in the original model grid to test for statistical significance (modified ThorntonWilhelmi (2008) method) 5. Cr eation of least RMSE interpolations of GCM outputs 6. Sampling of least RMSE GCM interpolation and bias correction 7. Interpolating biascorrected GCM interpolation samples to yield downscaled future projections 8. Deriving products from downscaled, biascorrected projections

PAGE 61

61 I n brief, this process selects least RMSE geostatistical models, which are unique to each explored variable and month, to interpolate 19712000 CCSM outputs. The resulting temperature and precipitation surfaces are sampled at calibration locat ions (meteorological stations) for assessment of GCM biases and identification of model specific bias adjustment constants. These constants are applied to each point sample, adjusting their value to remove GCM biases on a per s tation basis (individual bias correction). A second interpolation analogous to a previously produced baseline model of 1971-2000 observational data for the climate variable and month of interest, is performed between calibration locations using bias corrected values to yield the final, downscaled surface. For the baseline period, observed and individually biascorrected map surfaces will be identical. A lternatively, bias correction can be used on a study area mean -basis using a single adjustment constant that is applied to all calibration locations or, if operating with rasters, applied to the raster surface clipped to the boundaries of the study area (average bias correction). This single constant is equal to the mean GCM bias within the study area. In this case, mea ns of observed and modelderived interpolation surfaces will be identical for the baseline period, but values at individual locations will vary. This approach makes fewer assumptions about causes of differe nces in observed climate between meteorological st ations, but also yields a much more generalized surface that reduces but does not entirely remove regional GCM biases. Bias correction constants, for average as well as individual methods, are carried forw ard from the baseline period to future projections Options for visualization of results include temperature and precipitation surfaces, anomaly maps, and profiles along transects, which may be annotated and combined with other spatial information in GIS layers as preferred by the end user. Uncertainty estimates are produced via

PAGE 62

62 prediction standard error mapping T ests for statistical significance of changes over time are performed on unmodified CCSM data in their original model grid using Students ttests While applicable to any month or combination of months, this study focuse s on projections for the extreme seasons, represented by the months of January and July. It is during these months that much of Florida experiences its lowest and highest temperatures, respectively (Henry et al. 1994), exerting maximum physiological temperature stress on biota. Further, evaluation of these months allows for direct comparison with results of the Coenen (2005) study on recent climatic variability in Florida as determined by analysis of surface lake water temperatures. I ncorporation of elevation slope and aspect data was not considered necessary, since the highest point in the state is a mere 10 5 m above sea level and slopes a re with few exceptions, low Assuming a mean environmental lapse rate of 6.49C/1000 m, this corresponds to an elevation caused temperature differential of 0.67C, which is negligible compared to latitudinal and near coastal temperature gradients. Further this information is in part encapsulated in the individual bias correction procedure, which is based on the assumption that differences in temperatures between weather stations are a reflection of their location and topography. Applicability beyond Florida This methodology was intentionally designed to be flexible and pose minimal requirements on input data. Therefore, it is proposed that it can be applied, with some modifications, to any landmass with a sufficiently dense network of weather stations or appr opriate substitutes, such as satellite based data or lake surface temperature records (Coenen, 2005), to build calibrated and bias corrected interpolation models M odel accuracy will suffer with increasing distance or clustering between sampling stations ( Rosenthal et al ., 1998). CCSM output data is readily available for many other land and atmospheric variables that can be incorporated into more complex kriging or regression models.

PAGE 63

63 Where topographic relief is significant, a digital elevation model (DEM) LIDAR altimetry or field sampling using global positioning system (GPS) may be required to incorporate adiabatic and orographic effects into the model, necessitating more complex multivariate techniques (Boer et al ., 2001) that may be performed by cokrig ing within ArcGIS or using kriging extensions requiring statistical software such as S or R. Chuanyan et al (2005) suggest that incorporating topographic slope and aspect are important to account for local biases due to north -facing slopes being cooler th an south facing slopes at the same elevation in the northern hemisphere. Multivariate analyses are not discussed outside the literature review in my study, but are discussed in great detail by Moral (2010) with a focus on pre cipitation. It should be noted that Nalder & Wein (1998) cite disagreement between studies on the importance of incorporating elevation into models in non mountainous regions, and some have questioned use of the standard adiabatic lapse rate to determine altitudinal temperature gradien ts. Construction of Downscaled Climate Models for Florida Prerequisites: Input Data and Analysis Software The GCM providing climate projections for all downscaling models developed for my study is the Community Climate System Model CCSM 3.0, released in 2004 (Collins et al. 2006). Two major advantages of CCSM 3.0 are free availability of outputs in point shapefile format and a high spatial resolution of approximately 1.4 degrees latitude by 1.4 degrees longitude which is the second highest among the 23 m odels widely utilized by the IPCC for its 4th assessment report (IPCC, 2007a). The analysis s oftware used for this project was ESRI ArcGIS 9.3 under an ArcView license, including Geostatistical Analyst, Spatial Analyst and 3D Analyst extensions. The Maplex labeling engine was used for label placement and formatting. Third -party plugins used were Hawths Analysis Tools Version 3.27 by Hawthorne L. Beyer (2007) and ET Geo W izards

PAGE 64

64 Ver sion 9. 9 by Ianko Tchoukanski (2009). Microsoft Excel 2007 was used to complement ArcGIS for some attribute table calculations. NASA WorldWind and Google Earth software applications were used to retrieve satellite imagery and aerial photography for visual guidance, and to aid in interpretation of results, but were not part of the modeling workflow Step 1: Data Collection and Preparation National Climatic Data Center (NCAR) climate normals O f ficial NCAR climate normals for 1971 -2000 were obtained from the S outheast Regional Climate Center [SERCC] for weather stations in Florida, southern Georgia and southern Alabama (SERCC, 2010). Data from stations beyond the political borders of Florida were included to minimize interpolative edge effects (Brown & Comrie, 2002), particularly in the narrow panhandle. Due to their separation from Florida by more than 100 k m wide ocean channels, data from the Bahamas and Cuba were not considered. Metadata for each weather station were carefully evaluated and any station with extensive data gaps excluded from further analysis One hundred three stations were selected for inclusion among these 83 from Florida, 8 from southern Alabama and 12 from southern Georgia (Figure 4-1). A p oint shapefile of station locations w as created f rom SERCC metadata, which had a level of precision of 0.5 minutes of arc, or roughly 0.93 km. Temperature means, average minima and average maxima were copied from SERCC into separate attribute tables in an editing session and converted from degrees Fahren heit to Celsius. Precipitation normals were processed in the same way and converted from inches to millimeters T he spatial arrangement of selected weather stations is such that it allow ed for interpolation for nearly all of Florida except for a narrow sli ver along the Palm Beach County coast, where values had to be extrapolated. T heir spatial distribution was further evaluated using the Average Nearest Neighbor function of the ArcGIS Spatial Statistics Toolbox to determine

PAGE 65

65 w hether interpolation models wo uld benefit from declustering. T his was not found to be the case as the very slight degree of clustering indicated by a zscore of 1.51 was not found to be statistically significant (Figure 4 2 ). In order to apply this methodology to study areas where sam pling stations are significantly clustered, a declustering step may be required. C o mmunity C limate System Model (CCSM) output data CCSM output data for the variable ts ( mean surface temperature) were obtained in ArcGIS point shapefile format from the NCAR GIS Initiative via their GIS Climate Change Scenarios web portal, http://www.gisclimatechange .org for scenarios B1, A1B, A2, and the 20th Century Experiment (NCAR GIS Initiative, 2010 a). Free registration and acceptance of a data disclaimer are prerequisites for data access. Each shapefile point marks the centroid of a CCSM grid polygon (Wilhelmi, pers. comm. 30 January 2008). To precl ude interpolation artifacts that commonly occur at map edges from exerting i nfluence on projections for the Florida landmass (Brown & Comrie, 2002), CCSM data were obtained for a generous extent stretching between 23N and 33N latitude and 78 W to 90 W longitude, incorporating a total of 72 grid centroids. This provides a substantial buffer of land and water based cells surrounding the political borders of Florida (Figure 4-3). While each model run is equally plausible, ensemble averages are preferable in modeling since they average runspecific initial conditions and simu lated future weather events, resulting in a signal that represents the effects of changing boundary conditions more clearly. Therefore, a ll CCSM outputs used in this study were ensemble averages comprised of 5 independent model runs under the same scenario and boundary conditions, but different initial conditions (NCAR GIS Initiative, 2010b). The initial conditions of each ensemble member were determined by the point at which it branch ed from a 500 year control run that operate d under constant boundary cond itions using the year 1870 as a baseline, and was therefore subject only to a narrow range of

PAGE 66

66 s imulated natural variation. This branching at different points (elapsed years 360, 380, 400, 420 and 440) produced 5 slightly different models for conditions dur ing 1870-1999 known as the 20th Century Experiment. Unlike the 500 y ear control run, the 20th Century Experiment was subject to temporally and spatially varying forcings reflecting observed changes in insolation, atmospheric sulfates, ozone, GHGs volcanic activity, and other quantities during the 1870-1999 period (NCAR GIS Initiative, 2010b). At the endpoint of each of the 1870-1999 models i.e. the year 2000, a second branching occurred at which point boundary conditions changed according to expected forcings under each of the three IPCC benchmark scenarios ( Figure 4-4). E ach downloaded shapefile included 20 years of ensemble temperature data for one scenario (e.g. B1) and one month (e .g. January). Problems were discovered with the precision of precipitation flux ( pr ) shapefiles, which lacked sufficient significant digits to construct meaningful models. For this reason, pr files were retrieved in comma delimited text format, which w as n ot subject to the same significant figure limitations, but required additional steps to produce shapefiles ready for GIS analysis. The text files were imported into Microsoft Excel using the text to columns tool and combined to yield two spreadsheets (on e for each month) containing data for the entire study period Alternatively, separate spreadsheets for calibration and projections may be created based on user preference. Averages and standard deviations for the time periods used for analysis in my study 1971-2000, 1990-1999, 2020-2029, 2040-2049, 2060-2069 and 2090-2099 were calculated and converted from units of precipitation flux to precipitation in mm/month according to equation 2-1. mm/month = precipitation flux 86 ,400 number of days in the month (2 -1) The Excel precipitation files, in .xls format, were imported into ArcGIS via the Add XY Data function, assigned the CCSM Geographic Coordinate system (based on an earth model

PAGE 67

67 assuming a perfect sphere with a radius of 6371.22 km (Thornton & Wilhelmi, 2008)), and finally exported as point shapefiles in the Albers coordinate system, which is commonly used for Florida specific applications and is the standard coordinate system used by the Florida Geographic Data Library [FGDL]. Analogous to the pr ecipitation file workflow, multi year average calculations and conversion from CCSM Geographic to the Albers coordinate system were applied to temperature shapefiles. Degrees Kelvin was converted to degrees Celsius to facilitate interpretation. It was foun d that the ArcGIS Spatial Analyst extension, which was used in combination with Geostatistical Analyst for this study, could not produce interpolated raster files from input data in the default double precision format. For this reason, floating point preci sion was chosen as the preferred attribute table format. C alculation of two different baseline periods was intentional: The 1971-2000 baseline was used for calibration and biascorrection in the downscaling process in accordance with a World Meteorological Organization [ WMO ] recommendation advocating 30 y ear averages to obtain reliable baseline data upon which predictions can be based (WMO, 1967). The 1990 1999 b aseline was used for tests of statistical significance of temperature and precipitation anomalie s as projected by unmodified CCSM outputs. The shorter 10year time horizon maintain s similar variances about interval means even under accelerating climate change, allowing use of pooled variance t te sts. Use of different baseline periods for different pu rposes is common in the literature. For example, the IPCC (2007a) frequently uses 20year baselines, whereas Baigorria et al (2007) initially considered a 90year baseline for assessing spatial variability in southeastern U.S. r ainfall, but ultimately reduced it to 15 years due to concern s about climatic shifts during the instrumental record. It is generally not advisable to use baselines extending over more than 30 years if climatic trends are detectable or suspected

PAGE 68

68 Step 2: Exploratory Analysis of O bserved Data Prior to developing downscal ing models of CCSM data, it is important to develop present day baseline reference models interpolated temperature and precipitation surface s based on observed climate normals t hat serve as bases for CCSM calibration bias assessment and comparison with future projections, e.g. for the calculation of anomaly maps. The parameters of the baseline reference models are also used in the second interpolation of the individual bias correction procedure described i n step 7. D evelopment of this observed reference surface was preceded by extensive data exploration using the Explore Data options provided by ArcGIS Geostatistical Analyst, results of which provided outlier detection and guidance for which types of downscaling approaches and semivariogram models may be appropriate Th is include d evaluating data in original and transformed form via histograms, Voronoi mapping, trend analyses semivariogram clouds and normal QQplots This process is illustrated in figures 4-5 through 4-9 and discussed below for the example of Florida mean January temperatu res The same procedure was repeated for all variables and months under investigation The January mean temperature histogram ( Figure 4-5) exhibits a bimodal distribution, with values in the left most columns originating primarily from stations in the northern and northwestern areas of the state and values in the central and right -most columns being contributed by stations in the central and southern peninsula, where a strong, approximately latitudinal temperature gradient exists that is slightly skewed towards the northeast as isothermal lines approach the Atlantic coast, presumably due to the influence of the Gulf Stream. This pattern is reflected in s imple (Figure 4 -6A), m ean (Figure 4 -6B) and m edian (Figure 4 6C) Voronoi m aps, which all show a gradual north to -south progression from low (red) to high (blue) values T he t rend a nalysis ( Figure 4-7) and semivariogram cloud and surface (Figure 4-8)

PAGE 69

69 suggest strong autocorrelation in the same direction. It follows that detrending should be strongly considered for January mean temperature models. O nly a single station in the north of the study area was identified as not clustering with its neighboring cells ( Figure 4 6D). Review of the source data for the corresponding station did not reveal any data entry errors or other obvious reasons for this single outlier, and it was retained in the analysis accordingly. Entropy ( i.e. variability among a cell and its neighbors compared to variability of the entire dataset; Figure 46E) is low throughout the state except in the far western panhandle E ven though the histogram ( Figure 4-5) exhibits low skewness, its bimodal nature, platykurtic k urtosis and pronounced deviation from normality at the upper and lower tails of the Normal QQplot ( Figure 4-9) that did not improve with transformation of the data, led me to conservatively opt for ordinary kriging and detrended ordinary kriging for development of the baseline model, as this method places the fewest assumptions on the data while retaining many of the benefits of geostatistical modeling that deterministic models lack (ESRI, 2003). While a normal score transformation under s imple kriging may have been an option to address the non normality of the data, ArcGIS encountered data transformation errors during implementation removing 18 of 103 sampl es as a result T his loss of calibration reference points was not considered an acceptable compromise to achieve normality and therefore normal score transformed data under simple kriging was not further entertained For consistency, ordinary kriging or detrended ordinary kriging were selected for all steps involving interpolation b etween weather station location s with vari ogram models, anisotropy, lag size and neighborhood size optimized for each climate variable and month.

PAGE 70

70 Step 3: Creating the Baseline Reference Model from Observed Data Having thus prepared and explored the observed 1971-2000 climate normals, ordinary and detrended kriging prediction models were generated in ArcGIS Geostatistical Analyst to identify best fit models for present day temperature and precipitation patterns in Florida. Development of t hese was in part informed by data explo ration, and in part created by a constrained brute force approach, whereby m odel parameters that were changed, one at a time, included the semivariogram model (Spherical, t etraspherical, Gaussian, h ole e ffect and K-bessel, as these provided the best visual fit to the semivariogram cloud s), lag size and the number of lag bins (with their product approximating half the distance between the most distant points in the database for which the Euclidian distance line passed primarily over land; ca. 360 km), isotro py vs. anisotropy, and enabling firstand second -order trend removal as suggested by the histogram analysis, trend analysis, and Voronoi mapping. Software calculated values were retained for nugget, partial sill, and direction and shape of anisotropic ell ipses. T he resu lting models were compared via ArcGIS generated prediction error statistics produced by crossvalidation, giving preference to the model with the lowest root mean square prediction error, a standardized root mean square prediction error clos est to the value of 1, a standardized mean value closest to the value of 0, and the difference between root mean square and average standard errors approaching 0. T hese statistics indicate that predictions are as close to the observed values as possible, t hat they are unbiased, and that the variability of the predictions as measured by the prediction standard errors is assessed correctly (ESRI, 2003). As the final step in the model fitting process, search neighborhoods were optimized to cross v alidation statistics. While doing so reduces the utility of crossvalidation as an independent means for verification of the model (Nalder & Wein, 1998), several steps were taken to address this concern: First, all models used the same neighborhood type a single

PAGE 71

71 sect or. Second, determination of the number of sample points to include in the neighborhood was the last step in the model building process, after crossvalidation already confirmed validity of other model parameters. Third, the range of total and least accept able members within a neighborhood was restricted to no more than 20. Fourth, modeling parameters including optimized neighborhoods w ere retroactively checked by means of a training test dataset validation using a ratio of 75:25. Results are described in Chapter 5. Step 4: Analysis of CCSM Output Data Data exploration Similar to the exploratory steps taken to guide creation of the baseline reference model from observed climate normals, CCSM output data from the calibration perio d warrants exploration via the functionality of Geostati stical Analyst to inform model development in this case the first interpolation of the individual bias correction procedure, and the only interpolation of the average bias correction procedure (see s teps 6 and 7). Th e data exploration process is illustrated in figures 4-10 through 4-14 and discussed below for the exampl e of mean modeled January temperatures. The same process applies to any other month, climate variable or study area under investigatio n. It should be recalled that the extent of the 72 CCSM grid cells is substantially larger than that of the observed data, such that results of steps 2 and 4 are not directly comparable. Histogram ( Figure 4-10) and normal QQplot ( Figure 4-11) show that 19 71-2000 CCSM outputs are clearly non -normally distributed, and no data transformation improved this trend. Voronoi maps (Figure 4-12) and trend analyses ( Figure 4-13) indicate the same general near latitudinal temperature gradient that was identified in th e observed climate normals, although the information that can be derived from the Voronoi maps is limited due to the low spatial resolution of the grid polygons. Simple (Figure 4 -12A) and cluster (Figure 4 -12D) maps appear

PAGE 72

72 to primarily outline the land to ocean transition. The cluster map shows the two polygons corresponding to the southern top of the peninsula as outliers, likely also reflecting a combination of the land to ocean transition and the overall thermal gradient over the study area. E ntropy (Fig ure 4 -12E) is elevated for grid cells that correspond to south and southeast Florida, as well as coastal Georgia. The scattered appearance of the semivariogram cloud ( Figure 4-14) reflects that CCSM centroids are arranged along a regular grid Autocorrelat ion is strong. Despite the large difference in extent, t he semivariogram surface exhibits similar trend s to the observed data. To summarize, the non normality of the CCSM data warrants use of conservative methods similar to those applie d to the observed data, namely ordinary kriging. The existence of a clearly defined temperature gradient warrants exploration of detrended kriging models. Despite the dissimilarity in extent, patterns appear similar between observed and CCSM derived data. M odified ThorntonWilhelmi m ethod to v isualize CCSM -p rojected anomalies and test for statistical significance The most basic way to evaluat e projected future trends of climatic variables is calculation of anomaly maps based on raw, unmodified CCSM output da ta which illustrate changes relative to a baseline. This is convenient not only to understand patterns in CCSM projections for the study area prior to downscaling, but also makes tests for significance of anomalies easy to perform. For this project, a modified version of the Thornton Wilhelmi (2008) methodology was applied to evaluate mean temperature and precipitation anomalies relative to a 1990-1999 baseline based on data from the 20th Century E xperiment. Ensemble averages were utilized throughout. The ThorntonWilhelmi method was modified in four ways. First, the surface temperature parameter ts rather than air temperature ( tas ) was used, since one proposed

PAGE 73

73 application of the results of this study i s determination of ecological impacts, for which surface conditions are most pertinent. Secondly, evaluation of precipitation changes via model parameter pr ( precipitation flux) converted to mm/month, was added. Thirdly, c alculation of three month (season al) averages was abandoned in favor of evaluating January and July individually for consistency Finally 10year rather than 20 year averages were compared to maintain similar variances about interval means allow ing use of pooled variance t tests ( df =18) in assessing statistical significance of temperature anomalies. I f future temperature change a ccelerat es as suggested by most global studies, variances would increase proportionally to the number of years being averaged during the latter decades of the 21st century. Ten year averages represent a compromise between minimizing this effect while retaining a large enough sample size to obtain representative averages resistant to outliers (keeping in mind that each temperature value is itself an ensemble averag e that minimizes the influence of outliers ). T he downloaded and re projected shapefiles each containing 20 years of CCSMprojected temperature or precipitation data were imported to ArcMap. Ten year averages and associated standard deviations for the intervals 1990-1999, 2020-2029, 2040-2049, 2060-2069 and 20902099 were calculated into new attribute table fields via the Field Calculator. Subsequently, spatial join s w ere performed with a shapefile of the CCSM grid obtained from the GIS Climate Change S cenarios web portal, essentially converting the downloaded point shapefiles back into the original model polygons. A ttribute tables for the resulting shapefiles were joined and exported into a new shapefile E xtraneous fields were removed via Hawths Tools Delete Multiple Fields functionality such that a single, new polygon shapefile containing all five 10 y ear averages and their standard deviations was created. New fields were added for anomaly calculation s as well as the significance level of one taile d Students t-t est s with pooled variances

PAGE 74

74 and 18 degrees of freedom, which were performed in Excel for expedience, but can also be calculated via the Field Calculator function. Since calibration and bias correction adjustments in the downscaling process do not vary with time, statistical significance of anomalies in the raw climate data implies significance in the downscaled climate data as well. Ttests were performed on a per -polygon basis based on the means of each 10 -year sampling period and their pooled variance (df=18) Therefore the difference in spatial extent between the 72 CCSM polygons and the political boundaries of Florida is inconsequential. R esults are discussed in Chapter 5. Step 5: Creation of LeastRoot Mean Square Error (RMSE) I nterpolations of CCSM O utputs GCMs are designed to make projections at continental to global scales (IPCC, 2007a). R egional scale deviations between observed and modeled climate are frequently noted in hindcast analyses (Collins et al ., 2006). To produce downscaling models capable of projecting future temperatures and precipitation at high spatial resolution s, GCM interpolation followed by bias correction is required to compensate for such regional deviations In part guided by data exploration and in part by a systematic brute force approach, leastRMSE kriging models unique to each explored variable and month wer e identified to interpolate 1971-2000 CCSM outputs obtained from the 20th C entury E xperiment ensemble average for years 1971-1999, and from scenario A1B for the year 2000. The same modeling philosophy as was previously described for observed data was appli ed here, but an exponential semivariogram model was also entertained, appearing to be a plausible alternative judging by visual examination. Lag sizes and numbers were adjusted to account for the grid like arrangement of CCSM polygon centroids and the ir larger extent. It is very important to emphasize at this point that, unlike for model development based on observed data, prediction error statistics provided

PAGE 75

75 by ArcGIS during crossvalidation are not Step 6: Calibrating CCSM Outputs to the Obs erved Data: Bias Correction suitable to assess the quality of CCSM interpolation mode ls, since they are based on biased temperature or precipitation data that this procedure aims to reduce or eliminate. Since the degree of bias cannot be quantified a priori it must be assessed after each downscaling model has been run. To do so, each inte rpolated map surface was sampled at weather station locations using the ArcGIS Prediction function or GA Layer to Points tool The model that produced the smallest RMSE following average bias adjustment was selected as the method that deviates least from the 1971 -2000 observed climate normals and is therefore best suited for reproducing observed climate patterns over Florida. Multivariate downscaling was not considered necessary since a majority of the 72 gr id cells above Florida and surrounding areas are ocean based and therefore have an elevation of zero. Land based cells are located in the Gulf and Atlantic coastal plains, where relief and elevations are minimal. Two basic calibration options were developed for this study : average bias correction and individual bias correction, the latter of which was selected for production of all final temperature, precipitation and anomaly maps Both methods are based on assessing and adjusting for differences bet ween 19712000 climate normals ( observed climate) and interpolated GCM outputs averaged over the same timeframe (the calibration period), but differ in underlying assumptions and implementation Average bias correction Average bias correction (Figure 4-15) results in the removal of any absolute bias introduced by CCSM outputs on downscaled projection models for the Florida landmass For example, if interpolations based on raw CCSM data p rojec t Florida to have been, on average warmer during winter than it really was during the calibration period, the GIS model is adjusted by subtracting the appropriate value such that observed and modeled means for the study area, as

PAGE 76

76 sampled and averaged over the 83 weather station locations within the political borders of Florida are equal. In this case, use of sample data from beyond the boundaries of the study area is undesirable, as they would skew the mean calculation. The average difference between modeled /interpolated and observed temperatures at the sample points is defined as the fixed bias adjustment term k which is specific and constant to each kriging model and subsequently applied thereto by subtraction or addition, depe nding on the sign of k Essentially, this adjusts the interpolations in such a way that the modeled temperature sampled at and averaged over the 83 calibration locations is identical to the observed average, reducing absolute bias to 0. Being a mean based procedure, it frequently reduces but does not eliminate observed model d ifferences at individual sample points, and, unlike individual bias correction, produces very generalized map surfaces unless combined with kriging extensions to incorporate topographi c realities into the model. Individual bias correction Individual bias correction ( Figure 4-16) is founded on the assumption that temperature differences between stations have a physical basis and are a direct result of their relative locations, distance from the coast, altitude, aspect, and other geographical factors, many of which are not expected to change substantially, if at all, as a result of climate change. In other words, i ndividual bias correction implicitly takes topography into account to the extent that it is embodied in observed temperature differences between weather stations. Therefore each station location is assigned its own, individual adjustment term kSi that is derived from the difference between model projected and observed temperature at each station during the calibration period. This is equivalent to resampling and adjusting th e CCSM output in such a manner that observed and modeled temperature surfaces during the calibration period are identical, that is, the root mean square error is reduced to 0. Unlike average bias correction, sampling points outside the

PAGE 77

77 boundaries of the st udy area can be included, reducing edge effects. The combination of station specific adjustment factors and subsequent downscaling yields substantially more detailed map surface s than is the case for average bias correction. Step 7: Applying Bias Correctio n to Future Projections For either bias correction approach correction factors and downscaling model s selected during the calibration period are carried forward and applied to each future projection, allowing only automatically calculated parameters such as the nugget to vary. This reflects th e assumption that CCSM bias is caused by its limited spatial resolution that poorly delineates the coastline of the southeastern United States, and is not subject to change significantly, if at all, over time. It is a common assumption in climatology that if a model succeeds at reproducing observed climate at high fidelity, it is also capable of projecting future climates with a high degree of confidence, especially when using ensemble data (As hiq et al ., 2010). All CCSM outputs were prepared and interpolated as was done for calibration For individual bias correction, least RMSE downscaling model s selected during calibration were applied to interpolate CCSM projections, which w ere subsequently sampled at each weather station location via the Prediction or GA Layer to Points function s The resulting point samples were adjusted by each stations individual adjustment term, kSi. The final map was produced by interpolating the adjusted temperature values between s tation locations analogous to the production of the observed baseline model allowing for direct comparison and calculation of highresolution anomaly maps, allowing delineation of vulnerability zones within the study area. To facilitate map annotations and calculations of anomaly maps, the final geostatistical layers were exported to raster format with a cell size of 500 m x 500 m and the extent of the political borders of Florida. These rasters were then smoothed using the Neighborhood Statistics function with a circular search area measuring 15 cells. Some maps required greater degrees of

PAGE 78

78 smoothing to eliminate artifacts. Finally, the rasters were clipped to the outline of Florida, annotated with contour lines and labels, a nd given a consistent symbology format, as seen in the final maps presented in chapter 5. Step 8: Deriving Projections for Temperature Minima and Maxima T he GIS Climate Change Scenarios web portal only provides mean temperature data, but minima and maxima are equally, if not more important for projections of ecological impacts energy demand, human health, etc. Therefore, deriving estimates for minimum and maximum temperat ures is essential for applied modeling studies. Literature is somewhat contradictory regarding trends in diurnal temperature variability. Easterling (1997) note d that nighttime temperatures have increased at a greater rate than daytime temperatures whereas IPCC (2007a) concluded that Easterlings assessment was erroneous due to data limit ations and that maxima and minima are expected to increase at approximately equal rates in the future Based on the IPCCs position, a constant range between mean monthly minima and maxima was assumed for my study I ndividual bias correction factors kSi w ere further modified to also incorporate the difference between observed means and average minimum and maximum temperatures for each weather station, respectively. The resulting point shapefiles were processed analogously to mean temperature and precipitation analyses using individual bias correction If alternative hypothes e s of asymmetric warming are to be explored, this can simply be achieved by adjusting kSi constants accordingly

PAGE 79

79 Figure 4-1. Location of reference weather stat ions in Florida, Alabama and Georgia.

PAGE 80

80 Figure 4-2. Results of the Average Nearest Neighbor operation on weather station locations, establishing that their distribution is statistically random (screenshot) Figure 4-3. CCSM model grid and grid centroids over the southeastern United States.

PAGE 81

81 Figure 44. Illustration of difference s between CCSM ensemble members. [Reprinted with permission from NCAR GIS Initiative (2010b), Control Runs (Fig. 3)].

PAGE 82

82 Figure 4-5. Histogram of observed 19712000 mean January temperature normals for the 103 reference weather stations (screenshot)

PAGE 83

83 Figure 4-6. Voronoi maps illustrating geostatistical patterns in the data from the 103 reference weather stat ions for mean January temperatures A) Simple. B) Mean. C) Median. D) Cluster. E) Entropy. A B C D E

PAGE 84

84 Figure 4-7. Trend analysis of observed 1971-2000 mean January temperature normals for the 103 reference weather stations, indicating a firstor second order latitudinal trend previously identified in lake temperature data by Coenen (2005).

PAGE 85

85 Figure 4-8. Semivariogram cloud and semivariogram surface of 1971-2000 January temperature normals. Figure 4-9. Normal QQ-plot of observed 1971-2000 mean Janua ry temperature normals for the 103 reference weather stations.

PAGE 86

86 Figure 4-10. Histogram of CCSM -generated 19712000 mean January temperatures for the 72 grid cells bounding and including Florida.

PAGE 87

87 Figure 4-11. Normal QQ-plot of CCSM -generated 1971-2000 mean January temperatures for the 72 grid cells bounding and including Florida.

PAGE 88

88 Figure 4-12. Voronoi m aps illustrating geostatistical patterns in the CCSM data over the calibration period for mean January temperatures A) Simple. B) Mean. C) Median. D) Cluster. E) Entropy. F) Albers projected reference map illustration position of grid cells relative to the southeastern U.S. A B C D E F

PAGE 89

89 Figure 4-13. Trend analysis on the CCSM -generated 1971-2000 mea n January temperatures for the 72 grid cells bounding and including Florida.

PAGE 90

90 Figure 4-14. Semivariogram cloud and semivariogram surface of 1971-2000 CCSM outputs for mean January temperatures

PAGE 91

91 Figure 4-15. Average bias correction workflow summary for calibration. The left branch contains observed climate normals, the right branch CCSM outputs, whereas the center branch details determination of bias adjustment factors and the selection criterion for the best model. Download 1971-2000 Climate Normals Sample projected temps at each station for each model via Prediction function. Average temperature across all stations Average temperature across all stations for each model Calculate difference = adjustment factor k for each model Apply k to remove average bias from each downscaling model Select the downscaling model with the smallest RMSE Produce suite of downscaling models using ordinary kriging Download 1971-2000 CCSM Outputs

PAGE 92

92 Download 1971-2000 Climate Normals Sample projected temps at each station via Prediction function Calculate difference for each pair = adjustment factor kSi for each individual station Use least-RMSE downscaling model on prepared CCSM data Note observed temps at each station Figure 4-16. Individual b ias c orrection workflow

PAGE 93

93 CHAPTER 5 CLIMATE CHANGE PROJECTIONS FOR FLORIDA THROUGH 2100 GENERATE D BY GIS BASED DOWNSCALING OF A GENERAL CIRCULATION MODEL Florida ( Figure 5-1) consists of a long and narrow up to 235 km wide peninsula and a panhandle that marks the easternmost exten t of the East Gulf Coas tal Plain physiog raphic region (Fenneman, 1917). It is bordered by Alabama and Georgia to the north, the Atlantic Ocean to the east, the Gul f of Mexico to the west and the Florida Straits to the south. Its highest elevation is Britton Hill (Walton County) near the Alabama border, at 105 m. While definitions of the boundary between panhandle and peninsula vary, Madison and Taylor counties, and all areas to their west, are attributed to the panhandle, whereas areas to the ir east and south are counted as part of the peninsula for descriptive purposes in this study. Including the panhandle Florida spans over 5.3 degrees longitude, or approximately 582 km between the Alabama border and the Atlantic coastline near Fernandina Beach (Nassau County). The peninsula is separated from the Florida Keys, an archipelago to its south, by Florida Bay, a shallow estuary receiving runoff from the Everglades and f lood control structures in south ern Florida. T he distance between the Georgia border near Boulogne ( Nassau County ) and Key West ( Monroe County ) is ca. 6.2 degrees latitude, or approximately 700 km. Floridas westernmost point is at 8738W in Escambia County, its easternmost at 800W at Palm Beach ( Palm Beach County) its northernmost at 3 10N in Nassau County, and its southernmost at 2427N in the uninhabited islands to the west of Key West ( Monroe County ). Literature Descriptions of Late 20th Cent ury Climate of Florida1A brief description of late -20th century climate conditions in Florida follows. It is intended to familiarize the reader with major temperature and precipitation patterns during January and 1 This section is an updated and expanded version of a discussion originally written for Coenen (2005), to which the author retains copyright.

PAGE 94

94 July, and also serves as a qualitative mea ns to assess and verify downscaling model capabilities by mutual comparison, supplementing quantitative analyses. This discussion is primarily based on three sources: Morton Winsbergs Climate of Florida (1990), The Climate and Weather of Florida by James Henry, Kenneth Portier, and Jan Coyle (1994), and James Henrys chapter on Weather and Climate in the Water Resources Atlas of Florida (Fernald & Purdum (eds.), 1998). Note should be taken of differences in baseline periods between the sources: 1961 1990 for Henry (1998) and 1957-1986 for Winsberg (1990). Henry et al (1994) do not explicitly state their baseline, although statements in the text allude to them referencing the entire available instrumental record. In Florida, a comprehensive network of meteorological stations was established by 1948 (Henry et al ., 1994), although individual stations existed at least since 1903 (SERCC, 2010). Considering Coenen (2005) found only a marginally significant warming trend restricted to winters, and others (Henry, 1998; Marshall et al ., 2004) found no evidence for recent climate warming in Florida, differences in baseline period choice are not expected to introduce strong bias, and all three sources should be regarded as generally representative of Floridas late 20th century climate. Since the sources use imperial units of measurement, they were converted into SI units and rounded. The Kp pen climate classification system ( Figure 5-2A ), divides Florida into two segments : The panhandle and northern three -quarters of the peninsula are considered humid subtropical (Cfa), whereas the southern quarter is a tropical savanna, or tropical wet and -dry region (Aw) (Henry 1998). The Thornthwaite system ( Figure 5-2B ), which incorporates the balance of precipitation and potential evapotranspiration, recognizes three distinct zones: A humid mesothermal panhandle and northern peninsula, a humid tropical southern tip, and a subhumid mesothermal zone in between (Henry 1998).

PAGE 95

95 Zonal boundaries slope from the southwest to the northeast due to the influence of the Gulf Stream Its location immediately offshore Floridas east coast strongly influences regional climate and results in temperatures along the eastern seaboard exceeding those in the interior and the western seaboard at equal latitudes (Henry 1998). Both Kppen and Thornthwaite classific ation schemes are rather coarse, lacking sufficient resolution to resolve climatic differences between the northern peninsula and the panhandle. They cannot capture substantial seasonal and diurnal intrazonal gradients between coastal and inland locations (Henry et al 1994) due to differential heat ing between land and ocean. Further, t hey cannot account for the fact that Florida s climate is highly variable intra and interannually, particularly for winter temperatures and precipitation (Henry 1998). Surface Air Temperature As described by Winsberg (1990), major controls on air temperature within Florida are latitude and distance from the coast. The temperature mediating effects of the Atlantic Ocean and, to a lesser degree, the Gulf of Mexico restrict diurnal and annual temperature ranges in coastal areas, wi th amplitudes increasing inland. Large lakes such as Lake Okechobee, similarly moderate the temperature of their immediate surroundings. Annual mean temperatures range from 19C to 26C, a gradient of slightly more than 7 C between the northwestern panhan dle and Key West. The statewide mean is 22 C (Henry et al ., 1994). During January, the coldest month for most parts of Florida (Henry et al ., 1994), latitudinal temperature gradients predominate, and the signature of the Gulf Stream becomes clearly apparent. The difference in average maximum January temperatures between Jacksonville and Key West is 5.6C according to Winsberg (1990), a value that may be somewhat conservative based on 19712000 climate normals used as the baseline for my study (6.6C). Temperatures reach average maxima of 16C in southeastern Escambia County to 25C along the border of

PAGE 96

96 Miami -Dade and Monroe counties. Interior locations exceed coastal maxima by 1C to 2C. Average minima range from 3C at Fountain in Bay County to 18C at Key West. Minima in the states interior are lower by up to 4C compared to coastal locations (Henry et al ., 1994). During July the warmest month in nearly all parts of the state, temperatures reach average highs between 31C and 33C with very little region al variability (Henry et al ., 1994). Mean July maxima in Key West and Jacksonville for instance, differ by only 0.6C (Winsberg, 1990). Summer sea breezes reduce maxima along the east coast by up to 2C compared to the interior of the state, but this effect does not typically penetrate far inland (Winsberg 1990). At night, the interior cools more than coastal areas, resulting in greater diurnal temperature range s with increasing distance from the coast July minima show greater latitudinal differentiation, with a range from 21C in northern Baker and Columbia counties to 26C at Key West (Henry et al., 1994). The interseasonal incongruit y between strong temperature gradients in January and their de facto absence during July can be explained in part by the interplay of angle and duration of incoming solar radiation During summer, a lower angle of incidence over north Florida is compensate d for by slightly longer duration of daylight compared to the south meaning that the total amount of solar energy received is roughly equal throughout the state In winter, these effects are additive instead of compensatory, such that the north experience s both shorter days and a more acute angle of incidence than the south (Winsberg 1990). N orth Florida is strongly affected by incursions of cold continental air masses during winter that commonly depress temperatures for several days before rebounding to milder conditions mediated by the maritime influence of the Gulf of Mexico and Atlantic Ocean. Effects of cold fronts decrease with decreasing latitude (Henry 1998). The boundary between

PAGE 97

97 continental and peninsular (maritime) winter climate was defined by Winsberg (1990) as a line extending from St. Augustine (St. Johns County) west to the mouth of the Suwannee River in Dixie County. Because frequency, timing, and strength of cold fronts vary there is considerable day to day and interannual variability in winter temperatures. Thus, in contrast to summer, winter temperatures are likely to deviate substantially from the long term average. S pring and autumn fall between summer and winter in terms of temperature variability (Winsberg 1990). The beginning and duration of seasons from a meteorological standpoint varies by latitude and distance from the coast. Winter, defined by Winsberg (1990) as the first week after July during which average maximum air temperatures fall below 24C, commences by the first week of November in the northern panhandle, whereas it is delayed until after the first week of January in extreme southern Florida (Figure 5-3A). Mild to moderate overnight frost associated with cold fronts is a common occurrence in n orth Flo rida, but it rarely persists during daytime. Spring, defined as the first week of the year when average maximum temperatures exceed 24C, begins by the first week of February south of Lake Okeechobee, but not until after early April in coastal areas of the western panhandle ( Winberg, 1990; Figure 5-3B). Summer commences when mean maximum temperatures first exceed 31C (Figure 5-3C). This occurs in May for almost the entire peninsula, with the exception of coastal areas of the Atlantic and western panhandle, where it does not begin until early June. A wedgeshaped area in the southwestern peninsula between Sumter and Hardee counties exceeds 31C prior to the first week of May (Winsberg 1990). Fall begins when average minimum temperatures drop below 16C for the first time after July (Figure 5-3D). This occurs by mid October in the panhandle and not until midDecember along the southeast coast. By this definition, there is no distinct autumn in the Keys, as regional minimum temperatures never fall below 16C (Winsberg, 1990) As with most climatic

PAGE 98

98 variables there is considerable interannual variability, and departures from the stated dates by several weeks are common for any given year. The mean annual temperature range between warmest and coldest month exceed s 30C in north -central Florida, whereas it is less than 14C in the Keys (Winsberg, 1990) This reflects the fact that the seasons are clearly distinguished by pronounced temperature differences in the north, but seasonality progressively decreases southw ard to where diurnal fluctuations frequently exceed interseasonal variability. Precipitation Patterns M any landmasses of similar latitude experience arid to desertlike conditions (Henry et al ., 1994), whereas Florida, due to its peninsular shape and position between the Gulf of Mexico and Atlantic Ocean that supply the state with maritime tropical air (Chen and Gerber 1990), is humid. Mean annual rainfall, based on the period 1961-1990, varies from 102 cm in Key West (Monroe County) to 175 cm in the northwestern panhandle. The southeast coast is a second area of high precipitation, reaching 160 cm near Hialeah ( Miami Dade County ) The statewide average is 137 cm (Henry, 1998). While winter snowfalls occur sporadically in the north they become exceedingly rare with decreasing latitude and contribute only trace amounts in most years. Annual precipitation is dominated by locally generated convective summer thunderstorms (Henry 1998). Onset and termination of the convective season are controlled by position a nd strength of the Bermuda -Azores high pressure system. During winter, it induces subsidence of dry air over Florida, which stabilizes the atmosphere and impedes convection. In summer, subsidence is reduced, generating conditions conducive to thunderstorm formation (Winsberg 1990). Thunderstorms are typically brief, intense and exhibit a distinct afternoon peak in activity that coincides with maximum sea breeze intensity and thermal heating of the land surface

PAGE 99

99 Winter precipitation is lower in intensity a nd about equally likely to occur at any time of the day (Henry 1998). Due to the flat topography of Florida, orographic effects are not profound, although topography has been identified by Chen and Gerber (1990) as contributing to a mosaic of microclimates (p. 12). With the exception of the Keys, which comprise a distinct precipitation regime, seasonality of rainfall increases with decreasing latitude, resulting in very distinct wet and dry seasons toward the south. Annual totals are highly variable, res ulting in frequent droughts of moderate to severe intensity, as well as exceedingly wet years with localized flooding (Henry 1998). Henrys (1998) analysis of longterm data from some stations suggests a weak 5 7 year cycle roughly correlating to the inte rval between successive ENSO events. Tropical cyclonic activity is suppressed during El Nio conditions, an effect that is, however, more than compensated for by greatly increased winter rainfall of up to 300% above longterm means. A La Nia, to the contr ary, is associated with particularly severe drought conditions but increased tropical cyclone activity Atypical jet stream positions and high pressure ridges originating from anticyclonic systems over the central United States and the Bermuda Azores syst em also contribute to rainfall variability. These anomalies rarely affect the entire state equally, such that drought and flooding can occur simultaneously in different areas (Obeysekera et al ., 1999). According to Winsberg (1990), the southeast coast from Key Wes t to Vero Beach, as well as the G ulf coast between Tampa Bay and Pensacola, are subject to the greatest interannual variability. Precipitation R egimes Three distinct rainfall regimes are outlined by Henry (1998), differentiated primarily by the na ture and timing of secondary (i.e. nonconvective) precipitation

PAGE 100

100 Panhandle and n orthern peninsula The northern zone, which encompasses the panhandle and peninsula north of Tampa Bay, is heavily influenced by frontal systems during winter. Squall lines associated with continental cold fronts tend to progress on an easterly to southeasterly trajectory before stalling and eventually dissipating, resulting in a progressive reduction in the amplitude of the winter precipitation peak with decreasing latitude an d, to a lesser degree, decreasing longitude. Low pressure systems moving northeast from the Gulf of Mexico contribute additional moisture to the area. The region between Panama City and Pensacola is second only to the southeast coast in being affected by t ropical storms and hurricanes, which contribute on average, about one third of fall precipitation. Seasonal differences in precipitation are low, but increase with decreasing latitude. Near the southern margin of the northern rainfall regime, dry and wet seasons are clearly distinguishable. The share of s ummer precipitation range s from less than 50% of annual totals in the Pensacola area to greater than 60% near Tampa Bay. Rainfall minima occur in April and October, marking the transition between convection and frontaldominated precipitation patterns Maxima are reached in July for the panhandle, August in the central peninsula and Gulf coast, and September along the Atlantic c oast (Henry 1998). The most frequent excursions from the mean are observed during autumn due to irregular incidence of tropical cyclones (Winsberg 1990). Southern peninsula The southern half of the peninsula experiences a distinct five month wet season lasting from late spring to early fall that is dominated by convective storms enhanced near the southeast coast by strong sea breezes carrying warm, moist air from the Gulf Stream (Henry 1998; Obeysekera et al ., 1999). Midsummer precipitation is sligh tly depressed along the Atlantic coast

PAGE 101

101 due to development of a highpressure ridge associated with the Bermuda Azores system, increasing subsidence and thereby temporarily stabilizing the atmosphere in the region (Henry 1998). Seasonality is pronounced, with dry season rainfall between November and April accounting for only 25% of annual totals, which can be attributed to cold fronts that infrequently progress sufficiently far south to affect the area (Obeysekera et al ., 1999). Important additional sources of precipitation are tropical cyclones that predominantly affect the southern and southeastern coast between Florida Bay and Melbourne (Brevard County) during late summer and early f all (Winsberg 1990). Long term averages suggest that tropical storms reach the area once annually, whereas hurricanes make landfall once to twice per decade (Obeysekera et al ., 1999), although there is significant variability: c onsecutive hurricanes Betsy (1965) and Andrew (1992) were separated by 27 years, whereas Charley, Frances Ivan and Jeanne occurred within a period of six weeks in 2004. Whereas wet season r ainfall maxima vary regionally and typically occur in June, August or September d ry season precipitation tends to be fairly equitably distributed and lacks a distinct secondary peak (Henry 1998). Deviations from the long term average are most frequent during winter (Winsberg 1990). The Keys The Keys form a third, distinct rainfall regime. The relatively dry conditions in this area can be explained by the small size of the islands, which is insufficient to produce significant convection. Thus, the summer precipitation peak is greatly depressed, and mean monthly totals are more equitably distrib uted, such that there is no clear distinction between wet and dry seasons (Winsberg 1990). In summary, the climate of Florida is characterized as highly variable both spatially and temporally. In the north, seasons are defined by clear temperature differences and equitable

PAGE 102

102 distribution of precipitation, whereas the south experiences distinct wet and dry seasons and a much smalle r annual temperature range. Interpolated Baseline Models Derived From Observed Climate Normals B aseline model s representing current climatic conditions patterns in Florida were created by interpolating 19712000 climate normals published by the NCDC for 103 meteorological stations in Florida, southern Georgia and southern Alabama ( Figure 5-4, Table A -1). Mean, minimum, and maximum temperatures as well as precipitation were evaluat ed for January and July, the months during which the climatically most extreme conditions occur (Henry et al ., 1994). Interpolated prediction surfaces were assessed based on quantitative prediction error statistics and qualitative comparison with literature descriptions and formed the foundation for production of geospatially differentiated maps of temperature and precipitation anomalies to identify most vulnerable areas. T o visualize the predominant north south climatic gradient and track changes to its shape and slope over time temperature and precipitation pro files were produced for a 65 7.5 km transect departing from the weather station at Jasper (Hamilton County) near the Georgia border at a heading of 157.6 degrees until reaching the station at Tavernier on Key Largo (Monroe County; Figure 5-5). This transect spans the Florida peninsula along its entire length, encompassing all climate and life zones, and approaching or intersecting multiple areas of ecological and socioeconomic interest These include the increasingly urbanized I nterstate -4 (I 4 ) corridor, th e Lake Wales r idge, the Everglades and Florida Bay. The transect parallels I -75 until crossing it between Lake Panasoffkee and Bushnell in Sumter County (216 km from the origin ) and proceeds along the western edge of the Lake Wales ridge. It bypasses Lake Okeechobee approximately 35 km of its western shore and crosses I -75 again in the northwestern corner of Collier County (km 517) before entering Everglades National Park at For tymile Bend

PAGE 103

103 o n the Tamiami Trail (km 566). It enters Florida Bay at km 633. Meteorological stations located within 10 km of its path include High Springs ( Alachua County, km 83.1), Bushnell 2E ( Sumter County, km 221.9), Winter Haven (Polk County, km 293.6), Bartow (Polk County, km 309.6), Avon Park 2W ( Highlands County, km 352.0), Archbold Biological Station (Highlands County, km 401.6), Devils Garden (Hendry County, km 469.7) and Tamiami Fortymile Bend ( Miami D ade County, km 568.5). B aseline Models for O bserved January Temperatures January mean temperature baseline model After applying the methodology described in the corresponding section of chapter 4, the most suitable interpolation model for observed January mean temperatures during 1971-2000 was iden tified by prediction error statistics to be an anisotropic ordinary Gaussian krige detrended by a second order global polynomial, with 12 lags measuring 15 ,000 m each to measure r esidual variation. A one -sector search neighborhood with 12 members, including at least 6 if a full contingent of sample points was unavailable within the search neighborhood for any location was selected As with all models produced for this study, software calculated values were retained for nugget, partial sill, and direction and shape of anisotropic ellipses. The resulting temperature surface (Figure 5-6) shows a steep temperature gradient across the state with isothermal lines running ne arly parallel to each other, densely stacked in a north eastern to south western progression. D istance s between isotherms are relatively constant with the exception of the area between the 16C and 17C contour lines, which is approximately twice as wide as the distance between any other pair of 1C increment isotherms. The northeastern slant of isotherms reflect s the influence of the G ulf S tream in warming the Atlantic coast relative to locations of equal latitude along the G ulf coast T emperatures ranged fr om 9.21C in extreme northwestern Escambia County near the Alabama border to 20.61 C in the

PAGE 104

104 southern Keys of Monroe County ; a total temperature gradient of 11.40C The statewide mean January temperature was 14.46 C with a SD of 3.04C The Jasper Tavernier transect t emperature profile (Figure 57) illustrate s the nearly linear nature of the gradient except for a flattening of the slope to nearzero around the 16C isotherm. This change in slope correspond s to the larger distance observed between the 16C and 17C isotherms on the temperature surface map Along the transect, the mean temperature gradient is 1.41C/100 km. The quality of the observed January mean temperature interpolation is high. Henry et al (1994) state d a similar tempe rature gradient of approximately 11.1C based on a range of 50F to 70F (10.0C to 21.1C), and t he Gulf Stream induced sloping of isothermal lines match es the description in the literature. T he interpolation model produces considerably smoother isotherms that do not display the coastal interior gradients in south Florida that Henry et al (1994) describe d. Another area of difference i s the western panhandle, where the arrangement of isotherms runs parallel to those in the northern peninsula whereas in the literature description, isotherms follow a more complex pattern that is longitudinal in the extreme west, and curving around locations in northern and southern Walton County ( Choctawhatchee Bay ), respectively. Despite these differences in isotherm orientation, t emperature values for the western panhandle region are nevertheless very similar to the literature E valuated quantitatively prediction e rrors were very slightly elevated where weather station density was lower namely a long the Atl antic coastline the Big Cypress Swamp and southern Everglades area s grading towards Florida Bay and the Keys in Lee, Monroe and southern Dade counties, the Big Bend area, the central panhandle, and a belt spanning north c entral Florida between Levy and St Johns counties (Figure 5 -8). Prediction standard errors ranged from 0.55 to 0.64, with higher values occurring in areas most distant from weather

PAGE 105

105 s tations, or areas where the density of weather stations is lowest. Assuming normality, a 95% confidence int erval for the true value of any point on the surface can be derived by taking the predicted value two times the prediction standard error (ESRI, 2003). The cross validation prediction regression model shows a near perfect 1:1 relationship to measured val ues (Figure 5 9 ). Cross validation further produced excellent prediction error statistics indicating that predicted values are, on average, unbiased, they are as close as possible to measured values within the constraints of the model type and parameters chosen, and the assessment of uncertainty according to prediction standard errors is vali d. M aximum prediction errors ranged from -1.45C (Live Oak, Suwannee County) to +1.39C (Flamingo Ranger Station, Monroe County). Visual examination of the spatial distribution of stations with high prediction errors revealed no patterns or systematic bia ses. The QQ plot indicates that standardized errors are distributed near -normally for this model (Figure 5-10). The downscaling model was successfully validated using a ratio of 75% training data points and 25% test data points. Mean observed (15.49C) and mean predicted (15.48C) temperatures over the 26 point test subset differed by a mere 0.01C with a SD about the mean error of 0.57C. January minimum temperature baseline model To interpolate observed January minima during the baseline period, an anisot ropic ordinary Gaussian krig ing model detrended by a firstorder global polynomial was selected w ith nine lags of 18,000 m each to measure residual variation. A one -sector search neighborhood with 16 members including at least 7 if a full contingent of sample points was unavailable within the search neighborhood for any location, was chosen. The resulting temperature surface (Figure 5-11) shows a very steep north -south temperature gradient across the state superimposed upon whic h are coastal interior gradients

PAGE 106

106 reflecting the maritime influence of the Gulf of Mexico and Atlantic Ocean that results in warmer conditions along the coastline relative to the interior of the state. As a result, isotherm arrangement changes compared to the interpolation of January means, being more densely stacked due to a steeper gradient and more concave due to the greater influence of the oceans on near shore areas. These coastal interior gradients are wellresolved south of the I -4 corridor, at the Ap alachicola delta and in coastal Escambia and Okaloosa counties. T emperatures ranged from 2.35C in northwestern Okaloosa County near the Alabama border to 17.40 C on Key West (Monroe County) for a total temperature gradient of 1 5.05C The average January minimum temperature was 8.08C with a SD of 3. 34C The Jasper T avernier transect temperature profile ( Figure 5-12) illustrate s that the gradient is steep and linear along most of its path However, between central Polk County (km 3 11) and southern Highlands County (km 3 87), where the transect runs a short distance west of the Lake Wales Ridge, the gradient reverses reaching a local minimum near Lake Placid. Henry et al (1994) also identify the Lake Wales Ridge area as cooler compared to its immedi ate surroundings. Along the transect, the mean temperature gradient is 1.8 1C/100 km. T he quality of the interpolation of observed January minima is good. It reproduces most features of the reference map produced by Henry et al (1994), including the north ward sloping isotherms along the coastlines of south Florida, the regional minimum near Lake Placid (although offset by several km to the south) and the regional maximum at the mouth of the Apalachicola river The model does not resemble Henry et al .s (1 994) reference in the arrangement of isotherms over Florida Bay and the southern Keys and does not reproduce a regional minimum at Fountain (Bay County) where Henry place d the statewide minimum. Henry and colleagues gave a gradient of approximately 1 5.5 C based on a range of 37F to 65F

PAGE 107

107 (2.8C to 18.3C) ; indistinguishable from the results generated by the interpolation model when taking unit conversion and precision limitations into account. E valuated quantitatively, prediction standard error values (Figure 5-13) were nearly twice as high as for January means and ranged from 0.89 to 1.23. The cross validation prediction regression model approximates a 1:1 relationship to measured values (Figure 5-14). Cross v alidation produced good prediction error sta tistics confirming minimal bias and validity of prediction standard errors Maximum prediction errors exceeded those for January means and ranged from -2.99C (Key West, Monroe County) to + 2.45C (Flamingo Ranger Station, Monroe County) The QQ -plot ( Figure 5-15) indicates that standardized errors are distributed near normally for this model except at the very tail ends of the distribution corresponding to the Clewiston U .S. Engineers (Hendry County ) and Flamingo Ranger Station ( Monroe County ) locati ons. There was no indication of these stations having faulty data, and they were therefore retained in the analysis. The downscaling model was successfully validated using a ratio of 75% training data points and 25% test data points. Mean observed (9.17C ) and mean predicted (9.23C) temperatures over the 26-point test subset differed by a mere 0.06C with a SD about the mean error of 0.97C. J anuary maximum temperature baseline model O bserved January maximum temperatures were most successfully modeled using an anisotropic ordinary tetraspherical krige detrended by a global second -order polynomial, and with nine lags of 18,000 m each. A onesector search neighborhood with a fixed number of 5 members was selected. The resulting temperature surface (Figure 5-16) shows a temperature gradient that is less steep than for January minima and means. I sothermal lines run nearly parallel to each other, stacked in a northeastern to southwestern progression. The distance between isotherms increases

PAGE 108

108 toward the south. The northeastern slant of isotherms observed for January minima and means is subdued, reflecting that winter daytime temperatures over land frequently exceed ocean temperatures Instead, isotherms curve northward in the interior of the state and southward al ong the G ulf coast Temperatures range d from 15.16 C in extreme northwestern Escambia County near the Alabama border to 2 4.93 C on Big Pine Key (Monroe County); a total gradient of 9.77 C. The average statewide January maximum was 20.80C with a SD of 2.81 C The Jasper Tavernier transect temperature profile (Figure 5-17) confirms that the gradient is steeper in the north than it is in the south, where i t plateaus and ultimately reverses slightly after km 618 until its terminus on Key Largo Its mean slope is 0.95 C /100 km. The quality of the interpolation of observed January maxima is high. The arrangement of isotherms is extremely close to the description by Henry et al (1994), with the exception of the central panhandle, where the model projects isotherm s parallel to those in the peninsula while the literature described a more complex pattern. Henry et al (1994) cite d a temperature gradient of approximately 8.9 C based on a range of 61F to 7 7F (1 6.1 C to 2 5.0C ), p lacing the maximum temperature over the Everglades in western Dade County, approximately 78 km north of Big Pine Key. The interpolation model successfully reproduces decreasing temperatures in the Keys west of Big Pine. P rediction e rrors followed a similar pattern as for January mean and minimum temperatures (Figure 5 -18) The distortion pattern resembling narrow ellipses with their major axes oriented in a north northeast to south southwest orientation is a reflection of the selected interpolation model, w hose search neighborhood was considerably more anisotropic than other models Prediction standard error s ranged from 0.63 to 0.68, similar to the January mean temperature model The cross validation prediction regression model show ed a near perfect 1:1

PAGE 109

109 re lationship to measured values ( Figure 5-19). Cross validation further produced excellent prediction error statistics. Maximum prediction errors ranged from -1.75C (Live Oak, Suwannee County) to +1.96C (Miami Beach, Miami Dade County) intermediate betwee n the mean and minimum models Visual examination of the spatial distribution of stations with high prediction errors revealed no patterns or systematic biases The QQplot indicates that standardized errors are distributed near normally for this model ( Figure 5-20). The downscaling model was successfully validated using a ratio of 75% training data points and 25% test data points. Mean observed (21.75C) and mean predicted (21.71C) temperatures over the 26point test subset differed by a mere 0.04C wit h a SD about the mean error of 0.59C. Baseline Models for Observed July Temperatures July mean temperature baseline model The most suitable interpolation model for observed July mean temperatures during the calibration period was identified as an anisotropic ordinary s pherical krige detrended by a second order global polynomial, with 12 lags of 15,000 m each to account for residual variation. A one sector search neighborhood with 6 members including at least 4 if a full contingent of sample points was una vailable within the search neighborhood for any location, was selected The resulting interpolated temperature surface shows that the state was nearly isothermal during July (Figure 5-21). T emperatures rang ed from 27.00C in extreme northern Escambia County near the Alabama border to 29.13C in the southern Keys of Monroe County, or a total gradient of 2.13C. Henry et al (1994) g ave a nearly identical temperature gradient of approximately 2 .2C based on a range of 80F to 84F (26.7 C to 28.9C). The mean July temperature was 27.59C with a SD of 0.34C Matching the description of Henry et al (1994), there was little north south variation and little difference between coastal and interior temperatures. This is reflected in the temperature profile al ong the Jasper Tavernier transect

PAGE 110

110 ( Figure 5-22), which shows no trend north of the Big Cypress Swamp area (km 507) before increasing rapidly and then more gradually by about 1C until the transect terminates at Key Largo. The quality of the interpolation of observed July temperatures is good, with low prediction standard error values rang ing from 0.40 to 0.44 (Figure 5-23). However, the final baseline model, as well as all alternative interpolation models produced for July, exhibited conditional bias intro duced by interpolation smoothing, whereby high values are underpredicted and low values overpredicted (Figure 5-24). While this is a feature common to kriging models (ESRI, 2003), the extent observed in this case is likely due to geospatial temperature variability during summer being primarily driven by local conditions rather than meso or macroscale temperature gradient s. Unable to resolve these local drivers without multivariate modeling, t he model smoothes out loc al spikes and valleys in the temperature surface. Cross -validation produced excellent prediction error statistics. Maximum prediction errors ranged from -1.43C (Hialeah, Miami Dade County) to +1.15C (Flamingo Ranger Station, Monroe County) Visual exami nation of the spatial distribution of stations with high prediction errors revealed no patterns indicating that the conditional bias and smoothing effect does not constitute a source of systematic biases in the prediction surface The QQ plot indicates th at standardized errors are distributed normally for this model ( Figure 5-25). The downscaling model was successfully validated using a ratio of 75% training data points and 25% test data points. M ean observed (27.65C) and mean predicted (27.75C) temperat ures over the 26 point test subset differed by 0.10C with a SD about the mean error of 0.33C. July minimum temperature baseline model An anisotropic ordinary Gaussian krige detrended by a second-order global polynomial, with nine lags of 18,000 m was sel ected as the downscaling mode for average July minima A

PAGE 111

111 one sector search neighborhood with 16 members including at least 7 if a full contingent of sample points was unavailable within the search neighborhood for any location, was selected. The resultin g interpolated temperature surface shows that while most areas of the state were near isothermal, some minor gradients can be discerned that are primarily coastal interior in nature ( Figure 5-26). Only in southeast Florida does a roughly northwest to southeast gradient become established. This is reflected in the temperature profile along the Jasper -Tavernier transect (Figure 5-27), which shows no clear trend north of a regional temperature minimum of 21.1C in the Lake Placid area (km 396) in Highlands County. South of this location, temperatures increase approximately linearly with distance until the transect terminates at Key Largo reaching 25.26C. Statewide, temperatures ranged from 2 1.03C a t the triple junction between the Georgia, Madison County and Hamilton County borders to 26.36C on uninhabited Marquesas Key 30 km west of Key West (Monroe County) or a total gradient of 5.33C. Henry et al (1994) ga ve a de facto identical temperature gradient of approximately 5.5C based on a range of 69F to 79F (2 0.6 C to 26.1C). The average July minimum was 2 2.15C with a SD of 0.75C The quality of the interpolation of observed July minima is good. P rediction standard errors ranged from 0.62 to 0.89, somewhat higher than the mean based model (Figure 5 -28). Similar to the interpolation model for July mean temperatures, conditional bias exists, albeit to a lesser degree (Figure 5-29). Cross -validation produced excellent prediction error statistics. Maximum prediction errors ranged from -1.76C (Pensacola FAA Airport, Escambia County) to +2.31C (Archbold Biological Station, Highlands County), somewhat higher than the mean temperature model The QQ plot indicates that standardized errors are distrib uted near normally ( Figure 5-30). The downscaling model was successfully validated using a ratio of 75% training

PAGE 112

112 data points and 25% test data points. The mean observed (22.28C) and mean predicted (22.36C) temperatures over the 26point test subset diffe red by 0.08C with a SD about the mean error of 0.57C. July maximum temperature baseline model To obtain a representative temperature surface for 1971-2000 July maxima, an anisotropic ordinary hole effect krige detrended by a firstorder global polynomial with eight lags of 20,000 m was used A one -sector search neighborhood with 4 members and a minimum of 2, was selected. The resulting temperature surface (Figure 5-31) shows that observed July maxima were approximately the same throughout Florida with m inor gradients between coastal and interior areas. Temperatures ranged from 31.85C in Key West to 33.64 C in northwestern Glades County, which corresponds to total temperature differential of merely 1.79C. The average January maximum was 32.99C with a SD of 0.29C The Jasper Tavernier profile ( Figure 5-32) undulates indeterminately between 33 and 33.6C from Jasper to northwestern Glades County (km 420) To the south, temperatures decrease slightly and near -linearly to 32.2C at Key Largo. The quality o f the interpolation of observed July maxima is good. Geospatial distribution of July maxima parallels Henry et al .s (1994) description Henry et al (1994) ga ve a slightly higher temperature differential of approximately 2.7 C based on a range of 87F to 92F ( 30.6 C to 33.3C). One notable flaw in the interpolation model is its in ability to resolve local temperature gradient s in the Miami area. D uring the 1971-2000 period, the station at Miami Beach recorded average July maxima of 30.6C, the coolest in t he state. Since the interpolation model is not an exact interpolator like an IDW model, it is influenced by the nearby stations at Hialeah (33.4C) and Miami WSCMO Airport (32.7C ) that elevate projected temperatures in Miami Beach above their true value It must be emphasized, however, that all interpolation

PAGE 113

113 models in my study are designed to capture regional trends and gradients, and cannot be expected to resolve what approaches the maximum temperature differential in the state occurring within a single m etropolitan area over a distance of less than 20 km. P r ediction standard errors ranged from 0.6 2 to 0.70 (Figure 5 -33). C ross validation while producing excellent prediction error statistics, produced a regression function that departed significantly from a 1:1 relationship to measured values signifying conditional bias ( Figure 5 34 ). Maximum prediction errors ranged from -1.33C (Hialeah Miami -Dade County ) to +2.25C (Miami Beach Miami Dade County ) The QQ plot indicates that standardized errors are di stributed near normally but depart from normal ity at the ends of the distribution, which correspond to the stations at Hialeah and Miami Beach for reasons described in the previous paragraph. (Figure 5-35). The downscaling model was successfully validated using a ratio of 75% training data points and 25% test data points. Mean observed (32.98C) and mean predicted (33.06C) temperatures over the 26-point test subset differed by 0.08C with a SD about the mean error of 0.61C. CCS M Bias Assessment for Temper atures January. The CCSM model generally exhibited moderate to strong warm bias during January. Comparing observed climate data to point samples collected from the least RMSE CCSM derived temperature surface the only station with a negative (cool) bias within Florida was Live Oak ( Suwannee County, -0.61C). Perry (Taylor County, +0.48C), Madison 4 N (Madison County, +0.56C) and Lake City 2 E (Columbia County, +0.59C) were the locations where the least difference between observed and CCSM modeled clima te normals occurred. The stations with the most positive ( warm ) bias were Archbold Biological Station (Highlands County, +3.44C ), Ver o Beach 4 W ( Indian River County, +3.40C ), Fort Pierce ( St. Lucie

PAGE 114

114 County, +3.31C), and Immokalee 3 NNW (Collier County, +3.20C). The mean bias for all 83 Florida based weather stations was +2.01C with a SD of 0.82C Based on analysis of smoothed interpolations of observational and CCSM data, a process which deemphasize s local anomalies and interpolation artifacts in order to more clearly resolve larger scale patterns ( Figure 5-36) model predictions were on average ( i.e. statewide) 2.01C warmer than observational data indicates ( SD 0.70C ). D iscrepancies range d from +0. 40C (northern Hamilton County) to + 3.70C at the mouth of the Apalachicola River Warm bias was most pronounced in the coastal regions of the state, particularly in the panhandle and the Space and Treasure C oast areas ( Broward and southeastern Palm Beach Counties), where it exceeded +3C Recalling that the CCSM model, while one of the higher resolution GCM s in use, is limited in its ability to resolve Floridas shoreline and that many CCSM grid cells upon which these interpolations are based are in real ity on terrestrial/oceanic hybrids, it is not surprising that CCSM overstates the influence of relatively warmer sea surface temperatures compared to the cooler land areas during January Without bias correction, CCSM would therefore project consistently w armer conditions than observed, particularly near the coastline. Bias correction addresses these deviations. July. CCSM generally exhibited a cool bias during July due to the same limitations of the GCM that cause d warm bias during winter although at a smaller magnitude Comparing observed climate data via point samples from the interpolated CCSM surface, the only stations with a positive (warm) bias within Florida were Panama City 5 NE ( Bay County, +0.18C) and Archbold Biological Station ( Highlands County, +0.14C). Archbold, Vero Beach 4 W ( Indian River County, -0.02C), Monticello 3 W ( Jefferson County, -0.13C), DeFuniak Springs ( Walton

PAGE 115

115 County, -0.13C) and Parrish (Manatee County, 0.13C) were the locations where the least difference between observ ed and modeled climate normals occurred. The stations with the most negative (cool) bias were Hialeah ( Miami -Dade County, -2.27C ), Key West WSO Airport ( Monroe County, -1.65C ), St. Petersburg (Pinellas County, -1.51C ), Live Oak ( Suwanee County, -1.46C ) and Fernandina Beach ( Nassau County, -1.35C ). The mean bias for all 83 Florida based weather stations was -0.68C with a SD of 0.82 C Analysis of smoothed interpolations of observational and CCSM data (Figure 5-37), reveals that model predictions were on average 0.62 C cooler than observational data indicates ( SD 0.25C ) The discrepancies range d from -1.59C to + 0.37C Warm bias was restricted to the extreme northwestern panhandle, encompassing the northern half of Escambia County and the northwester n corner of Santa Rosa County. C ool bias was greatest in northeast Florida (Nassau, northern Duval, northern Baker and northern Columbia counties), as well as in south Florida (southern and eastern Monroe, Dade, e astern Broward and southeastern Palm Beach counties), where it exceeded -1C. Baseline Models for Observed January Precipitation The best interpolation model for observed January precipitation was an anisotropic ordinary Gaussian krige detrended by a first -order global polynomial, with 18 lags measuring 10,000 m. A one-sector search neighborhood with 16 members including at least 5 if less than 16 sample points fell within the search neighborhood for any location, was selected. The resulting interpolated precipitation surface ( Figure 5-38) shows that latitude was the predominant control on rainfall during January while distance from the coast bec ame increasingly important in south central and southern Florida. The Atlantic coast was generally wetter than the G ulf c oast at the same lati tude. An area of precipitation below 60 mm arche d from the southern tip of the peninsula through its interior towards Cape Canaveral. Rainfall

PAGE 116

116 r anged from 44 mm along the central G ulf coast of mainland Monroe County to 157 mm in northwestern Escambia County. Mean January rainfall was 92 mm with a SD of 34 mm The precipitation profile along the Jasper Tavernier transect (Figure 5-39) supports a general north s outh trend that bottoms out after a distance of 600 km (west-central Dade County), before increasing slightly as the transect approaches the coast of Florida Bay. H enry et al (1994) present ed precipitation totals in inches, making identification of local s cale patterns more difficult due to the considerably less precise unit of measurement Nevertheless there is general agreement between literature and the interpolation model. V alues are, however, quite different : Henry et al (1994) ga ve a mean precipitation value of 2.7 in ches (69 mm) but t he interpolation models value of 92 mm is well -supported by 1971-2000 climate normals: The average value of the 83 weather stations located in Florida was 85 mm ( SD 31 mm) Since the distribution of Florida weather stations was determined to be statistically random ( p=0.24) by the Average Nearest Neighb or function in ArcGIS ( Figure 4-2) their mean value is generally representative of the mean value for the state as a whole. The cause for the difference in values between Henry et al (1994) and the 19712000 climate normals is unknown, but may include random variability affecting baseline values or impacts of climatic change that, while not evident in temperature projections, may be reflected in changing precipitation during the late 20th century T he quality of the interpolation of observed January pre cipitation is good. Per the prediction standard error map e rror values ranged from 5.23 to 8.83 (Figure 5 -40). The cross validation prediction regression model showed a near -perfect 1:1 relationship to measured values. Cross validation produced excellent prediction error statistics ( Figure 5-41). Maximum prediction errors ranged from 19 mm (West Palm Beach WSO Airport) to + 21 mm (Pensacola

PAGE 117

117 FAA Airport) The QQ plot indicates normality of standard errors for most of its range, with two stations at the high end (Pensacola FAA Airport, Tampa WSCMO Airport) departing from norma l ity ( Figure 5-42). The downscaling model was successfully validated using a ratio of 75% training data points and 25% test data points. The m ean observed (80.68 mm) and mean predicted ( 80.76 mm) precipitation totals over the 26 point test subset differed by 0.08 mm with a SD about the mean error of 7.03 mm. Baseline Models for Observed July Precipitation The most suitable interpolation model for observed July precipitation was an ordinar y isotropic Gaussian krige, with nine lags measuring 20,000 m. This model was not detrended. A one -sector search neighborhood with 19 members including at least 4 if less than 19 members were available within the search neighborhood around any location, was selected. T he resulting interpolated precipitation surface ( Figure 5-43) shows, with the exception of extreme southeastern Florida, that there was a pronounced east west gradient in July rainfall, undoubtedly due to the influence of the BermudaAzore s high pressure system. The Atlantic coast i s, therefore, much dryer than the G ulf c oast at the same latitude. Rainfall ranged from 97 mm at Marathon (Monroe County) to 227 mm in coastal Dixie County Mean July rainfall was 181 mm with a SD of 20 mm. T here is general agreement between the interpolation model and Henry et al (1994) with the exception of the western panhandle, where Henrys isohyets run in a north south direction and the interpolation models isohyets run from west to east V alues are close as well : Henry et al (1994) ga ve a mean precipitation value of 7.5 in ( 191 mm). The precipitation profile along the Jasper Tavernier transect (Figure 5-44) roughly parallels the 175 mm contour and crosses it in multiple locations. After a distance of 540 k m from the origin, over eastern Collier County in the Big Cypress National Preserve the transect enters an area

PAGE 118

118 marked by a regional northsouth declining gradient, likely along the transition between southern peninsula and Florida Keys precipitation regimes The quality of the interpolation of observed J ul y precipitation is fair as illustrated in the prediction standard error map in Figure 5-45, which exhibits patterns similar to those for the temperatures interpolations. Error values ranged from 18.69 to 25.81. The cross validation prediction regression model exhibited some conditional bias ( Figure 5-46) Maximum prediction errors ranged from -51 mm (Myakka River State Park, Sarasota County) to + 63 mm (Miami Beach, Miami -Dade County). Both standard and m aximum prediction errors exceed those for January by a large margin. The QQ -plot confirms normal ity of standard errors (Figure 5-47). The downscaling model was successfully validated using a ratio of 75% training data points and 25% test data points. The m ean observed (183.66 mm) and mean predicted (179.54 mm) precipitation totals over the 26 -point test subset differed by 4.12 mm with a SD about the mean error of 19.00 mm. CCSM Bias Assessment for Precipitation January. CCSM exhibited a moderate to strong dry bias in northwest Florida, and a neutral to wet bias in southeast Florida during January. Comparing observed climate data to the least RMSE CCSM derived interpolated precipitation surface, 30 stations exhibited wet bias and 53 a dry bias. The stations with the most negative (dry) bias were Milton Experiment Station (Santa Rosa County, -95 mm), Chipley 3 E (Washington County, 87 mm), Niceville ( Okaloosa County, -82 mm), and Panama City 5 NE (Bay County, 82 mm), all located in the panhandle The stations with the most positive (wet) bias were Miami WSCMO Airport ( Miami Dade County+30 mm), Flamingo Ranger St ation (Monroe County, +26 mm), Tamiami Trail Fortymile Bend ( Miami -Dade County, +24 mm), and Tavernier ( Monroe County, + 19 mm), all located in south Florida and the Keys. The mean bias for all 83 Florida based weather stations was 22 mm

PAGE 119

119 with a SD of 33 mm Comparison of smoothed interpolation surfaces of observational and CCSM data revealed that model predictions were on average 28 mm drier than observational data indicate ( SD 35 mm). Discrepancies ranged from 89 mm to + 26 mm. ( Figure 5-48). July. In July, the CCSM model exhibited a moderate to strong dry bias along the G ulf coast that was most pronounced in the panhandle, and a neutral to slightly wet bias in southeast and northeast Florida. Comparing observed climate data to the least RMSE CCSM derived interpolated precipitation surface, 16 stations exhibited wet bias and 67 a dry bias. The stations with the most negative (dry) bias were N iceville (Okaloosa County, -112 mm), Wewahitchka (Gulf County, -100 mm), Myakka River State Park ( Sarasota County, -96 mm), and Panama City 5 NE ( Bay County, -92 mm), all located in the coastal areas of the Panhandle and in Sarasota County (Myakka River) The stations with the most positive (wet) bias were Miami Beach ( Miami -Dade County, +52 mm), Tavernier (Monroe County, +51 mm), Key West WSO Airport ( Monroe County, +36 mm), and Jasper ( Hamilton County, +36 mm) the only station on this list not located in south Florida and the Keys. The mean bias for all 83 Floridabased weather stations was -29 mm ( SD 33 mm). Based on geospatial analysis of smoothed interpolation surfaces of observational and CCSM data model predictions were on average (statewide) 32 m m drier than observational data indicates ( SD 23 mm). Discrepancies ranged from -98 mm (Escambia County barrier islands) to + 32 mm (middle Keys, Monroe County) ( Figure 5-49). Projections of Climate Change in Florida Modified ThorntonWilhelmi Method to Visualize CCSM Projected Temperature Anomalies As discussed in chapter 4, evaluating raw CCSM outputs within the original model grid is an expedient and convenient way to visualize GCMprojected anomalies and their statistical significance before any inter polation or bias correction operations are applied to the data. The

PAGE 120

120 purpose of this evaluation is identif ication of large scale patterns and how they affect or drive changes occurring within the boundaries of the state. January T he 1990-1999 baseline map ( Figure 5-50A reproduced in Figures 4-51A and 3-52A ) skillfully reproduces many macroscale features of observed climate illustrated in the 1971-2000 baseline model of NCAR climate normals ( Figure 5-6) and discussed in the literature review. These include a general northwest to southeast climatic gradient and Atlantic grid cells being warmer than grid cells of equal latitude in the Gulf of Mexico. CCSM successfully resolves expected patterns of grid cells over land being associated with cooler temperatures during winter than ocean based polygons due to the higher heat capacity of water that warms overlying air via heat transfer processes. This explains both the positive temperature bias during January and its pattern of being at its greatest magnitude in coa stal areas ( Figure 5-36) Identification of smaller s cale patterns is prevented by Floridas geography relative to the 1.4 resolution of the grid cells, meaning that most of the polygons covering the political boundaries of the state are, in reality, hybr id land/ocean cells. S cenario B1. Anomaly maps 3 -50B through 3-50E, corresponding to 10year averaged temperatures over the intervals 2020 -2029, 2040-2049, 2060-2069 and 2090-2099, project greater warming over land than over the ocean. This is consistent w ith theory that suggests that heat absorbed by oceans is partially conveyed via thermohaline processes to deep water below the mixed surface layer that interacts with the atmosphere (IPCC, 2007a ). In terms of statistical significance examined by one tailed Students t tests with pooled variances, it is noteworthy that the level of significance of ocean based polygons tends to exceed that of land-based polygons. For example, Figure 550B shows that anomalies for 2020 -2029 are significant at the p =0.01 level for all ocean and several land/ocean hybrid cells, whereas only one fully land based cell

PAGE 121

121 shows significance at the same level. Over Florida, the level of significance appears to be a factor of the relative percentage of land to water each cell corresponds to. Based on the three almost entirely land -based polygons in the Big Bend area and in the central and southern peninsula, the level of significance is marginal, at about the p=0.1 level. With each progressive 10year interval evaluated, the m agnitude of the anomalies and their statistical significance increases, albeit slowly, to the point where all 72 grid cells exhibit significance at p=0.01 in 20902099. The magnitude of anomalies is generally greater in the northwest than in the east and s outh, suggesting a lessening of the climate gradient over the area. Scenario A1B. The intermediate scenario of climate change, A1B essentially shows the same patterns unfolding as under scenario B1, but at a more rapid rate and reaching a greater magnitud e by the end of the century ( Figure 551). The difference between the scenarios appears to increase with time; in fact, Figure 5-51B suggests that during the 2020-2029 period, scenario B1 may produce change of slightly greater magnitude and higher signific ance in south central Florida and along the Atlantic seaboard of Georgia and South Carolina. By 2040-2049, expected temperature warming exceeds projections for scenario B1, and statistical significance for the three nearly entirely land -based polygons in F lorida is at p=0.05 or below. By 2060-2069, all 72 grid cells show change significant at the p=0.01 level compared to the baseline. S cenario A2 Under this scenario, statistical significance at the p=0.01 level is achieved for all 72 polygons by 2020-2029, and is maintained through the rest of the century. Continentality appears to be the main driver of anomaly magnitudes, establishing a northwestto southeast pattern. Over parts of Florida, the rate of warming slows or reverses between 2060 2069 a nd 2090-2099, as reflected in the Big Bend and southcentral peninsula grid cells changing colors, indicating lower anomaly values. On the contrary, the western panhandle, the

PAGE 122

122 Gulf coast north of Tampa Bay, and southeast Florida are associated with higher anomalie s than during the preceding projection. Averaged over all 72 CCSM polygons, 2.25C of warming is expected to occur by 2060-2069 relative to the calibration period, whereas only an additional 0.42C is projected through the end of the century. July The July temperature baseline map for 1990-1999 temperatures as projected by the CCSM 20th Century Experiment ( Figure 553A) reflects that the breakdown of macroscale temperature gradients during summer is not limited to Florida alone, but extends over most of the 72polygon area with the exception of its northwesternmost extent corresponding to central Mississippi. Few patterns relevant for Florida can be discerned, except that the interior of the state, as reflected by the three previouslymentioned grid cells th at are primarily land based, are associated with the coolest temperatures in the region. Scenario B1. Unlike projections for January, which established anomaly magnitude s as a function of continentality for the southeastern United States, the opposite is the case for July : Anomalies tend to be greater and more statistically significant over Florida and surrounding ocean areas than in the states to the north and northwest ( Figure 5-53) With the exception of the extreme northwestern panhandle, which falls under grid cells located primarily in Alabama, statistical significance at p=0.01 is achieved by 20202029 for Florida and is maintained through the entire century. Portion s of Mississippi and Alabama where the greatest anomalies in the 72 polygon grid occur during winter, fail to achieve even the p=0.1 level under this scenario for July. Anomaly magnitudes for Florida are uniform throughout the second half of the century, and exceed those of surrounding land and ocean areas, especially in the 2060-2069 projection (Figure 5-53D), which is perhaps the best example of CCSMs ability to recognize Florida as a terrestrial feature under the right conditions.

PAGE 123

123 Scenario A1B. Scenario A1B is essentially a more rapidly unfolding version of scenario B1 on this spatial scale. Statistical significance at p=0.01 is reached by 2060-2069, when greater anomalies in Florida compared to the surrounding areas become apparent ( Figure 5-54). S cena rio A2 Scenario A2 resembles the projection for scenario A1B for the 2020-2029 period closely ( Figure 5-55B) By 20402049, all grid cell anomalies have achieved statistical significance at the p=0.01 level compared to the baseline ( Figure 5-55C). The pat tern of anomalies is somewhat irregular. While Florida, as reflected in the three reference polygons previously discussed, exhibits more positive anomalies than surrounding areas, several land based polygons to the north as well as ocean -based polygons in the Gulf and Atlantic show below a verage cooling. Throughout the rest of the century, north and central Florida remain among the regions with the highest anomalies, rivaled only by central Mississippi ( Figure 555D E ). Summer anomalies are greater than projected winter anomalies for this region, an anomaly of its own on the global scale, where warming during winters is generally projected to be greater than during summers (IPCC 2007a). R ecalling the brief discussion on dangerous climate change and eco logical and socioeconomic impacts in chapter 1, the magnitude of projected change in the region is a cause for concern under any of the three scenarios and in either season. This emphasiz es the need to examine climatic trends at finer spatial resolutions t han raw GCM data can provide. While CCSM is one of the first models that recognizes Florida as a land feature (most climate models have too coarse a spatial resolution to resolve the peninsula and consider it part of the ocean), projections at the state le vel and finer spatial resolutions, essential to gauge impacts on ecologically and policy relevant scales, require additional processing by geostatistical downscaling or dynamic regional climate modeling, the latter of which requires substantial

PAGE 124

124 computing r esources and highly specialized software programming, often rivaling GCMs in complexity To summarize, the Thornton Wilhelmi visualization of CCSM data establishes that substantial warming is expected for the area Scenario B1 projects greater warming during Januar y, whereas A1B and A2 project more warming during July. It has further been established that, in general, warming is expected to be greater over land than water During January, north Florida and particularly the panhandle are subject to greater warming than the south of the peninsula whereas the central to south central peninsula are projected to experience the greatest anomalies du ring July. Projections from D ownscaled, B ias -C orrected I nterpolations January means Scenario B1. Under the most conservative scenario for 21st century climate change, B1, projected temperature change for Florida is moderate. The mean statewide temperature is projected to change from the 1971-2000 baseline of 14.46 C ( SD 3.04C Figure 5-6) to 15.35C ( SD 2.92C ) in the 2020-2029 timeframe ( Figure 5-56), 15.36C by 2040-2049 ( SD 2.98 C ; Figure 5-57), 15.57C ( SD 3.01 C ) by 2060-2069 ( Figure 5-58) and 15.78 C ( SD 2.86C ) by 2090-2099 ( Figure 5-59). This equates to a change of 0.89C, 0.9C, 1.11C and 1.32 C compared t o the baseline, respectively. Anomaly maps suggest that the pattern of change will be predominantly forced by longitudinal location through midcentury ( Figure 5-60), with greater warming towards the west, whereas the trend changes to a lessening of the latitudinal temperature gradient during the second half of the century ( Figure 5-61), as predicted by the ThorntonWilhelmi analysis (Figure 5-50). N orthern portions of the state are projected to warm by up to 1.71C (in northwestern Escambia County) by 20 90-2099 compared to 1.01C in the south (the lower Keys of Monroe County). Further, the western half of the peninsula is projected

PAGE 125

125 to warm to a very slightly greater extent than the Atlantic coast and interior of the state at equal latitudes. As a result, the gradient between coolest and warmest location in the state is expected to decrease from 11.41C during the calibration period to 11.14C by 2040-2049 and 10.70C by 2090-2099. The Jasper Tavernier temperature profile changes little in shape, though its gradient is reduced from 1.41 C /100 km during the baseline period (Figure 5-7) to 1.39 C /100 km in 2040-2049 ( Figure 5-62) and 1.33 C /100 km in 2090-2099 ( Figure 5-63, Table 5-1). Scenario A1B. The intermediate climate change scenario, A1B projects large temperature change s for Florida. The mean statewide temperature is projected to change from the 1971 -2000 baseline of 14.46C ( SD 3.04C Figure 5-6) to 15.30C ( SD 2.94C ) in 2020-2029 timeframe ( Figure 5-64), 15.70C by 2040-2049 ( SD 2.96C Figure 5-65), 16.19C ( SD 2.87C ) by 20602069 ( Figure 5-66), and 16.47C ( SD 2.90C ) by 2090-2099 (Figure 5-67) This equates to a change of 0.84C, 1.24C, 1.73C and 2.01C compared to the baseline, respectively. Anomaly maps suggest that the pattern of chang e will be equally forced by l atitudinal and longitudinal location through midcentury ( Figure 568), with greater warming in the northwest than southeast, whereas longitudinal location predominates forcing during the second half of the century (Figure 5-69). The statewide temperature gradient is projected to lessen, as northwestern portions of the state are projected to warm by up to 2.29C (in northwestern Escambia County) by 2090-2099 compared to 1.76C in the south and southeast (the lower Keys of Monroe County). Further, the western half of the peninsula is projected to warm to a greater extent than the Atlantic coast and interior of the state at equal latitudes. As a result, the gradient between coolest and warmest location in the state is expected to de crease from 11.41C during the calibration period to 11.04C by 2040-2049 and 10.89C by 2090-2099. The Jasper Tavernier temperature profile changes little in shape, through its gradient is reduced from 1.41C/100 km during the baseline

PAGE 126

126 pe riod (Figure 5-7) to 1.38C/100 km in 2040-2049 ( Figure 5-70) and 1.35C/100 km in 2090 2099 ( Figure 5-71, Table 5-1). S cenario A2 The most aggressive scenario, A2, projects highly significant temperature change for all areas of Florida. The mean statewide temperature is projected to change from the 1971-2000 baseline of 14.46C ( SD 3.04C Figure 5-6) to 15.90C ( SD 2.90C ) in the 2020 2029 t imeframe (Figure 5-72), 15.99C by 2040-2049 ( SD 2.86 C Figure 5-73), 17.06C ( SD 2.75C ) by 2060-2069 (Figure 5 74) and 17.10C ( SD 2.82 C ) by 2090-2099 ( Figure 5-75). This equates to a change of 1.44C, 1.53C, 2.60C and 2.64C compared to the baseline, respectively. As with the other scenarios, the statewide temperature gradient is projected to lessen, although patterns of change are somewhat more complex under scenario A2. By midcentury, the pattern of change in the peninsula is similar to mid century projections under scen ario A1B, albeit of much greater magnitude. In the panhandle, to the contrary, patterns of change approximate end of -century projections under scenario B1, and despite the much shorter time horizon, are of slightly greater magnitude. Therefore, control seems to be due to a combination of latitudinal and longitudinal forcing in the peninsula, whereas control is strictly latitudinal in the panhandle (Figure 5-76) By the end of the century a unique pattern becomes established, where control is entirely longitudinal with the exception of the Atlantic coast, where it is mostly latitudinal. The 2.5C and 2.6C anomaly contours bend at a nearly right angle at a distance of approximately 65 km from the Atlantic shoreline. The 2.4C contour bends at a 45 angl e at Lake Okechobee, but assumes a northerly heading once again to reach the coast at an acute angle. The western panhandle is projected to warm by up to 3.34C (in western Escambia County) by 2090-2099 compared to 2.35C at Key Largo, but values are highl y similar all along the southeastern coast up to Palm Beach County (Figure 5-77). T he gradient between

PAGE 127

127 coolest and warmest location in the state is expected to decrease from 11.41C during the calibration period to 10.74C by 2040-2049 and 10.67C by 2090-2099. The Jasper Tavernier temperature profile changes little in shape, though its gradient is reduced from 1.41C/100 km during the baseline period (Figure 5-7) to 1.32C/100 km in 2040-2049 (Figure 5-78) and 1.36C/100 km in 2090-2099 (Figure 5-79, Table 5-1). Therefore, scenario A2 implies the most rapid temperature change, and the greatest decline in the temperature difference between northern and southern parts of the state by midcentury although the gradient stabilizes and reverts somewhat during the second half of the century January m inima and maxima Since minima and maxima were derived by performing simple arithmetic as part of the bias correction and calibration steps geospatial patterns and magnitude of change are identical to the above projections with a slight margin of error introduced by use of different interpolation models. Only absolute values differ. M inima. Average minima under scenario B1 are projected to increase from the 1971 -2000 average of 8.08 C ( SD 3.34C Figure 5-11) to 8.97C ( SD 3.29C ) by 2040-2049 (Figure 5-80) and 9.40 C ( SD 3.16C ) by 2090-2099 (Figure 5-81). Under scenario A1B, the change is expected to be 9.31 C by 2040-2049 ( SD 3.26C Figure 5-82) and 10.08 C ( SD 3.20 C ) by 2090-2099 ( Fi gure 5-83). Scenario A2 projects 9.60 C ( SD 3.16 C ) in 2040-2049 (Figure 5-84) and 10.72 C ( SD 3.12C ) by 2090-2099 (Figure 5-85). T he gradient between coolest and warmest location in the state is expected to decrease from 15.05 C during the calibration period to 14.79C by 2040-2049 under scenario B1, 14.72 C under scenario A1B and 14.37C under scenario A2. By 2090-2099, these values are reduced to 14.39 C 14.54C and 14.34C respectively As with the mean -based models, there is little change in the shape of the Jasper T avernier temperature profiles, whereas there are changes in slope, which changes from

PAGE 128

128 1.8 1C /10 0 km during the baseline period (Figure 5-12) to 1.78C /100 km (Scenario B1, Figure 5-86), 1.78C /100 km (Scenario A1B Figure 5-87) and 1.72/100 km (Scenario A2, Figure 5 88 ), respectively by midcentury. By 2090-2099, mean slopes are projected at 1.73 C /100 km ( Scenario Figure 5-89), 1.75C /100 km ( Scenario A1B Figure 5-90) and 1.75C /100 km ( Scenario Figure 5-91; Table 5 -2). M axima. Average maxima under scenario B1 are projected to increase from the 1971 2000 a verage of 20.80C ( SD 2.81 C Figure 5-16) to 21.70C ( SD 2.75 C ) in 2040-2049 ( Figure 5-92) and 22.12C ( SD 2.63C ) by 2090 2099 (Figure 5-93). Under scenario A1B, the change is expected to be 22.03C by 2040 2049 ( SD 2.72C Figure 5-94) and 22.80C ( SD 2.67C ) by 2090-2099 (Figure 5-95). Scenario A2 projects 22.33C ( SD 2.63C ) in 2040 2049 ( F igure 5-96) and 23.44C ( SD 2.57C ) by 2090-2099 (Figure 5-97) The gradient between coolest and warmest location in the state is expected to decrease from 9.77 C during the calibration period to 9.50C by 2040-2049 under scenario B1, 9.39C under scenario A1B and 9.02C under scenario A2. By 2090-2099, these values are reduced to 9.07C 9.26C and 8.71C The Jasper Tavernier profile slope is expected to change from 0.95 C /100 km ( Figure 5 17) to 0.92C /100 km (B1, Figure 5-98), 0.89C /100 km (A1B, Figure 5-99), and 0.87C /100 km (A2 Figure 5-100) by 2040-2049. By 2090-2099, these values will be 0.87C /100 km ( Scenario B1 Figure 5-101), 0.87C /100 km ( Scenario A1B Figure 5-102), and 0.93C /100 km ( Scenario A2 Fig ure 5-103, Table 5 -3). Ju ly m eans Scenario B1. Under scenario B1, projected temperature change for Florida in July is moderate and similar in magnitude to January The mean statewide temperature is projected to change from the 1971 -2000 baseline of 27.59 C ( SD 0.34C Figure 5-21) to 28.36C ( SD 0.30C ) in the 2020-2029 timeframe (Figure 5-104), 28.64C by 2040-2049 ( SD 0.30C Figure

PAGE 129

129 5 -105), 28.75C ( SD 0.32C ) by 2060-2069 ( Figure 5-106) and 28.81 C ( SD 0.32 C ) by 2090 2099 ( Figure 5 107). This equates to a change of 0.77C, 1.05C, 1.16C and 1.22 C compared to the baseline, respectively. There is little regional differentiation: By 2040 -2049, the difference between greatest and least change is merely 0.21C, with the greatest change occurring in the northern peninsula, the Apalachicola delta and Tampa Bay and the least change at the south ern tip of the peninsula and the Keys (Figure 5-108) The same pattern holds for the remainder of the century, with the difference between most and least warming declining to 0.17C ( Figure 5-109) As such, gradients and temperature patterns remain largely unchanged including the shape and slope of the Jasper Tavernier profile (1971-2000: 0.20C/100 km ( Figure 5-22), 2040-2049: 0.16C/100 km (Figure 5-110), 2090-2099: 0.18C/100 km ( Figure 5 111, Table 5 -4)). S cenario A1B. The intermediate climate change s cenario, A1B projects significant climate change for Florida during July that exceeds projected January warming by 0.25C by centurys end. The mean statewide temperature is projected to change from the 1971 -2000 baseline of 27.59C ( SD 0.34C Figure 5-21) to 28.52C ( SD 0.30C ) in the 2020 -2029 timeframe (Figure 5-112), 28.94C by 2040-2049 (SD= 0.38C Figure 5-113), 29.46C (SD= 0.27C ) by 2060-2069 (Figure 5-114) and 29.85C (SD= 0.30 C ) by 2090-2099 ( Figure 5 115). T his equates to a change of 0.93 C, 1.35C, 1.87C and 2.26C compared to the baseline, respectively. Regional differentiation is larger compared to scenario B1: By midcentury, the difference between greatest and least change is 0.38C, with the greatest change occurring in the Tampa Bay are a and the least change in the northwestern panhandle (Figure 5 116) By 20902099, the difference between greatest and least warming declines to 0.30C, with the greatest change occurring along the coastal areas of the panhandle between the

PAGE 130

130 A palachicola delta and the Alabama border, and the least change in southeastern Dade County ( Figure 5-117). The slope and shape of the Jasper Tavernier profile does not change appreciably (1971-2000: 0.20C/100 km (Figure 5-22), 2040-2049: 0.20C/100 km (Figure 5-118), 2090 2099: 0.17C/100 km (Figure 5-119, Table 5 -4)). S cenario A2. Scenario A2 projects highly significant temperature change exceeding January projections by an average margin of 1.09C when comparing 2090-2099 summer and winter anomalies Regional differe ntiation is appreciably larger than for the other scenarios in July and it is the only July scenario to project an increase in regional differentiation of anomalies over time that, incidentally, cause Florida to become more isothermal The mean statewide temperature is projected to change from the 1971-2000 baseline of 27.59C (SD= 0.34C Figure 5-21) to 28.37C (SD= 0.32C ) during 2020-2029 ( Figure 5-120), 28.99 C by 2040-2049 (SD= 0.29 C Figure 5-121), 29.69C (SD= 0.26C ) by 2060-2069 (Figure 5-122) and 30.97 C (SD= 0.21C ) by 2090-2099 (Figure 5-123) This equates to a change of 0.78 C, 1.40C, 2.10C and 3.38 C relative to the baseline, respectively. By midcentury, the difference between greatest and least change is 0.30C, with the greatest change occur ring in the western panhandle, and the least change in southeastern Dade County (Figure 5-124). By 2090-2099, the difference between greatest and least change increases to 0.86C, with the greatest change occurring in the western panhandle to the Alabama b order, and the least change the coastal areas between Key Largo and northern Palm Beach County in southeast Florida (Figure 5-125). This causes the difference between warmest and coolest areas of the state to declin e from 2.13C to 1.65C. This is reflected in a lessening of the average slops of the JasperTavernier temperature profile from 0.2C/100 km during the baseline (Figure 5-22) to 0.17C/100 km in 2040-2049 ( Figure 5-126) and 0.09C/100 km in 2090-2099 (Figure 5-127, Table 5 -4)

PAGE 131

131 Ju ly m inima and maxima M inima. Average July minima under scenario B1 are projected to increase from the 1971 2000 a verage of 22.14C (SD= 0.75 C Figure 5-26) to 23.19C (SD= 0.71 C ) by 2040 2049 ( Figure 5-128) and 23.36C (SD= 0.73 C ) by 2090-2099 ( Figure 5-129). Under scenario A1B, the change is expected to be 23.48C (SD= 0.78C ) by 2040-2049 ( Figure 5-130) and 24.40C (SD= 0.71C ) by 2090-2099 (Figure 5-131). Scenario A2 projects 23.54C (SD= 0.71C ) by 2040 2049 (Figure 5-132) and 25.51C (SD= 0.63 C ) by 2090-2099 ( Figure 5 133) The gradient between coolest and warmest location in the state is expected to change from 5.33 C during the calibration period to 5.21 C by 20402049 under scenario B1, 5.58 C under scenario A1B and 5.2 C under scenario A2. By 2090-2099, these values change to 5.33 C 5.27C and 5.19 C The Jasper -Tavernier profile slope ( Figures 3-134 through 3-139) is expected to remain within 0.04C /100 km of the baseline slope of 0.60C /100 km (Figure 5-27) regardless of scenario or time horizon ( Table 5 -5). M axima. Average maxima under scenario B1 are projected to increase from the 1971 2000 a verage of 32.99 C (SD= 0.29 C Figure 5-16) to 34.04 C (SD= 0.31 C ) by 20402049 ( Figure 5-140) and 34.21C (SD= 0.30C ) by 2090-2099 (Figure 5-141) Under scenario A1B, the change is expected to be 34.33 C (SD= 0.31 C ) by 2040-2049 ( Figure 5-142) and 35.25C (SD= 0.30C ) by 2090-2099 (3 -143). Scenario A2 projects 34.39C (SD= 0.30C ) by 2040-2049 ( Figure 5-144) and 36.36 C (SD= 0.37 C ) by 2090-2099 ( Figure 5-145) The gradient between coolest and warmest location in the state is expected to increase slightly from 1.79C during the calibration period to 1.98 C by 20402049 under scenario B1, 1.83 C under scenario A1B and 1.88 C under scenario A2. By 2090 -2099, these values change relatively little to 1.85C 1.89C and 2.32 C respectively The Jasper Tavernier profile slope is negative but near 0 for all scenarios ( Figures 5-146 through 5-151). Slopes remains stable within 0.04C/100 km

PAGE 132

132 o f the baseline value of -0.17C/100 km with the exception of scenario A2 during 2090-2099 ( 0.28 C/100 km, Figure 5-151, Table 5 -6). P recipitation Important disclaimers about precipitation projections are that they tend to be more uncertain than temperature due to the ir discontinuous, highly localized and temporally variable nature, as was discussed in the literature review of chapter 4. They are considered here in isolation of temperature projections. More concretely, this means that a projection of increased preci pitation for some area of the state at a future time may correspond to a decrease in water availability for domestic, industrial and ecohydrological purposes if increases in evapotranspiration exceed increased precipitation, even assuming constant withdraw als for human activities. Integration of ecohydrological modeling is planned for future versions of my downscaling models project pending stakeholder interest and funding availability. Modified ThorntonWilhelmi Method to Visualize CCSM Projected Precipita tion Anomalies Unlike temperature projections, which initial data exploration revealed to follow a clear trajectory towards warming, precipitation anomalies follow a less linear pattern. For this reason, pooled variance ttests for statistical significance of anomalies relative to the baseline were calculated with two tails, rather than just one as was done for temperatures. In the maps, therefore, the color of a grid cells shows the sign and magnitude of the anomaly, whereas the overlain patterns signify whether or not the anomaly is significantly different (either higher or lower) from baseline conditions. When observed in combination as presented in the ThorntonWilhelmi maps discussed below, it is readily apparent in which manner (high or low) an anomaly is significantly different from the baseline.

PAGE 133

133 January The 19901999 baseline map ( Figure 5-152A, reproduced in Figures 4-153A and 4-154A) bears little resemblance to observed climate as illustrated in the 1971 -2000 baseline model of NCAR climate normals ( Figure 5-38) and discussed in the literature review with the exception of wetter conditions in the extreme northwestern panhandle. The CCSM baseline projects the area of lowest precipit ation over Tampa Bay, which observed data do not support. Southeastern Florida receiving more precipitation than the rest of the state is only indirectly supported by Gulf Stream influence, which CCSM appears to overestimate. Clearly CCSM performs poorly at replicating precipitation patterns over Florida, making bias correction essential to producing credible precipitation projections. Even though the uncorrected baseline model is faulty, analysis of anomalies is nevertheless of interest, since they are un affected by bias correction procedures. Scenario B1. Anomaly maps 3 -152B through 3 -152EE, corresponding to 10year averaged precipitation over the intervals 2020 -29, 2040-49, 2060-69 and 2090-99, show no significant changes or patterns relevant to Florida. Magnitudes of anomalies over Florida are approximately neutral, tending towards slight increases in precipitation in the south and slight decreases in the north, with some slight variation between 10 year intervals. The sole exception occurs in the far western panhandle and the southeastern tip of the state (roughly corresponding to Collier, Monroe Lee and Hendry counties), which show increased precipitation at a marginal significance level of p=0.1 in 2060-2069 (Figure 5-152D). During no other time do a nomalies approach statistical significance in grid cells over or near Florida. In fact, anomalies for none of the 72 polygons during 2020-2029 and 2090-2099 show significance levels better than p=0.1. During 2040-2049, portions of Alabama and Mississippi a re projected to experience reductions in precipitation in excess of 10 mm, and grid cells in the area indicate statistical significance at p=0.05 to p=0.1 (Figure 5152C). These cells revert to near -baseline conditions by 2060-2069,

PAGE 134

134 whereas increased preci pitation is expected over the Atlantic Ocean and portions of the Gulf of Mexico southwest of Florida and south of Louisiana, Mississippi and Alabama at various levels of significance up to p=0.01 ( Figure 5-152D). Scenario A1B. The intermediate scenario of climate change, A1B, essentially shows a similar patterns unfolding as under scenario B1, but with even less regional differentiation ( Figure 5-153B-E ). On average, the scenario tends to be slightly drier than B1, including Florida during 2060-2069 ( Fig ure 5-153D) and 2090-2099 ( Figure 5153E). Still, anomalies are close to 0, and no cells above or near Florida approach statistical significance at any time. S cenario A2 Scenario A2 produces the most complex pattern of change for January precipitation amo nalies. By 2020-2029 ( Figure 5-154B), a trend of drying over much of the panhandle, neutral to slightly drier conditions over the northern peninsula, and neutral to slightly wetter condition in the south can be discerned, albeit lacking significance. Proje ctions for 2040 2049 ( F igure 5-154C) and 2060-2069 ( Figure 5-154D) are highly similar, and show neutral to slightly drier conditions in the central and southern peninsula, whereas the Keys, panhandle and northern peninsula are subject to drying at signific ance levels between p=0.1 and p=0.05. By 2090-2099 ( Figure 5-154 E), the trend reverts to a pattern similar to Figure 5-154B, but marginal statistical significance at the p=0.1 level is retained for portions of the panhandle. J uly A s was the case with for J anuary, t he 1990-1999 CCSM baseline map ( Figure 5 15 5A, reproduced in Figures 4-156A and 3-157A) bears little resemblance to observed climate as illustrated in the 1971 -2000 baseline model of NCAR climate normals ( Figure 5-43) and discussed in the literature review S cenario B1. CCSM projects Florida to dry progressively during the 21st century, with the magnitude of anomalies increasing toward the south, whereas conditions in the panhandle remain

PAGE 135

135 relatively stable. During 2020-2029 (Figure 5-155B), neutral to slightly wetter conditions predominate in the panhandle, whereas neutral to slightly dryer conditions are encountered in the peninsula, excluding a marginally significant hybrid landocean cell in east Florida that projects greater precipitatio n losses. By 2040-2049, statistical significance at the p=0.05 level or better is established for most of the Florida peninsula, the Atlantic, and large portions of Mississippi and Alabama. Whereas the more continental areas of the 72 polygon area are projected to receive in excess of 40 mm more rainfall, the southern and central portions of the Florida peninsula will be subject to declining precipitation. North Florida, including the panhandle, remains approximately neutral, with statistical significance v arying from grid cell to grid cell. For the remainder for the century, the trend of highly negative (and highly significant) anomalies in the central and southern peninsula intensifies, whereas conditions in the north remain neutral to slightly wetter. By 2090-2099, all grid cells encompassing the peninsula show significance levels below p=0.01, whereas changes in the western panhandle have p-values above 0.1. Nearly the entire Gulf of Mexico, with the exception of areas close to the coast of Florida, does not show significant changes. S cenario A1B. As was the case for January precipitation, scenario A1B (Figure 5-156B E ) shows a similar pattern unfolding as under scenario B1, but at a faster rate and cumulating in greater (dryer) anomalies over the entire Florida peninsula, whereas the panhandle, remains in near neutral territory, lacking statistical significance. During 2020-2029 ( Figure 5-156B) and 2040-2049 ( Figure 5-156C) anomalies over Florida are neutral to slightly drier in the central and southern peninsula at levels of significance during 2040-2049 ranging from p> 0.1 for the Tampa Bay area and the shore to its north, to p=0.01 in the southeast. For the remainder of the century, anomalies increase and the number of statistically significant cells increases for all areas

PAGE 136

136 of Florida except the panhandle, as well as the Atlantic off Floridas coast. As under scenario B1, the Gulf of Mexico, aside from areas near the coast of Florida, is not projected to experience significant changes in rainfall. Miss issippi, Alabama and northern Georgia grid cells reach high levels of significance in the 2060 -2069 ( Figure 5-156D) and 2090-2099 ( Figure 5-156E) projections with anomalies indicating in excess of 40 mm of additional precipitation relative to current conditions. Scenario A2 Scenario A2 (Figure 5-157) maintains the patters of nearneutral anomalies lacking statistical significance in the panhandle over the entire century, whereas the peninsula, particularly its southern reaches, experience s a very strong dr ying trend that is significant at the p=0.01 level by 2040-2049 (Figure 5-157C). By 2090-2099, southeastern Florida is projected to lose in excess of 80 mm of rainfall during July. The entire peninsula except the grid cell between the Big Bend and Tampa Ba y ( p=0.05) shows anomalies significant at the p=0.01 level. Projections from Downscaled, Bias -C orrected Interpolations January S cenario B1 Under scenario B1, changes in precipitation patterns and magnitude through the year 2100 are minor and statewide averages do not exhibit a clear trend toward either dryer or wetter conditions Unlike temperature projections, precipitation changes are not unidirectional between 10 year intervals. A verage statewide rainfall is projected to change from the 197 1-2000 baseline of 92 mm (SD= 34 mm Figure 5-38) to 90 mm (SD= 33 mm) in the 2020-2029 (Figure 5-158) timeframe, 89 mm (SD= 31 mm) by 2040 2049 (Figure 5-159), 97 mm (SD= 34 mm) by 2060-2069 ( Figure 5-160) and 92 mm (SD= 32 mm) by 2090-2099 (Figure 5-161) This equates to a mean statewide c hange of -2 mm, 3 mm and +5 mm and 0 mm compared to the baseline, respectively. There is little regional differentiation during 2040-2049, maximum anomalies range from -5.43 mm in the upper Keys to + 4.71 mm in coastal Collier County ( Figure 5-162).

PAGE 137

137 B y the end of the century, anomalies remain low, ranging from -7.44 mm in southwestern Escambia County to +5.37 mm along the coast of Indian River County (Figure 5-163). Expressed as percentages relative to the baseline, th e scenario produces values ranging from 90.5% in the upper Keys to 109.5 % in coastal Collier County by 2090-2099. Regionally, these data suggest slightly dryer conditions in the Keys and much of north Florida, particularly the Big Bend area, whereas other regions can expect no changes or slightly wetter conditions. As was the case for temperatures, the Jasper Tavernier precipitation profile changes little in shape over time, and changes to its slope are inconsistent. By 20402049, the mean slope of the grad ient is reduced slightly from -10.6 mm/100 km during the baseline period (Figure 5-39) to 10.3 mm/100 km in 2040-2049 ( Figure 5-164). By 2090-2099, this trends reverts, and the slope is -11.3 mm/100 km (Figure 5-165, Table 5 7). These results match the ThorntonWilhelmi anomaly analysis, which did not find any consistent trends nor statistical significance of grid cells over or near Florida (Figure 5-152). S cenario A1B. Under scenario A1B, changes in precipitation patterns and magnitude are minor to mode rate for most of Florida point ing towards slightly drier conditions A verage statewide rainfall is projected to change from the 1971-2000 baseline of 92 mm (SD= 34 mm, Figure 5-38) to 93 mm (SD= 34 mm) in the 2020-2029 timeframe (Figure 5-166), 87 mm (SD= 32 mm) by 2040-2049 ( Figure 5-167) 87 mm (SD= 36 mm) by 2060-2069 (Figure 5-168) and 85 mm (SD= 33 mm) by 2090-2099 ( Figure 5-169) This equates to a mean statewide change of +1 mm, 5 mm, -5 mm, and 7 mm compared to the baseline, respectively. Indeed, projections for the 2060-2069 and 20902099 timeframes are highly similar in patterns and values. Regional differentiation is similar to scenario B1, but with the mean shifted by several millimeters towards drier conditions. By midcentury, the nort heastern panhandle, along the border to Georgia, is

PAGE 138

138 e xpected to experience the greatest losses in precipitation by up to 9.5 mm less than is presently the case (Figure 5-170), while t he barrier islands off the coast of Indian River and St. Lucie County are projected to lose less than 1 mm of precipitation during January. For the 2090-2099 timeframe, greatest losses are recorded in northeastern Jackson County (and, more generally, the entire northern panhandle and northern peninsula with the exception of Escambia County) and the lower Keys of Monroe County (Figure 5-171) While at opposite ends of the state geographically, they share projected losses in precipitation up to 9.5 mm/month for January. Expressed as percentages relative to the baseline period, the scenario produces values ranging from 82.3% in the lower Keys to 95.4 % in western Escambia County. The mean slope of the Jasper Tavernier precipitation profile remains nearly constant, changing from -10.6 mm/100 km during 1971-2000 ( Figure 5-39) to 10.3 mm/1 00 km by midcentury ( Figure 5-172) and 10.8 mm/100 km during 2090-2099 ( Figure 5-173, Table 5 -7). S cenario A2. S cenario A2 projects a progressive drying trend that reaches its greatest anomaly in 2060-2069, before reverting to less dire conditions in 2090-2099. Scenario A2 also exhibits more regional differentiation than the other scenarios. A verage statewide precipitation is projected to change from the 1971-2000 baseline of 92 mm (SD= 34 mm, Figure 5 38 ) to 89 mm (SD= 32 mm) in the 2020-2029 timeframe ( Figure 5-174), 85 mm (SD= 35 mm) by 2040-2049 ( Figure 5-175), 82 mm (SD= 32 mm) by 2060-2069 ( Figure 5-177) and 89 mm (SD= 28 mm) by 2090-2099 ( Figure 5-178). This equates to a change of -3 mm -7 mm -10 mm and -3 mm compared to the baseline, respectively. By midcentury, drying in excess of 9.5 mm is expected for the lower Keys, with other regions experiencing precipitation losses in excess of 8 mm including Taylor County and several counties in northeast Florida. In northwestern Escambia County, relative dr ying of less than 1 mm is projected (Figure 5-178). By 2090-2099, anomalies

PAGE 139

139 r ange from less than -16 mm in northern Jackson County to more than +7.5 mm at Cape Canaveral (Figure 5-179). Expressed as percentages relative to the baseline period, the scenario produces values ranging from 89.0% in the middle and lower Keys (slightly higher (i.e. less dry) than under scenario A1B) to + 112% at Cape Canaveral. The mean slope of the Jasper Tavernier precipitation profile remains constant between the baseline (Figure 5-39) and 2040-2049 ( 10. 6 mm/100 km Figure 5-172) before decreasing to -9.9 mm/100 km during 2090-2099 ( Figure 5-173, Table 5 -7). J uly S cenario B1. Under scenario B1 changes in precipitation patterns and magnitude tend towards progressively drier conditions in south and east Florida, whereas rainfall in the panhandle remains largely unchanged. A verage statewide rainfall is projected to change from the 1971-2000 baseline of 181 mm (SD= 20 mm, Figure 5-43) to 174 mm (SD= 26 mm) in the 2020 2029 timefr ame (Figure 5-182), 166 mm (SD= 31 mm) by 2040-2049 ( Figure 5-183), 165 mm (SD= 26 mm) by 2060-2069 ( Figure 5-184) and 160 mm (SD= 28 mm) by 2090-2099 (Figure 5 185) This equates to a change of 7 mm, -15 mm, 16 mm and 21 mm compared to the baseline, respectively. There is substantial regional differentiation. Through midcentury, precipitation in the panhandle changes little in pattern and magnitude. In the peninsula particularly south of the I 4 corridor, substantial drying is noted. T he range of anomalies is -38 to +19 mm, with only the central and western panhandle receiving more precipitation than is presently the case. Negative anomalies reach their greatest values along the coast in Palm Beach County (Figure 5-186). The 2040-2049 ( Figure 5-183) and 2060-2069 (Figure 5-184) precipitation surfaces are nearly identical, indicating that the CCSM model projects a stabilization of precipitation changes for Florida during the middle of the 21st century under scenario B1 before re initiating a slight drying trend by 2090-2099. Statewide, anomalies

PAGE 140

140 b etween the calibration period and the end of the century range from -42 mm in coastal Palm Beach County to + 10 mm in the northernmost extent of the panhandle, paralleling the Alabama and Georgia borders (Figure 5-187). Expressed as percentages relative to the baseline period, the scenario produces values ranging from 67.9% in North Key Largo t o 106.6% in the northern panhandle The mean slope of the Jasper Tavernier precipitation profile increases consistently, changing from 9.0 mm/100 km during 1971-2000 ( Figure 5-44) to 12.0 mm/100 km by midcentury ( Figure 5-188) and 13.1 mm/100 km during 2090-2099 ( Figure 5-189, Table 5 8). S cenario A1B. Under scenario A1B, conditions tend towards significant drying in southeast Florida, where some areas are projected to receive less than half the current amount of rainfall, to largely unchanged conditions in the western panhandle. A verage statewide rainfall is projected to change from the 1971 -2000 baseline of 181 mm (SD= 20 mm, Figure 5-43) to 167 mm (SD= 26 mm) in the 2020-2029 timeframe (Figure 5-190), 170 mm (SD= 32 mm) by 2040-2049 ( Figure 5-191), 152 mm (SD= 33 mm) by 2060-2069 (Figure 5-192) and 153 (SD= 35 mm) by 2090-2099 ( Figure 5-193). This equates to a change of 14 mm 11 mm 29 mm and 28 mm compared to the baseline, respectively. As under scenario B1, precipitation in the panhandle is r elatively stable, whereas the western half and southern tip of the peninsula are projected to dry significantly T he range of anomalies between the calibration period and midcentury is from -40 mm in coastal Broward County to + 26 mm in the northern panhandle ( Figur e 5-194). Statewide, anomalies between the calibration period and the end of the century range from -63 mm in coastal Broward County to + 17 mm in north eastern Escambia County ( Figure 5-195). Expressed as percentages relative to the baseline period, the sce nario produces values ranging from 46.2% in the upper Keys to 108.9% in northe a stern Escambia County. The mean slope of the Jasper Tavernier precipitation profile increases consistently, and to a greater

PAGE 141

141 e xtent than is the case under scenario B1. Whereas the slope was 9.0 mm/100 km during the baseline period ( Figure 544), it is projected to reach 14.8 mm/100 km by midcentury ( Figure 5 1 96) and -15.7 mm/100 km during 2090-2099 ( Figure 5-197, Table 5 8). S cenario A2. Under scenario A2, average statewide precipitation is projected to decrease greatly from the 1971-2000 baseline of 181 mm (SD= 20 mm, Figure 5-43) to 170 mm (SD= 78 mm) in 2020-2029 ( Figure 5-198), 157 mm (SD= 31 mm) by 2040-2049 ( Figure 5-199), 156 mm (SD= 31 mm) by 2060-2069 ( Figure 5-200) and 139 mm (SD= 40 mm) by 2090-2099 ( Figure 5 201) This equates to a change of 11 mm 24 mm -25 mm and 42 mm compared to the baseline, respectively. By 2040-49, the western panhandle will be up to 7 mm wetter relative to the baseline, and southeast Florida up to 50 mm drier (Figure 5-202). The 20 40-2049 ( Figure 5 199) and 2060-2069 ( Figure 5-200) precipitation surfaces are nearly identical, indicating that the CCSM model projects a temporary stabiliza tion of precipitation regimes affecting Florida during the middle of the 21st century under scenario A2, as was the case for scenario B1 By 2090-2099, anomalies reach -89 mm along the coast of Broward County to + 7 mm in northwestern Okaloosa County ( Figur e 5 -201). Expressed as percentages relative to the baseline period, the scenario produces v alues ranging from 23.6% in the middle Keys to 103.5% in northern Okaloosa County. The mean slope of the Jasper Tavernier precipitation profile initially increases at a slower rate than scenario A1B, but exceeds it by a margin of 2.2mm/100 km by 2090-2099. Whereas the slope was .0 mm/100 km during the baseline period (Figure 544), it is projected to reach .3 mm/100 km by midcentury (Figure 5-204) and -17.9 mm/1 00 km during 2090 2099 ( F igure 5-205, Table 5 -8).

PAGE 142

142 Having established projected precipitation and temperature changes in Florida for the 21st century, as well as assessments of statistical significance and uncertainty, initial assessments on the implication s of these changes can be discussed, as will be done in the next chapter. .

PAGE 143

143 Figure 51. Political map of the State of Florida.

PAGE 144

144 A B Figure 5-2. Florida climate classifications. A) Kppen system, B) Thornthwaite system. [Adapted from Henry (1998) Weather and climate (pp. 16-17). Water resources atlas of Florida (ed. by E.A. Fernald and E.D. Purdum), pp. 16-37. Florida State University Institute of Science and Public Affairs, Tallahassee, FL. ]

PAGE 145

145 A B C D Figure 5-3. Beginning of meteorological seasons (month/week) in Florida. A) Winter, the first week after July during which average maximum air temperatures fall below 23.9C, B) Spring when average maximum temperatures first exceed 23.9C, C) Summer, when average maximum te mperatures first exceed 31.1C, D) Fall, the first week after July during which average minimum temperatures fall below 15.6C. [Adapted from Winsberg (1990). Florida weather University of Central Florida Press, Orlando, FL. A) p 29, Fig. 2:2. B) p. 54, Fig 3:2. C) p. 77, Fig. 4:1. D) p 108, Fig. 5:2.]

PAGE 146

146 Figure 5-4. Location of reference meteorological stations in Florida, Alabama and Georgia.

PAGE 147

147 Figure 5-5. Location of the Jasper Tavernier transect and distance from its origin at Jasper.

PAGE 148

148 Figure 5-6. Interpolation of observed January mean temperatures 1971-2000.

PAGE 149

149 Figure 5-7. January m ean t emperature profile for 1971-2000 along the Jasper Tavernier transect

PAGE 150

150 Figure 5-8. January m ean t emperature baseline model p rediction standard e rror m ap

PAGE 151

151 Figure 5-9. Error s tatistics and regression function for the January mean temperature baseline model (screenshot) Figure 5-10. QQplot for the January mean temperature baseline model (screenshot)

PAGE 152

152 Figure 5-11. Interpolation of observed January minimum temperatures.

PAGE 153

153 Figure 5-12. January minimum temperature profile for 1971-2000 along the JasperTavernier transect

PAGE 154

154 Figure 5-13. January minimum temperature baseline model prediction standard error map.

PAGE 155

155 Figure 5-14. Error statistics and regression function for the January minimum temperature baseline model (screenshot). Figure 5-15. QQplot for January minimum temperature baseline model (screenshot).

PAGE 156

156 Figure 5-16. Interpolation of observed January maximum temperatures 1971-2000.

PAGE 157

157 Figure 5-17. January maximum temperature profile for 1971-2000 along the JasperTaverni er transect.

PAGE 158

158 Figure 5-18. January maximum temperature baseline model prediction standard error map.

PAGE 159

159 Figure 5-19. Error statistics and regression function for the January maximum temperature baseline model (screenshot). Figure 5-20. QQplot for January maximum temperature baseline model (screenshot).

PAGE 160

160 Figure 5-21. Interpolation of observed July mean temperatures 1971-2000.

PAGE 161

161 Figure 5-22. July m ean temperature profile for 1971-2000 along the JasperTavernier transect

PAGE 162

162 Figure 5-23. July m ean t emperature baseline model p rediction standard e rror m ap

PAGE 163

163 Figure 5-24. Error s tatistics and regression function for the July mean temperature baseline model (screenshot) Figure 5-25. QQ-plot for J uly mean temperature baseline model (screenshot)

PAGE 164

164 Figure 5-26. Interpolation of observed July minimum temperatures 1971-2000.

PAGE 165

165 Figure 5-27. July minimum temperature profile for 1971-2000 along the JasperTavernier tr ansect

PAGE 166

166 Figure 5-28. July minimum temperature baseline model prediction standard error map.

PAGE 167

167 Figure 5-29. Error statistics and regression function for the July minimum temperature baseline model (screenshot). Figure 5-30. QQplot for July minimum temperature baseline model (screenshot).

PAGE 168

168 Figure 5-31. Interpolation of observed July maximum temperatures 1971-2000.

PAGE 169

169 Figure 5-32. July maximum temperature profile for 1971-2000 along the JasperTavernier transect

PAGE 170

170 Figure 5-33. July maximum temperature baseline model prediction standard error map.

PAGE 171

171 Figure 5-34. Error statistics and regression function for the July maxim um temperature baseline model (screenshot). Figure 5-35. QQplot for July maximum temperature baseline model (screenshot).

PAGE 172

172 Figure 5-36. CCSM bias for January temperatures: Point samples at weather station locations and interpolated, smoothed bias surface.

PAGE 173

173 Figure 5-37. CCSM bias for July temperatures: Point samples at weather station locations and interpolated, smoothed bias surface.

PAGE 174

174 Figure 5-38. Interpolation of observed January precipitation 1971-2000.

PAGE 175

175 Figure 5-39. January precipitation profile fo r 1971-2000 along the JasperTavernier transect

PAGE 176

176 Figure 5-40. January precipitation baseline model prediction standard error map.

PAGE 177

177 Figure 5-41. Error statistics and regression function for the January precipitation baseline model (screenshot) Figure 5-42. QQ-plot for the January precipitation baseline model (screenshot)

PAGE 178

178 Figure 5-43. Interpolation of observed July precipitation 1971-2000.

PAGE 179

179 Figure 5-44. July precipitation profile for 1971-2000 along the JasperTavernier transect

PAGE 180

180 Figure 5-45. July precipitation baseline model prediction standard error map.

PAGE 181

181 Figure 5-46. Error statistics and regression function for the July precipitation baseline model (screenshot). Figu re 5 -47. QQ-plot for July precipitation baseline model (screenshot).

PAGE 182

182 Figure 5-48. CCSM bias for January precipitation: Point samples at weather station locations and interpolated, smoothed bias surface.

PAGE 183

183 Figure 5-49. CCSM bias for July precipitation: Point samples at weather station locations and interpolated, smoothed bias surface.

PAGE 184

184 Figure 5-50. CCSM projected January temperature anomalies and their statistical significance under scenario B1 relative to 1990-1999 baseline A) 1990-1999 baseline. B) 20202029 anomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 185

185 Figure 5-51. CCSM projected January temperature anomalies and their statistical significance under scenario A1B r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 20202029 anomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 186

186 Figure 5-52. CCSM projected January temperature anomalies and their statistical significance under scenario A 2 relative to 1990-1999 baseline A) 1990-1999 baseline. B) 20202029 anomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 187

187 Figure 5-53. CCSM projected J uly temperature anomalies and their statistical significance under scenario B1 r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 2020-2029 anomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 188

188 Figure 5-54. CCSM projected J uly temperature anomalies and their statistical significance under scenario A1B r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 2020-2029 anomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 189

189 Figure 5-55. CCSM projected J uly temperature anomalies and their statistical significance under scenario A2 r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 2020-2029 anomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 190

190 Figure 5-56. Projected January mean temperature 2020 -2029, Scenario B1.

PAGE 191

191 Figure 5-57. Projected January mean temperature 20 40-2049, Scenario B1.

PAGE 192

192 Figure 5-58. Projected January mean temperature 20 60-2069, Scenario B1.

PAGE 193

193 Figure 5-59. Projected January mean temperature 20 90-2099, Scenario B1.

PAGE 194

194 Figure 5-60. Anomaly map for January temperatures, comparing 2040-2049 projections under scenario B1 to the 1971-2000 baseline.

PAGE 195

195 Figure 5-61. Anomaly map for January temperatures, comparing 2090-2099 projections under scenario B1 to the 1971-2000 baseline.

PAGE 196

196 Figure 5-62. Projected 2040-2049 January mean temperature profile under scenario B1 along the Jasper Tavernier t ransect. Figure 5-63. Projected 2090-2099 January mean temperature profile under scenario B1 along the Jasper Tavernier transect.

PAGE 197

197 Table 51. Projected January mean temperature gradient and mean slope between transect origin and terminus. Scenario Interv al Gradient Mean Slope Baseline 1971 2000 9.28 C 1.41 C/100 km B1 204 0 20 49 9.11 C 1.39 C/100 km 2090 2099 8.77 C 1.33 C/100 km A1B 2040 2049 9.08 C 1.38 C/100 km 2090 2099 8.90 C 1.35 C/100 km A2 2040 2049 8.71 C 1.32 C/100 km 2090 2099 8.91 C 1.36 C/100 km

PAGE 198

198 Figure 5-64. Projected January mean temperature 2020 -2029, Scenario A1B.

PAGE 199

199 Figure 5-65. Projected January mean temperature 2040 -2049, Scenario A1B.

PAGE 200

200 Figure 5-66. Projected January mean temperature 2060 -2069, Scenario A1B.

PAGE 201

201 Figure 5-67. Projected January mean temperature 2090 -2099, Scenario A1B.

PAGE 202

202 Figure 5-68. Anomaly map for January temperatures, comparing 2040-2049 projections under scenario A1B to the 1971-2000 baseline.

PAGE 203

203 Figure 5-69. Anomaly map for January temperatures, comparing 2090-2099 projections under scenario A1B to the 1971-2000 baseline.

PAGE 204

204 Figure 5-70. Projected 2040-2049 January mean temperature profile under scenario A1B along the Jasper Tavernier transect. Figure 5-71. Projected 2090-2099 January mean temperature profile under scenario A1B along the Jasper Tavernier transect.

PAGE 205

205 Figure 5-72. Projected January mean temperature 2020 -2029, Scenario A2.

PAGE 206

206 Figure 5-73. Projected January mean temperature 2040 -2049, Scenario A2.

PAGE 207

207 Figure 5-74. Projected January mean temperature 2060 -2069, Scenario A2.

PAGE 208

208 Figure 5-75. Projected January mean temperature 2090 -2099, Scenario A2.

PAGE 209

209 Figure 5-76. Anomaly map for January temperatures, comparing 2040-2049 projections under scenario A2 to the 1971-2000 baseline.

PAGE 210

210 Figure 5-77. Anomaly map for January temperatures, comparing 2090-2099 projections under scenario A2 to the 1971-2000 baseline.

PAGE 211

211 Figure 5-78. Projected 20402049 January mean temperature profile under scenario A2 along the Jasper Tavernier transect. Figure 5-79. Projected 20902099 January mean temperature profile under scenario A2 along the Jasper Tavernier transect.

PAGE 212

212 Figure 5-80. Projected January minimum temperature 20 40-2049, Scenario B1.

PAGE 213

213 Figure 5-81. Projected J anuary minimum temperature 20 90-2099, Scenario B1.

PAGE 214

214 Figure 5-82. Projected January minimum temperature 20 40-2049, Scenario A1B.

PAGE 215

215 Figure 5-83. Projected January minimum temperature 20 90-2099, Scenario A1B.

PAGE 216

216 Figure 5-84. Projected January minimum temperature 2040 -2049, Scenario A2.

PAGE 217

217 Figure 5-85. Projected January minimum temperature 2090 -2099, Scenario A2.

PAGE 218

218 Figure 5-86. Projected 2040-2049 January minimum temperature profile under scenario B1 along the Jasp er Tavernier transect. Figure 5-87. Projected 2040-2049 January minimum temperature profile under scenario A1B along the Jasper Tavernier transect.

PAGE 219

219 Figure 5-88. Projected 2040-2049 January minimum temperature profile under scenario A2 along the Jasper Tavernier transect. Figure 5-89. Projected 2090-2099 January minimum temperature profile under scenario B1 along the Jasper Tavernier transect.

PAGE 220

220 Figure 5-90. Projected 2090-2099 January minimum temperature profile under scenario A1B along the Ja sper Tavernier transect. Figure 5-91. Projected 2090-2099 January minimum temperature profile under scenario A2 along the Jasper Tavernier transect.

PAGE 221

221 Table 5 -2. Projected January minimum temperature gradient and mean slope between transect origin and terminus. Scenario Interval Gradient Mean Slope Baseline 1971 2000 11.88C 1.81C /100km B1 204 0 20 49 11. 73 C 1. 78 C/100km 2090 2099 11.3 8 C 1.73C/100km A1B 2040 2049 11.70C 1.78C/100km 2090 2099 11.51C 1.75C/100km A2 2040 2049 11.33C 1.72C/100km 2090 2099 11.52C 1.75C/100km

PAGE 222

222 Figure 5-92. Projected January maximum temperature 2040-2049, Scenario B1.

PAGE 223

223 Figure 5-93. Projected January maximum temperature 2090-2099, Scenario B1.

PAGE 224

224 Figure 5-94. Projected January maximum temperature 2040-2049, Scenario A1B.

PAGE 225

225 Figure 5-95. Projected January maximum temperature 2090-2099, Scenario A1B.

PAGE 226

226 Figure 5-96. Projected January maximum temperature 2040-2049, Scenario A2.

PAGE 227

227 Figure 5-97. Projected January maximum temperature 2090 -2099, Scenario A2.

PAGE 228

228 Figure 5-98. Projected 2040-2049 January maximum temperature profile under scenario B1 along the Jasper Tavernier transect. Figure 5-99. Projected 2040-2049 January maximum temperature profile under scenario A1B along the Jasper Tavernier transect.

PAGE 229

229 Figure 5-100. Projected 2040-2049 January maximum temperature profile under scenario A2 along the Jasper Tavernier transect. Figure 5-101. Projected 2090-2099 January maximum temperature profile under scenario B1 along the Jasper Tavernier transect.

PAGE 230

230 Figure 5-102. Projected 2090-2099 January maximum temperature profile under scenario A1B along the Jasper Tavernier transect. Figure 5-103. Projected 2090-2099 January maximum temperature profile under scenario A2 along the Jasper Tavernier transect.

PAGE 231

231 Table 5 -3. Projected January maximum temperature gradient and mean slope between transect origin and terminus. Scenario Interval Gradient Mean Slope Baseline 1971 2000 6.25C 0.95C/100km B1 204 0 20 49 6.07C 0.92C/100km 2090 2099 5.73C 0.87C/100km A1B 2040 2049 5.87C 0.89C/100km 2090 2099 5.75C 0.87C/100km A2 2040 2049 5.75C 0.87C/100km 2090 2099 6.13C 0.93 C/100km

PAGE 232

232 Figure 5-104. Projected July mean temperature 2020-2029, Scenario B1.

PAGE 233

233 Figure 5-105. Projected July mean temperature 20 40-2049, Scenario B1.

PAGE 234

234 Figure 5-106. Projected July mean temperature 20 60-2069, Scenario B1.

PAGE 235

235 Figure 5-107. Projected July mean temperature 20 90-2099, Scenario B1.

PAGE 236

236 Figure 5-108. Anomaly map for July temperatures, comparing 2040-2049 projections under scenario B1 to the 1971-2000 baseline.

PAGE 237

237 Figure 5-109. Anomaly map for July temperatures, comparing 2090-2099 projections under scenario B1 to the 1971-2000 baseline.

PAGE 238

238 Figure 5-110. Projected 20402049 July mean temperature profile under scenario B1 along the Jasper Tavernier transect. Figure 5-111. Projected 20902099 July mean temperature profile under scenario B1 along the Jasper Tavernier transect.

PAGE 239

239 Table 5 -4. Projected July mean temperature gradient and mean slope between transect origin and terminus. Scenario Interval Gradient Mean Slope Baseline 1971 2000 1.30 C 0.20C/100 km B1 204 0 20 49 1.02 C 0.16C/100 km 2090 2099 1.18 C 0.18C/100 km A1B 2040 2049 1.33 C 0.20C/100 km 2090 2099 1.10 C 0.17C/100 km A2 2040 2049 1.12 C 0.17C/100 km 2090 2099 0.60 C 0.09C/100 km

PAGE 240

240 Figure 5-112. Projected July mean temperature 2020 -2029, Scenario A1B.

PAGE 241

241 Figure 5-113. Projected July mean temperature 2040 -2049, Scenario A1B.

PAGE 242

242 Figure 5-114. Projected July mean temperature 2060 -2069, Scenario A1B.

PAGE 243

243 Figure 5-115. Projected July mean temperature 2090 -2099, Scenario A1B.

PAGE 244

244 Figure 5-116. Anomaly map for July temperatures, comparing 2040-2049 projections under scenario A1B to the 1971-2000 baseline.

PAGE 245

245 Figure 5-117. Anomaly map for July temperatures, comparing 2090-2099 projections under scenario A1B to the 1971-2000 baseline.

PAGE 246

246 Figure 5-118. Projected 2040-2049 July mean temperature profile under scenario A1B along the Jasper Tavernier transect. Figure 5-119. Projected 20902099 July mean temperature profile under scenario A1B along the Jasper Tavernier transect.

PAGE 247

247 Figure 5-120. Projected July mean temperature 2020 -2029, Scenario A2.

PAGE 248

248 Figure 5-121. Projected July mean temperature 2040 -2049, Scenario A2.

PAGE 249

249 Figure 5-122. Projected July mean temperature 2060 -2069, Scenario A2.

PAGE 250

250 Figure 5-123. Projected July mean temperature 2090 -2099, Scenario A2.

PAGE 251

251 Figure 5-124. Anomaly map for July temperatures, comparing 2040-2049 projections under scenario A2 to the 1971-2000 baseline.

PAGE 252

252 Figure 5-125. Anomaly map for July temperatures, comparing 2090-2099 projections under scenario A2 to the 1971-2000 baseline.

PAGE 253

253 Figure 5-126. Projected 20402049 July mean temperature profile under scenario A2 along the Jasper Tavernier transect. Figure 5-127. Projected 20902099 July mean temperature profile under scenario A2 along the Jasper Tavernier transect.

PAGE 254

254 Figure 5-128. Projected July minimum temperature 20 40-2049, Scenario B1.

PAGE 255

255 Figure 5-129. Projected July minimum temperature 20 90-2099, Scenario B1.

PAGE 256

256 Figure 5-130. Projected July minimum temperature 2040 -2049, Scenario A1B.

PAGE 257

257 Figure 5-131. Proj ected July minimum temperature 2090 -2099, Scenario A1B.

PAGE 258

258 Figure 5-132. Projected July minimum temperature 2040 -2049, Scenario A2.

PAGE 259

259 Figure 5-133. Projected July minimum temperature 2090 -2099, Scenario A2.

PAGE 260

260 Figure 5-134. Projected 2040-2049 July minimum temperature profile under scenario B1 along the Jasper Tavernier transect. Figure 5-135. Projected 2090-2099 July minimum temperature profile under scenario B1 along the Jasper Tavernier transect.

PAGE 261

261 Figure 5-136. Project ed 20402049 July minimum temperature profile under scenario A1B along the Jasper Tavernier transect. Figure 5-137. Projected 20902099 July minimum temperature profile under scenario A1B along the Jasper Tavernier transect.

PAGE 262

262 Figure 5-138. Projected 2040-2049 July minimum temperature profile under scenario A2 along the Jasper Tavernier transect. Figure 5-139. Projected 2090-2099 July minimum temperature profile under scenario A2 along the Jasper Tavernier transect.

PAGE 263

263 Table 5 -5. Projected July minimum tem perature gradient and mean slope between transect origin and terminus. Scenario Interval Gradient Mean Slope Baseline 1971 2000 3.97 C 0.60 C/100 km B1 204 0 20 49 3.71 C 0.56 C/100 km 2090 2099 3.87 C 0.59 C/100 km A1B 2040 2049 4.03 C 0.61 C/100 km 2090 2099 3.79 C 0.58 C/100 km A2 2040 2049 3.78 C 0.57 C/100 km 2090 2099 3.87 C 0.59 C/100 km

PAGE 264

264 Figure 5-140. Projected July maximum temperature 2040-2049, Scenario B1.

PAGE 265

265 Figure 5-141. Projected July maximum temperature 2090-2099, Scenario B1.

PAGE 266

266 Figure 5-142. Projected July maximum temperature 2040-2049, Scenario A1B.

PAGE 267

267 Figure 5-143. Projected July maximum temperature 2090-2099, Scenario A1B.

PAGE 268

268 Figure 5-144. Projected July maximum temperature 2040-2049, Scenario A2.

PAGE 269

269 Figure 5-145. Projected July maximum temperature 2090-2099, Scenario A2.

PAGE 270

270 Figure 5-146. Projected 2040-2049 July maximum temperature profile under scenario B1 along the Jasper Tavernier transect. Figure 5-147. Projected 2090-2099 July maximum temperature profile under scenario B1 along the Jasper Tavernier transect.

PAGE 271

271 Figure 5-148. Projected 2040-2049 July maximum temperature profile under scenario A1B along the Jasper Tavernier transect. Figure 5-149. Projected 2090-2099 July maximum temperature profile under s cenario A1B along the Jasper Tavernier transect.

PAGE 272

272 Figure 5-150. Projected 2040-2049 July maximum temperature profile under scenario A2 along the Jasper Tavernier transect. Figure 5-151. Projected 2090-2099 July maximum temperature profile under scenario A2 along the Jasper Tavernier transect.

PAGE 273

273 Table 5 -6. Projected July maximum temperature gradient and mean slope between transect origin and terminus. Scenario Interval Gradient Mean Slope Baseline 1971 2000 1.15C 0.17C/100 km B1 2040 20 49 1.40C 0.21C/100 km 2090 2099 1.25C 0.19C/100 km A1B 2040 2049 1.07C 0.16C/100 km 2090 2099 1.33C 0.20C/100 km A2 2040 2049 1.34C 0.20C/100 km 2090 2099 1.83C 0.28C/100 km

PAGE 274

274 F igure 5-152. CCSM projected January precipitation anomalies and their statistical significance under scenario B1 r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 2020 2029 a nomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 275

275 F igure 5-153. CCSM projected January precipitation anomalies and their statistical significance under scenario A1B r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 2020 2029 a nomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly A B C D E

PAGE 276

276 F igure 5-154. CCSM projected January precipitation anomalies and their statistical significance under scenario A2 r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 2020 2029 a nomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 277

277 F igure 5-155. CCSM projected Jul y precipitation anomalies and their statistical significance under scenario B1 r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 2020 2029 a nomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 278

278 F igure 5-156. CCSM projected Jul y precipitation anomalies and their statistical significance under scenario A1B r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 2020 2029 a nomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 279

279 F igure 5-157. CCSM projected Jul y precipitation anomalies and their statistical significance under scenario A2 r elative to 1990-1999 baseline A) 1990-1999 baseline. B) 2020 2029 a nomaly. C) 2040-2049 anomaly. D) 2060-2069 anomaly. E) 2090-2099 anomaly. A B C D E

PAGE 280

280 Figure 5-158. Projected January precipitation 20 20-2029, Scenario B1.

PAGE 281

281 Figure 5-159. Projected January precipitation 20 40-2049, Scenario B1.

PAGE 282

282 Figure 5-160. Projected January precipitation 20 60-2069, Scenario B1.

PAGE 283

283 Figure 5-161. Projected January precipitation 2090-2099, Scenario B1.

PAGE 284

284 Figure 5-162. Anomaly map for January precipitation, comparing 2040-2049 projections under scenario B1 to the 1971-2000 baseline.

PAGE 285

285 Figure 5-163. Anomaly map for January precipitation, comparing 2090-2099 projections under scenario B1 to the 1971-2000 baseline.

PAGE 286

286 Figure 5-164. Projected 2040-2049 January precipitation profile under sc enario B1 along the Jasper Tavernier transect. Figure 5-165. Projected 2090-2099 January precipitation profile under scenario B1 along the Jasper Tavernier transect

PAGE 287

287 Table 5 7. Projected January precipitation gradient and mean slope between transect origin and terminus. Scenario Interval Gradient Mean Slope Baseline 1971 2000 70 mm 10.6 mm/100 km B1 204 0 20 49 68 mm 10.3 mm/100 km 2090 2099 74 mm 11.3 mm/100 km A1B 2040 2049 68 mm 10.3 mm/100 km 2090 2099 71 mm 10.8 mm/100 km A2 2040 2049 70 mm 10.6 mm/100 km 2090 2099 65 mm 9.9 mm/100 km .

PAGE 288

288 Figure 5-166. Projected January precipitation 20 20-2029, Scenario A1B.

PAGE 289

289 Figure 5-167. Projected January precipitation 20 40-2049, Scenario A1B.

PAGE 290

290 Figure 5-168. Projected January precipitation 20 60-2069, Scenario A1B.

PAGE 291

291 Figure 5-169. Projected January precipitation 2090-2099, Scenario A1B.

PAGE 292

292 Figure 5-170. Anomaly map for January precipitation, comparing 2040-2049 projections under scenario A1B to the 1971-2000 ba seline.

PAGE 293

293 Figure 5-171. Anomaly map for January precipitation, comparing 2090-2099 projections under scenario A1B to the 1971-2000 baseline.

PAGE 294

294 Figure 5-172. Projected 2040-2049 January precipitation profile under scenario A1B along the Jasper Tavernier transect. Figure 5-173. Projected 2090-2099 January precipitation profile under scenario A1B along the Jasper Tavernier transect.

PAGE 295

295 Figure 5-174. Projected January precipitation 20 20-2029, Scenario A2.

PAGE 296

296 Figure 5-175. Projected January precipitation 20 40-2049, Scenario A2.

PAGE 297

297 Figure 5-176. Projected January precipitation 20 60-2069, Scenario A2.

PAGE 298

298 Figure 5-177. Projected January precipitation 20 90-2099, Scenario A2.

PAGE 299

299 Figure 5-178. Anomaly map for January precipitation, comparing 2040-2049 projections under scenario A2 to the 1971-2000 baseline.

PAGE 300

300 Figure 5-179. Anomaly map for January precipitation, comparing 2090-2099 projections under scenario A2 to the 1971-2000 baseline.

PAGE 301

301 Figure 5-180. Projected 2040-2049 January precipitation profile under scenario A2 along the Jasper Tavernier transect. Figure 5-181. Projected 2090-2099 January precipitation profile under scenario A2 along the Jasper Tavernier transect.

PAGE 302

302 Figure 5-182. Projected July precipitation 20 20-2029, Scenario B1.

PAGE 303

303 Figure 5-183. Projected July precipitation 20 40-2049, Scenario B1.

PAGE 304

304 Figure 5-184. Projected July precipitation 20 60-2069, Scenario B1.

PAGE 305

305 Figure 5-185. Projected July precipitation 2090-2099, Scenario B1.

PAGE 306

306 Figure 5-186. Anomaly map for July precipitation, comparing 2040-2049 projections under scenario B1 to the 1971-2000 baseline.

PAGE 307

307 Figure 5-187. Anomaly map for July precipitation, comparing 2090-2099 projections under scenario B1 to the 1971-2000 baseline.

PAGE 308

308 Figure 5-188. Projected 2040-2049 July precipitation profile under scenario B1 along the JasperTavernier transect. Figure 5-189. Projected 2090-2099 July precipitation profile under scenario B1 along the JasperTavernier transect.

PAGE 309

309 Table 5 -8. Projected July precipitation gradient and mean slope between transect origin and terminus Scenario Interval Gradient Mean Slope Baseline 1971 2000 59 mm 9.0 mm/100 km B1 204 0 20 49 79 mm 12.0 mm/100 km 2090 2099 86 mm 13.1 mm/100 km A1B 2040 2049 97 mm 14.8 mm/100 km 2090 2099 103 mm 15.7 mm/100 km A2 2040 2049 94 mm 14.3 mm/100 km 2090 2099 118 mm 17.9 mm/100 km

PAGE 310

310 Figure 5-190. Projected July precipitation 2020-2029, Scenario A1B.

PAGE 311

311 Figure 5-191. Projected July precipitation 20 40-2049, Scenario A1B.

PAGE 312

312 Figure 5-192. Projected July precipitation 20 60-2069, Scenario A1B.

PAGE 313

313 Figure 5-193. Projected July precipitation 2090-2099, Scenario A1B.

PAGE 314

314 Figure 5-194. Anomaly map for July precipitation, comparing 2040-2049 projections under scenario A1B to the 1971-2000 baseline.

PAGE 315

315 Figure 5-195. Anomaly map for July precipitation, comparing 2090-2099 projections under scenario A1B to the 1971-2000 baseline.

PAGE 316

316 Figure 5-196. Projected 2040-2049 July pre cipitation profile under scenario A1B along the Jasper Tavernier transect. Figure 5-197. Projected 2090-2099 July precipitation profile under scenario A1B along the Jasper Tavernier transect.

PAGE 317

317 Figure 5-198. Projected July precipitation 20 20-2029, Scenario A2.

PAGE 318

318 Figure 5-199. Projected July precipitation 20 40-2049, Scenario A2.

PAGE 319

319 Figure 5-200. Projected July precipitation 20 60-2069, Scenario A2.

PAGE 320

320 Figure 5-201. Projected July precipitation 2090-2099, Scenario A2.

PAGE 321

321 Figure 5-202. Anomaly map for July precipitation, comparing 2040-2049 projections under scenario A2 to the 1971-2000 baseline.

PAGE 322

322 Figure 5-203. Anomaly map for July precipitation, comparing 2090-2099 projections under scenario A2 to the 1971-2000 baseline.

PAGE 323

323 Figure 5-204. Projected 2040-2049 July precipitation profile under scenario A2 along the JasperTavernier transect. Figure 5-205. Projected 2090-2099 July precipitation profile under scenario A2 along the JasperTavernier transect.

PAGE 324

324 CHAPTER 6 LIMITATIONS TO BI OCLIMATIC MODELING Unfortunately, the power of climatic modeling is not unlimited, even as scientific monitoring of climate processes and computing power increase. This chapter describes some of the fundamental limitations of climatic modeling that need to be taken into account by researchers and stakeholders expecting precise forecasts analogous to short term weather report s. Climate projections, particularly when combined with ecological variables are powerful tools to constrain the range of possible fu ture scenarios and can help guide long term ecological management. They remain, however, projections rather than prescriptive forecasts, and are subject to multiple assumptions and limitations that should be understood by all users. Scientific U ncertainty Earths climate system is highly complex, consisting of atmosphere, hydrosphere, cryosphere, biosphere and land surface interacting with each other and responding to external forcings on timescales spanning many orders of magnitude. Not all components of the system are well understood. The IPCC (2007a) lists the following climate elements controlling the planetary energy budget as having a medium or low scientific level of understanding: stratospheric and tropospheric ozone, stratospheric water vapor from CH4, surface albedo changes due to land use and black carbon deposition on snow, aerosol effects, linear contrails and solar irradiance. These gaps in knowledge contribute to uncertainty about climate sensitivity (the expected temperature change associated with doubling CO2 concentrations), quantification of which is central to any GCM projection. Of particular interest to Florida are implications of climate change for hurricane frequency and intensity which are still subject to considerable debate, with published research yielding conflicting results about the relative contributions of increasing sea surface

PAGE 325

325 temperatures and wind shear in the Atlantic O cean (IPCC, 2007a ; Stanton & Ackerman, 2007; Vecchi & Soden, 2007; Saunders & Lea, 2008). The pace of climatology research is currently swift, and multiple satellite observation platforms are in development that will provide additional monitoring data to further enh ance scientific understanding. While a possibility of climate surprises, i.e. unexpected responses to increasing atmospheric GHG concentrations cannot be excluded, the skill that climate models have demonstrated at reproducing current and past climates, a nd their broad agreement for future climate scenarios, suggests that current understanding is sufficient to project future temperature changes with substantial confidence, while other climate variables, such as precipitation, a re subject to wider uncertain ty envelopes. It is well established that habitat ranges of species are fundamentally constrained by climatological, topographical and soil variables, as well as biotic interactions, each acting on distinct scale domains (Pearson & Dawson, 2003) Ecological systems remain imperfectly understood, exhibiting complex, non-linear, emergent and unpredictable behaviors (Harris et al ., 2006) causality of which is often difficult to identify. M ost systems and constituent species are infrequently or never monitored, and data availability tends to be poor (Williams & Jackson, 2007). This is exemplified by the geographic distribution of observational studies on climate change impacts on biota as summarized by the IPCC (2007b) which is heavily biased towards Europe and has substantial gaps in South America, Africa, Australia and Southeast Asia. Model Limitations Any model of complex, imperfectly known systems incorporates assumptions, lim itations, parameterizations and excluded variables. The same is true of climate and ecological models. GCMs require significant, expensive supercomputer time to run that increases exponentially with increasing resolution and the number of model components. The relatively coarse resolution of

PAGE 326

326 GCMs, while vastly improved in recent years, continue s to be inadequate to actively model micro and mesoscale weather patterns such as convective cloud formation and sea breezes, which contribute significantly to the annual water budget in Florida (Winsberg, 1990; Henry et al ., 1994). Model bias es in areas wh ere such processes are important tend to be inflated. Substantial knowledge gaps remain to be filled in the field s of meteorology and climatology as to the nature and mechanisms of natural climate cycles, predictability of frequency and intensity of precip itation events on multiple spatial scales, and formation, trajectories and intensities of tropical cyclones before precipitation/evaporation models can approach the utility of temperature models. Generally, climate models, including CCSM are limited in their ability to incorporate threshold events and rapid transitions to alternate stable states, instead producing relatively smooth outputs with gradual changes in rate. It is therefore imperative to be aware of the specter of relatively abrupt, nonlinear ch anges, the likelihood of which increases the longer comprehensive mitigation efforts are delayed. Thus, emphasis should be placed on mitigation as well as flexible adaptation approaches that can be adjusted to absorb the most serious impacts of abrupt change (Lenton et al ., 2008). Some evidence also suggests that IPCC projections, and the models upon which they are based, have been conservative and underestimate observed trends (Rahmstorf, 2007; Raupach et al ., 2007). Models used to assess climate change impacts on biota are also subject to substantial limitations. Advantages and shortcomings of bioclimate envelope models ( BCEMs ) are addressed in detail by Pearson & Dawson (2003) from which the following summary is derived. BCEMs assess changes in the distribution of species by drawing on ecological niche theory and the well established fact that climate is a primary constraint to habitat range at large spatial

PAGE 327

327 scales. Biotic interactions, evolutionary change and dispersal ability are generally not considered in BCEMs D istributions of many species are strongly correlated to climate variables, which define their bioclimatic envelope based on their currently realized niche under the assumption that it is in equilibrium with climate (Box et al ., 1993). Alternatively, a fundamental nichederived envelope may be derived by determining constraining climate variables experimentally, whereby care must be taken to include populations throughout the species range to incorporate their full phenotypic diversity. BCEM s are not suitable for projections at small spatial scales w here biotic interactions are increasingly important, for species with highly fragmented distributions, and for species with very short generation times for which the potential for rapid adaptation via evolutionary mechanisms is significant and may lessen t he need to disperse. Poor dispersers may not be capable of tracking rapid climatic changes and thus fail to colonize their shifting bioclimate envelope fully. G eographic and anthropogenic barriers as well as habitat fragmentation further undermines migration. In those cases, BCEM projections are likely to overestimate future habitat ranges. Assisted migration may be utilized by conservation managers to assist poor dispersers in expanding their range margins. Correlation -based BCEMs have a tendency to undere stimate potential future ranges if contemporary distributions are not in equilibrium with climate, or if climate is not a significant control at all. Physiology -based BCEMs, to the contrary, tend to overestimate future envelopes by modeling the entire fundamental niche, even if it is unlikely to be fully occupied. Nevertheless, both types of BCEMs have shown substantial skill at reproducing observed species distributions on large spatial scales for many species. Box et al (1993) have demonstrat ed that climatic envelope modeling is an appropriate tool for assessing species distributions in Florida.

PAGE 328

328 Many limitations of BCEMs can be addressed by implementing hierarchical mo deling that employ different types of models at different spatial scales. The power of BCEMs to provide a rapid initial assessment of potential future range shifts cannot be denied, but neither can its limitations. Combining BCEMs with ancillary informatio n such as soil type, slope, aspect, projected land use, etc., as is possible using GIS analysis, has the potential to greatly improve BCEM projection fidelity. Inherent S ystem P roperties A number of recent studies have highlighted important limitations to model ing climate and ecological systems in a fully quantitative way, not only because of remaining gaps in knowledge, but also due to inherently uncertain or chaotic system properties. Roe & Baker (2007) discuss the fundamental uncertainty of climate sensitivity. The heterogeneity and nonlinearity of the climate system implies that sensitivity can take on a range of values that may be increasingly narrowly constrained as scientific understanding advances, but cannot be precisely resolved. The discussion by Lenton et al (2008) on tipping elements in the climate system underlines th e difficulty anticipating thresholds beyond which non linear changes in climate system components are triggered. In some cases, the final drop that causes the proverbial bucket to overflow may not be a climate associated process at all. It is conceivable t hat sudden changes to dynamic flow of sensitive ice sheets in Greenland and West Antarctica, resulting in their partial or total disintegration, could ultimately occur in response to a tectonic event or a transient, localized weather extreme while climate system internal processes would not have caused a similar cascade for multiple years or decades. Beninca et al (2008) established that food webs can assume chaotic properties in the long term. While early ecological models had predicted such behavior and simplified experiments provided empirical support, the results by Beninca et al (2008) were based on the first such

PAGE 329

329 observation in a complex food web a plankton community isolated from the Baltic Sea. Their results, which the authors suggest may be applicable to other food webs of similar structure, show ed that predictability of community dynamics weakened progressively, and was lost within 15 to 30 days, corresponding to 5 to 15 plankton generations. Their findings have important implications for ecology in general, as well as climate change ecology specifically. Long term stability of a food web in spite of significant fluctuations in relative specie s abundance may suggest substantial resilience to functional modification, although the extent to which this can be extrapolated to more complex, dynamic ecosystems in general is unclear, particularly since the experimental environment was kept stable. More significantly, the progressive decline in species abundance predictability with time suggests that modeling future biotic interactions, and by extension structure and function of future ecosystems to which no current analogs exist may not be possible (see also Williams & Jackson, 2007). Beninca et al .s (2008) findings are highly significant if confirmed for other ecological systems, since they strong ly suggest that a major criticism of bioclimate envelope modeling lack of consideration of biotic interactions (Pearson & Dawson, 2003) may be overemphasized. C ollective human behavior, which impacts climate and biota at all levels of organization, is also inherently uncertain. The implication is that prescriptive bioclimate forecasts are unlikely to ever be produced, highlighting the impor tance of accepting and managing uncertainty as a key element of climate change ecology and conservation A combination of scenario -based, hierarchical modeling, extensive monitoring, and adaptive management embedded in a longterm visioning framework are considered essential elements of on the ground management of climate impacts to ecological and socioeconomic systems (Hannah et al ., 2002b; Pearson & Dawson, 2003).

PAGE 330

330 Figure 61. Percentage of climate models projecting globally and regionally non analog climates by 2100. A) Globally novel climates under scenario A2. B) Globally novel climates under scenario B1. C) Regionally novel climates under scenario A2. D) Regionally novel climates under scenario B1. Florida is identified as likely to be affected under each scenario. [Reprinted with permission from Williams & Jackson (2007) Novel climates, no analog communities, and ecological surprises. (p. 480, Fig. 4) Frontiers in Ecology and the Environment 5, 475-482].

PAGE 331

331 CHAPTER 7 ECOLOGICAL AND SOCIOECONOMIC IMPACTS ON FLORIDA Cha pter 3 introduced some of the processes through which climate ch ange and the biosphere interact on various spatial scales along with implications for conservation and development initiatives While many of the same principles apply to Florida, this chapte r integrates a review of local research with the results of the downscaling models presented in chapter 5 allowing more specific conclusions about 21st century climate change impacts on Florida Literature Review of Climate Change Impact Assessments for Florida Magnitude and Patterns of Recent and Future Change In general, the state of knowledge on how Florida has been and will continue to be impacted by climate change is poor, although academic and government agency interest in the topic has increased du ring the past five years, as indicated by a series of workshops, conferences, and special reports commissioned by governmental and non governmental organizations. The realization that Floridas climate will be affected by global climate change is not new; in fact, it was discussed, albeit mostly qualitatively over 15 years ago by Henry et al (1994) and Henry (1998). Mulholland et al (1997) discussed climate change in the southeastern United States from a freshwater ecology perspective based on RCM projec tions for future temperature and precipitation. Henry et al (1994) emphasize, among many other sources, that Floridas paleoclimatic history has been highly variable, ranging from total inundation to cool and arid, nearly desert like during periods of low sea level, as was the case during the last glacial maximum. The reader is referred to Watts (1980), Watts & Stuiver (1980) and Watts et al (1992) for detailed overviews of Floridas climatic evolution since the last glacial maximum from a

PAGE 332

332 paleolimnological perspective. Floridas climate stabilized at near modern conditions approximately 5000 years ago, and has fluctuated comparatively little since then. Henry et al (1994) state that Floridas present climate is strongly affected by change s in atmospheri c circulation and interannual variability, making it more difficult to extract signals of anthropogenic climate change from the background of natural cycles and variability. Particularly relevant are the position of the jet stream, phase of ENSO, land use changes, and a curious relationship between solar activity and prevailing wind patterns on Florida temperatures: High solar activity accompanied by highlevel winds from the west is associated with cooler air temperatures, whereas high solar activity associated with easterly winds tends to increase temperatures. During periods of low solar activity, the reverse effects are observed (Henry et al 1994). During the 20th century, the ge nera l trend in the southeastern United States including Florida, has been gradual cooling, particularly in rural areas, whereas rapidly expanding urban centers experienced increasing temperatures due to heat island effects averaging about 1.1C but on oc casion exceeding 8C (Henry et al ., 1994). The cooling trend was explained by a change in the position of the jet stream around 1950, allowing cold frontal air masses to penetrate further south and negating warming that was observed d uring the first half o f the century. Between 1950 and 1984, mean annual statewide temperatures dropped by 1.7C with the exception of Key West, where a slight warming trend of +0.3C continued. The decrease in winter temperatures between 1920 and 1990 was stronger in the panhandle and northern peninsula (-1.9 to -2.5C) than in the south (Henry et al ., 1994). (-1.1C to -1.9C) where widespread land cover change throughout the 20th century has been implicated as a contributing cause (Marshall et al. 2004; H. Swain pers. comm. May 2007).

PAGE 333

333 A renewed trend of warming began in the mid1980s, when a string of years with record temperatures began that continued into the 1990s. This transition in the mid1980s likely explains the variation in winter temperatures between study intervals observed b y Coenen (2005), since the 1985-1994 study period coincides with many of the years during which record temperatures were recorded. Henry (1998) updated his and co-authors 1994 assessment by adding that annual statewide precipitation has exhibited substantial variability, but no clear trends over the instrumental record could be discerned. However, rainfall during the month of Septemb er declined with statistical significance. Regionally, precipitation in parts of the panhandle and northeast Florida has increased (Henry, 1998). The trend towards record temperatures continued through the mid 1990s, with four of the warmest years on recor d falling between 1990 and 1996. According to Henry (1998), f uture summer precipitation reductions due to the influence of a stronger Bermuda -Azores system may be partially offset by the positive correlation between higher temperatures and higher frequency and intensity of summer rainstorms. He nry acknowledge d that while early models projected more frequent and intense tropical storm activity, more recent models have arrived at mixed conclusions, an area of uncertainty t hat persists to the present day (see Chapter 3). Mulholland et al (1997), basing their analysis in part of the IPCCs first assessment report released in 1990, disagree with the previous assessments on rainfall while acknowledging the general uncertainty of precipitation projections. Their analys is foresee s increased precipitation that would likely exceed evapotranspiration during summer, increasing water availability. Hydrographs would assume a more peaked shape, with more intense rainfall events followed by

PAGE 334

334 prolonged periods of drought. During winter, water stress would be increased, particularly in the panhandle and northern peninsula. Ecological Impacts Effects of climate change according to Henry et al (1994) are expected to be a slight drying trend and increasing temperatures, in part b ecause of a relocation of the jet stream to pre 1950s configuration and projected intensification and expansion of the BermudaAzore s high pressure system over the Atlantic Ocean. Water availability would suffer due to combination of decreased precipitatio n and higher temperatures, and, in particular, increased variability in precipitation, paradoxically resulting in more floods as well as droughts. Henry et al (1994) also discuss ed ecological impacts on habitat ranges, projecting an average northward shi ft of 97-145 km per 1 .0C increase in temperatures, restoring some of the tropical crops that thrive d in north -central Florida during the late 1800s to this area while warm temperat e species may be displaced northward by a total of 200 and 1000 km depending on the scenario. The 97 -145 km northward shift per 1C warming is remarkably close to an independently derived estimate from Coenen (2005), who projected shifts of 152 km per 1C warming based on existing thermal gradients Henry et al (1994) described that Florida would become viable habitat for several tropical species such as vampire bats with relatively minor increases in temperatures. Coral reefs are described as increasingly vulnerable to stress via bleaching with minor increases in ocean temperatures of 1.1C sustained, or 1.7C to 2.2C for short periods. Coral skeleton degradation due to acidification is also mentioned. Local s ea level rise has paralleled the global average as of the early 1990s while south Florida has been subsiding at a rate of 64 cm/century primarily due to compaction, oxidation and erosion of drained wetland soils Henry et al (1994) conclude d their discussion with some dire prognoses under extreme scenarios of climate change that include flooding of most of Floridas major cities

PAGE 335

335 should the West Antarctic ice sheet disintegrate, to massive losses of wetlands if sea level rise exceeds 0.5 m. Other issues raised include salt water intrusion, beach erosion, isolation/ degradation of barrier islands and loss of habitat for fish and shrimp. Lakes and streams would be subject to increased nutrient loading and eutrophication (Henry, 1998). Ecological impacts for the Gulf region according to Mulholland et al (1997) include increased rates of primary production due to lo nge r growing seasons, reduction in water quality and coldwater freshwater habitats, altered hydroperiods of wetlands, lakes and streams resulting in decreased organic matter storage and biotic stress, northward habitat range shifts, expansion of wetlands, inc reased eutrophication of lakes and changes in estuarine salinity regimes and water quality. Box and colleagues have produced a series of publications on modeling climate vegetation interactions based on their Florida Plant Species-Climate Model used to pre dict distributions of 125 woody plant species based on climatic variables, achieving median success rates of 85% to 88% (Box et al ., 1993). This early BCEM incorporated eight climate variables as measured over non -contiguous baseline periods for 106 weather stations in Florida and adjacent regions: Mean temperature of the warmest month Mean temperature of the coldest month The difference between the two (int ra -annual temperature range) Average annual precipitation Driest month precipitation Annual moisture index (mean annual precipitation divided by mean potential evapotranspiration) Mean minimum winter temperature Absolute minimum winter temperature

PAGE 336

336 It shou ld be emphasized at this point that all of the above variables can be derived from the downscaling methodology developed for my study Validation and verification of results comparing projected distributions to observational data was performed after five m odel iterations. With the exception of Key West, the model performed with high accuracy, with most discrepancies being due to near misses relative to low resolution species distribution maps used as ground truth data. The mean prediction error for individu al species was 29 km and the maximum prediction error 116 km, with the model performing approximately equally well within each of Floridas four forest types (temperate panhandle/upper peninsula, warm temperate, subtropical, and coastal subtropical). I ntra -annual temperature range, average annual precipitation and absolute minimum winter temperatures were found to be redundant after analysis. With their model, Box et al (1993) confirmed that the distribution of most woody plant species in Florida is primarily controlled by climatic variables, that their distribution c ould modeled with acceptable precision even when non climatic factors are not taken into account, and that effects of climate change c ould successfully be projected using the same approach with an accuracy of about 100 km, which could be further improved by incorporating topography, soil types, and anthropogenic barriers into the model They caution ed that modeling rare and exotic species is considerably more difficult because of a lack of infor mation on their climatic tolerance, particularly for exotics that are still expanding their range and have not yet reached equilibrium with climate in their new environment. Subsequently, species distributions for 124 native trees and shrubs were evaluated under 12 study specific scenarios of climate change, including a more detailed analysis of a subset of 28 species under 6 scenarios using the Florida Plant SpeciesClimate Model (Box et al ., 1999). The climate scenarios were limited to a maximum of 2C warming and are therefore

PAGE 337

337 conservative compared to most IPCC scenarios, which have in turn been revealed as conservative by more recent investigations of emission trends. ( C anadell et al ., 2007; Raupach et al ., 2007). Results quantitatively confirm the hypothesis of range collapses of temperate (76% areal loss) and warm temperate climate envelopes (36% areal loss), whereas subtropical climate envelopes increase in areal extent by an average of 82% across the six scenarios. Even the most conservative scen ario (1C of warming and no change in moisture balance) indicates a reduction in areal extent of panhandle/northern peninsular climate envelope s by 41%, increasing to 69% if the mean annual temperature anomaly is distributed asymmetrically, i.e. greater wa rming during winter than summer. When combined with a reduction in precipitation of 20% these values increase further, approaching 97% for one scenario. Range expansion of the subtropical climate envelopes is particularly strong under the asymmetric scena rio (up to 145% increase in area), reflecting that the main constraining factor on the distribution of subtropical and tropical taxa are minimum winter temperatures. Crumpacker et al (2001a) further elaborated on Florida Plant SpeciesClimate Model projec tions by summarizing conservation implications for native Florida trees projecting temperate hardwood forests and swamp species as most severely affected Scenarios assuming reduction in precipitation would result in conversion of large swaths o f currentl y forested areas to lightning and wildfire controlled scrub and savanna systems. Adjacent 100x100 km grid cells at times produced very different projections and, therefore, imply regionally di verse conservation challenges. Offered solutions parallel those mentioned for global conservation efforts in Chapter 3 (Coenen et al ., 2008) and Coenen (2005; described below), and include boosting overall system resilience by controlling exotic invasive species restoring more natural hydrological regimes, and, with extensive monitoring and safeguards, engaging in assisted colonization where

PAGE 338

338 maximum rates of natural migration are slower than projected habitat range shifts. Concrete examples for potential species transfers between conservation areas are given. They fur ther introduce d the notion of early war ning systems by monitoring for initial signs of biotic resp onses, a thread they explored in a separate publication (Crumpacker et al ., 2002 ). They also applied the model to estimate disruption of extant plant communities, concluding that such will be common occurrences in the panhandle and northern two thirds of the peninsula, whereas dominance of tropical and subtropical species in south Florida will increase (Crumpacker et al ., 2001b). Coenen (2005) drew the following conclusions from his research on recent and future impacts on freshwater ecosystems1 1 This section is an updated discussion originally written for Coenen (2005), to which the author retains copyright. : b ecause species differ in sensitivity to climatic changes and habitat modification, as well as their ability to adapt by migration altered phenology, behavior or evolutionary mechanisms, it is expected that disruption of community structures, including food webs and coevolved symbiotic relationships, are likely to occur (Hughes, 2000, Crumpacker et al ., 2001; McCarty 2001; Parmesan 2006). The constancy of mean summer temperatures between 1975 and 2004 implies no immediate threat of exceeding thermal tolerance limits of warm temperate, heatlimited species near their southern distributional limit. Winters are, however, in many respects the mo re influential season in terms of ecology, because winter temperatures constrain habitat ranges of cold limited exotic invasive species (Thomas et al ., 2001), and, in lentic, monomictic ecosystems, the duratio n and timing of winter mixing (Beaver et al ., 1981), affecting nutrient cycling and hypolimnion oxygenation. In eutrophic lakes with high rates of biomatter decay, earlier onset of stratification and later breakdown thereof increases the potential for bottom water deoxygenation, adversely impacting benthic communities.

PAGE 339

339 T he observed pattern of strong warming followed by a moderate temperature reversion suggests highly dynamic ecological processes along the subtropical-transitional zonal boundary. It is possible that succession dynamics analogous to th e seasonal changes in relative abundance of subtropical and temperate assemblages of Cladocera in Florida lakes, as described by Frey (1982), occur on interannual to interdecadal timescales in response to temperature variability. Specifically, tropical and subtropical species would be expected to colonize new habitats north of their present range following successive mild winters, with transitional and temperate species rebounding following prolonged cold (cf. Walther et al ., 2002). Since the net trajectory through 2004 indicated winter warming it is possible that some range expansion of subtropical species, paralleling zonal boundary shifts, has in fact already occurred but is awaiting observational confirmation Associated ecosystem structural adjustments and microevolutionary adaptation are likely to have increased resilience and capacity for recovery of invading species following future disturbances. It must be emphasized, however, that Coenen (2005) utilized temperature means as the basis for analysis, whereas species tend to be more sensitive to minima and maxima, as well as extreme weather conditions that are often transient in duration and impossible to predict more than a few days into the future. A ny impacts directly caused by global climate change have to date been subtle relative to more acute, local stresses, particularly land cover modification. Generalists capable of rapid expansion are likely to be at a competitive advantage over more sedentary organisms (Thomas et al ., 2001), particularly specialists whose preferred foods or shelter may display different thermal tolerances or rates of dispersal, disrupting coevolved species interactions. Such rapid change in ecosystem structures is likely to result in undesirable consequences for ecosystem functions, including utilitarian uses. Conservation planners are urged to incorporate projections for climate forced habitat adjustments into design and

PAGE 340

340 management plans to ensure that conservation goals are reachable and sustainable in a time of accelerat ing global change where historical benchmarks are increasingly unreliable conservation targets (Harris et al ., 2006). According to Williams & Jackson (2007), magnitude and spatial distribution of community disruption risks have not yet been evaluated. Their own projections compared climate dissimilarity by grid cell (2.8 x 2.8 degrees) and assigned the label globally novel to each cell for which changes in an index comprised of mean summer temperature, winter temperature, summer precipitation and winter precipitation exceeded a threshold. Results indicate a high concentration of novel climates in tropical and subtropical regions, including Florida for IPCC scenarios A2 and B1. They additionally defined regionally novel climates as those grid cells for w hich no current analog exists within a 500 km radius, which has been estimated as the maximum distance over which plants can disperse without human interference by the beginning of the next century. More than 70% of climate models queried estimate that Flo rida will feature regionally novel climates by 2100, with a greater percentage agreeing for scenario A2 ( Figure 61). Of particular concern to lake managers will be increased incidence of deoxygenation, particularly in eutrophic lakes. In addition to the p reviously described mechanism for deoxygenation of hypolimnia of stratified lakes associated with shorter winter mixing periods, shallow, non stratified lakes will likely experience more frequent oxygen stress and fish kills: Higher temperatures are associated with decreased dissolved oxygen storage capacity and increased biological productivity (and, therefore, greater biological oxygen demand). Periods of reduced photosynthesis due to prolonged summer cloudiness, input of high quantities of allochthonous organic matter following heavy precipitation events and other conditions that result in respiration exceeding photosynthetic oxygen production for periods of multiple days

PAGE 341

341 will induce hyp oxia more rapidly than is presently the case. Installation of aerato rs in the most sensitive system may provide relief. A more peaked hydro logy particularly extended droughts, would depress water levels, increase benthic temperatures, reducing connectivity with fringing wetlands and other ecotones, and possibly result in temporary desiccation of shallow lakes and wetlands. Mulholland et al. s (1997) statement of likely expansion in wetland area is an optimistic one, as many coastal wetlands are constrained between rising seas and urban development, and expansion of freshwater wetlands is likely to come at the expense of openwater habitats, since lake shores continue to be prime real estate and are often highly developed. Socioeconomic Impacts The most comprehensive assessment on Florida specific socioeconomic impacts to date has been by Stanton & Ackerman (2007), who tabulated economic costs of failing to take mitigative and adaptive measures by comparing a business as usual with a rapid GHG stabilization scenario under which atmospheric concentrations w ould s tabilize at 450 ppm. While basing their assessment on the latest IPCC reports they used slightly different scenarios that assume 1.2C of warming and 18 cm of sea level rise for the optimistic vision, and 5.4C of warming along with 115 cm of sea level ri se relative to year 2 000 temperatures by 2100 under the pessimistic scenario, which assumes peak CO2 concentrations of 850 ppm. The 450 ppm scenario would result in socioeconomic and ecological pressures on Florida being primarily caused by population growth and unsustainable development patterns, with climate change being a n additional, but not dominant, contributor. The business as usual scenario, to the contrary, would result in massive impacts, including 9% of Floridas current landmass being inundated by the seas by 2060, most notably nearly 100% of Monroe County and more than two thirds of Miami-Dade County. This includes much of the Everglades, which state and federal government s are currently investing substantial financial resources in for restorat ion

PAGE 342

342 efforts. Presently 1.5 million people inhabit the vulnerable areas, along with numerous major infrastructure facilities, nuclear reactors, Superfund sites, cultural and historical sites as well as many of Floridas beach resorts important to the tour ism industry. Preventing sea level rise in Florida is complicated by its porous, limestone based geologic platform that is not conducive to protecting large areas via levees and storm surge barriers such as the ones protecting the Netherlands. Instead, protection will likely be limited to relatively small sites of particular significance, using a combination of engineering techniques to maintain dry conditions. A particularly interesting perspective is given by Stanton & Ackermann (2007) on natural resourc e use, which unlike previous studies, integrates all known climatic factors rather than looking at them individually. Unless technological solutions such as salt water desalinization can be deployed economically on a wide scale, the results are economic lo sses in agriculture, forestry and fisheries primarily cause d by a mismatch of required freshwater for irrigation and its increased scarcity due to higher evapotranspiration and salt water intrusion. Ecosystem impacts parallel those described by previously summarized sources. Globally, the total cost of unmitigated climate change in the form of direct human impacts and indirect impacts resulting from the disruption of ecosystem services ha s been estimated to reach 20% of the global gross domestic product by 2100, implying that the cost of inaction vastly exceeds the cost of comprehensive mitigation (Stern, 2007). In Florida, the cost of inaction assessed on impacts on tourism, electricity consumption, real estate and hurricane damage alone w as estimated to amount to $345 billion in 2006 U.S. dollars, or 5% of projected GSP by 2100 (Stanton & Ackerman, 2007). Implications of Downscaling Results My study presents climate projections for Florida in unprecedented quantitative detai l. It allows for production of customiz ed change maps and is ready to integrate additional GIS layer

PAGE 343

343 data to yield highresolution impact projections. As has been shown, bio climatic model s are essential for assessing ecological impacts and guiding conserv ation efforts despite their previously discussed limitations and as yet limited power of such models. Until necessary ecological ground truth become s available and a fully integrated bioclimatic model for Florida is developed, implication s can be assessed qualitatively by compari ng existing impact literature to the results of this study Areas Subject to Greatest Change Temperature All temperature downscaling models reveal a flattening of the January climatic gradient, with the northwestern panhandle warmi ng the most and anomalies decreasing in a generally southwestern direction. Scenarios are differentiated primarily by magnitude of average statewide warming whereas there is little difference between scenarios as far as the steepness of the Jasper Taverni er and total temperature gradients are concerned. In Thornton Wilhelmi anomaly maps, January warming reaches statistical significance levels at p<0.01 for all Floridabased grid cells, although the timing at which this high degree of s ignificance is achieved varies spatially. Unlike January, patterns of vulnerability to temperature change are less consistent in July reaching their maxima in different locations between scenarios Generally speaking, summer temperature anomalies under scenarios B1 and A1B show very little regional differentiation, affecting the southern panhandle, north ern Florid a and the Gulf C oast north of Tampa Bay just slightly more than the rest of the state. Scenario A2 projects a pattern similar to January temperatures, meaning t hat the northwestern panhandle will be subject to the greatest temperature change. Under all scenarios, July warming reaches statistical significance levels at p<0.01 in ThorntonWilhelmi anomaly maps for all Florida -based grid cells.

PAGE 344

344 These findings lend s trong support to the projections by Box et al (1999). Temperate and warm temperate taxa, including endemic flora, fauna, and relict populations in the panhandle will experience significant to total habitat loss due to a combination of two climate related factors, which act in addition to existing, non climatic stresses. 1. General p hysiological temperature stress, reducing fitness of species at the southern margin of their distribution exacerbated by the fact that two out of three scenarios project gre ater s ummer than winter warming 2. Reduction in climatic barriers to northward expansion of subtropical and tropical taxa due to the decreasing slopes of winter temperature gradients The se point s support Williams & Jacksons (2007) projection of Florida being among the most vulnerable areas for the formation of transient and novel ecosystems without past or present analogs. As subtropical and tropical taxa expand the northern margin of their range, they will increasingly compete with and exploit resources made avail able by failing temperate and warm temperate communities. Barriers to the expansion of tropical taxa vary by means of propagation method and habitat specificity. Many freshwater species will experience little difficulty migrating northward, as connectivity between lakes and wetlands is high due to numerous water control structures (primarily canals), streams, and human recreational activities that inadvertently or intentionally introduce species to new systems However, different water chemistry between sys tems and the specter of increased incidence of hypoxia or temporary desiccation of freshwater and wetland systems during hot and dry summers places limits on migration potential. Terrestrial specialists will experience greater difficulty crossing anthropogenic barriers such as the densely developed I4 co rridor. Figure 71, for instance, shows a magnified temperature projection for the Tampa Bay area under scenario A2. Figure 7 -2 demonstrates integration of projected temperature change, 2 m of sea level rise, and future development

PAGE 345

345 projections for 2060 to illustrate how multiple agents of change act simultaneously to promote or retard ecological and socioeconomic adaptation to climatic change. Precipitation Downscaling results reveal that assumptio ns of homogenous precipitation change throughout the state are unrealistic since study results demonstrate that changes in precipitation expressed as amounts in mm or percentage relative to baseline conditions vary widely between seasons and scenarios. Ja nuary anomalies are close to neutral for scenarios B1 and A1B, never reaching statistical significance (with the possible exception of the Collier County area for 20602069 under scenario A1B). The indications, however tenuous, point to very slightly dryer conditions in North Florida and the Keys and slight wetting elsewhere T he magnitude of projected January anomalies is so small relative to internannual variability that a trend would be very difficult to confirm. Scenario A2 produces similar patterns, b ut of greater magnitude, reaching statistical significance during 2060 -2069 before reverting. Given these results, January precipitation is expected to change relatively little, with drying possible in some of Floridas most ecologically sensitive areas in the panhandle Since the panhandle is also projected to experience high er than average warming, water budgets will inevitably be altered, but more detailed hydrologic investigation will be required to quantify impacts on soil moisture levels, runoff rates wetlands hydroperiods, lake levels and aquifer recharge. The remainder of the state, with the exception of the Keys, appears to be less affected by winter precipitation changes, but hydrological models will need to be run here as well to draw conclusions on changes to regional water budgets. Naturally, future development, agricultural practices and other direct human impacts will also need to be taken into account, necessitating an additional scenario -based modeling step.

PAGE 346

346 During July, strengthening of the BermudaAzores high pressure system is implicated in substantial drying of oceanbased grid cells over the western Atlantic extending to the Florida peninsula. N o evidence for a significant compensating effect due to increased convective precipitation as was suggested by Henry (1998), is evident. Anomalies generally increase to the southeast, in sharp contrast to Mulholland et al .s (1997) projections. The panhandle, to the contrary, is projected to experience few changes in summer rainfall, tending towar ds a slight increase. Declines in expected precipitation are particularly severe in the Florida Keys, which are projected to lose more than 50% of current rainfall under scenario A1B, and in excess o f 75% under scenario A2, strongly suggesting that while t emperature projections generally favor tropical species in South Florida, precipitation projections favor droughtresistant, water efficient taxa typical of scrub and dry tropical savanna biomes. Tropical species lacking adaptations to drought will come un der increasing pressure to adjust the trailing edge of their bioclimatic envelopes to the northwest, a scenario that has not previously been considered for Florida. Further, i t is a distinct possibility that some native subtropical scrub and savanna species will be able to expand their habitat ranges into areas most affected by drought. While the general pattern of habitat range shifts would follow the temperature -forced south to north pattern, the results of this study now raise the possibility of alternate trajectories for some taxa, potentially even in the opposite direction to the general trend. This adds an extra level of complexity for bioclimatic models and conservation m anagers. While general conclusions of Box et al .s (1999) BCEM are supported, none of their six main climate scenarios match CCSM downscaling results well. Box et al .s (1999) scenarios a ssumed temperature change is generally lower than projected for Florida G reater warming

PAGE 347

347 during winters than summers contradicts projections under scenarios A1B and A2, and assumptions of a 20% statewide decline in precipitation do not r eflect substantial spatial differentiation in July precipitation anomalies. Inter -S cenario Uncertainties The inter scenario range between mean statewide temperature anomalies in the 2090 -2099 period is 1.32C during January and 2.16C during July. Scenario B1 projects slightly greater warming during January than July, whereas the other s cenarios project more summer than winter warming. In terms of precipitation, inter scenario ranges are more fluid between study intervals and are probably less meaningful for practical application due to the discontinuous nature of rainfall. During January the inter scenario uncertainty is 7 mm in 2040 -2049 and 2090-2099. During July, the mean inter -scenario difference in 20402049 is 13 mm, increasing to 21 mm in 2090-2099. In summary, most literature published on climate change and its impacts o n Florida, while valuable, has been qualitative in nature. Too frequently, authors downscaled linearly from global projections without taking local ground truth or model biases into account. Quantitative studies tended to be based on studyspecific climate change scenarios that make mutual comparisons difficult and do not always reflect the projections generated for this study. They generated partially contradictory impact projections. They frequently considered variables in isolation rather than as compone nts of a complex and dynamic system. Ground truth, specifically concerning range shift and phenological monitoring data, as well as fundamental physiological data for many species and the communities within which they function, remains sparse and largely l imited to species presence lists in conservation areas and dated, coarse resolution range maps.

PAGE 348

348 Figure 7-1. High-resolution detail of mean January temperature projection for the Tampa Bay area for scenario A2 during 2090-2099.

PAGE 349

349 Figure 7-2. High -resolution detail of mean January temperature projection for the Tampa Bay area for scenario A2 during 2090-2099, incorporating anticipated urban growth and 2 m of sea level rise.

PAGE 350

350 CHAPTER 8 CONCLUSIONS Outlook In 1990, Florida was the ni nth largest emitter of GHGs in the United States By 2004, emissions had increased by an additional 38% (EPA, 2008) due to a combination of rapidly increasing population and political apathy, making the state the sixth -ranked GHG polluter in the U.S. (Mulkey, 2007). The results of this study clearly indicate that i mplementation of comprehensive mitigation and adaptation measures is beyond due, since inaction would result in disintegration of Floridas socioeconomic, ecological and cultural fabric. Being amo ng the most vulnerable areas to climate change, Florida must lead by example and begin to realize its considerable potential for mitigation and adaptation using principles of sustainable resource use with heavy investment in carbon neutral infrastructure a nd climate change integrated conservation strategies In doing so, resilience to climate change impacts would be enhanced while simultaneously accruing substantial synergistic benefits in the short term improving human and environmental health, aesthetics and quality of life. Planning for and monitoring of climate change impacts based on results of this study are desperately needed measures. The design of this study was intentionally flexible and fit for adaptation by stakeholders across ecological, economic, infrastructure and other sectors. While the costs of addressing climate change comprehensively are not trivial, they increase with each passing year of apathy while chances for success are progressively reduced. Future Directions and Challenges CCSM i s just one of more than two dozen GCMs widely used to project Earths climate into the future. As resolutions of other climate models increase sufficiently to join CCSM as capable of rendering Florida as peninsula, downscaled projections based on multimod el

PAGE 351

351 ensemble s will be come possible, allowing better quantification of uncertainties, identification of outliers, and overall more refined results. It is hoped that more modeling groups will make their output data available in readily accessible GIS format while the 5th assessme nt report of the IPCC is being prepared. In the intermediate future, work will resume on my GIS based GCM downscaling method, focusing on the following areas of improvement: Development of automated scripts to perform data preparatio n and downscaling steps more rapidly, allowing products for more months and more c limate variables to be produced Development of a data viewer through which endusers can rapidly assess impacts relevant to their sector Evaluating alternative models to redu ce prediction errors for precipitation maps, including IDW, multivariate kriging, regression kriging, and discontinuous kriging Integrating a mesoscale hydrology model to assess net impact s of higher temperatures and changing precipitation regimes on water availability for ecological, agricultural and infrastructure systems Regional, very high resolutions assessments of ecological transition areas such as the I -4 corridor Integrating biotic data to yield an adva nced BCEM capable of generating rapid assessments of future plant and animal habitat range shifts Compared to conventional BCEMs, results will be improved by incorporating additional (i.e. non climate related) habitat requirements and projections of availabilit y of those requirements under various scenarios of climate change and population growth/urban expansion. In practice, this will supply agencies and conservation managers with high resolution estimates on the magnitude of anticipated change, including likely impacts on key conservation objectives, and guidance on suitable, cost effective adaptation measures. This includes identification of at risk species and ecosystems as well as delineation of migration corridors to prepare fixed -boundary conservation area s and the unprotected matrix of private, ever more urbanized land between

PAGE 352

352 them to the reality of increasingly dynamic ecosystems responding to species individualistic habitat range shifts and phenological changes. The model will be expanded beyond Florida to encompass adjacent states and other land areas on a per -project basis. Complementing this modeling work, monitoring ecological processes at the increasingly dynamic transition zones between tropical and subtropical, and subtropical to warm temperate climate zones is a future objective Via multiyear monitorin g, the biotic components of combined downscaling/BCEM model s will be validated and crucial information about alteration in ecosystem structure and function in response to shortterm weather perturb ations and long term climatic trends gained in situ Opportunities for controlled experiments in the laboratory or field plots include quantifying thermal tolerance limits and other abiotic limiting factors for species for which this information is not yet known, as well as assessments of climate change impacts on plants, soils and soil microorganisms by monitoring their responses to modified environmental conditions as projected by climate models in a controlled environment. The results of these experiment s will in turn inform and revise model projections. Combined, the modeling monitoring, field and laboratory aspects of my post doctoral research plan will provide opportunities for collaboration with and funding by a multitude of agencies and private industry ranging from insurance to utilities.

PAGE 353

353 APPENDIX NAMES AND LOCATIONS OF REFERENCE WEATHER STATIONS Table A -1. Official n ames and locations of reference weather stations in Florida, Alabama and Georgia. Station Name Latitude (N) Longitude (E) ANDALUSIA 3 W, AL 31.30 -86.53 APALACHICOLA WSO ARPT 29.73 -85.03 ARCADIA 27.23 -81.85 ARCHBOLD BIOLOGICAL STN 27.18 -81.35 AVON PARK 2 W 27.60 -81.53 BARTOW 27.90 -81.85 BAY MINETTE 3 NNW, AL 30.92 -87.78 BELLE GLADE EXP STN 26.65 -80.63 BRADENTON 5 ESE 27.45 -82.47 BREWTON 3 SSE, AL 31.07 -87.05 BROOKSVILLE CHIN HILL 28.62 -82.37 BRUNSWICK FAA, GA 31.15 -81.38 BRUNSWICK, GA 31.17 -81.50 BUSHNELL 2 E 28.67 -82.08 CAMILLA 3 SE, GA 31.18 -84.20 CANAL POINT USDA 26.87 -80.62 CHIPLEY 3 E 30.78 -85.48 CLERMONT 7 S 28.45 -81.75 CLEWISTON US ENGINEERS 26.75 -80.92 COLQUITT 2 W, GA 31.17 -84.77 CROSS CITY 2 WNW 29.65 -83.17 DAYTONA BEACH WSO AP 29.18 -81.05 DE FUNIAK SPRINGS 30.73 -86.07 DELAND 1 SSE 29.02 -81.30 DEVILS GARDEN 26.60 -81.13 ENTERPRISE 5 NNW, AL 31.38 -85.90 EVERGLADES, FLORIDA 25.85 -81.38 FAIRHOPE 2 NE, AL 30.55 -87.88 FEDERAL POINT 29.75 -81.53 FERNANDINA BEACH 30.65 -81.47 FLAMINGO RANGER STN 25.15 -80.92 FOLKSTON 9 SW, GA 30.73 -82.13 F ORT DRUM 5 NW 27.58 -80.83 FORT LAUDERDALE 26.10 -80.20 FORT MYERS FAA/AP 26.58 -81.87 FORT PIERCE 27.47 -80.35 FRISCO CITY 3 SSW, AL 31.38 -87.42

PAGE 354

354 Table A -1. Continued Station Name Latitude (N) Longitude (E) GLEN ST MARY NURSERIES 30.27 -82.18 HEADLAND, AL 31.35 85.33 HIALEAH 25.83 80.28 HIGH SPRINGS 29.83 -82.60 HOMERVILLE 3 WSW, GA 31.03 82.80 IMMOKALEE 3 NNW 26.47 81.43 INVERNESS 3 SE 28.73 -82.32 JACKSONVILLE BEACH 30.28 81.40 JACKSONVILLE WSO AP 30.50 81.70 JASPER 30.52 -82.95 KEY WEST WSO AIRPORT 24.55 81.75 KISSIMMEE 2 28.28 81.42 LA BELLE 26.75 -81.43 LAKE ALFRED EXP STN 28.10 81.72 LAKE CITY 2 E 30.18 82.60 LISBON 28.87 -81.78 LIVE OAK 30.28 82.97 MADISON 4 N 30.53 83.43 MAYO 30.05 -83.17 MELBOURNE WSO 28.12 80.65 MIAMI BEACH 25.78 80.13 MIAMI WSCMO AIRPORT 25.80 -80.30 MILTON EXPERIMENT STN 30.78 87.13 MONTICELLO 3 W 30.53 83.92 MOORE HAVEN LOCK 1 26.83 -81.08 MOULTRIE 2 ESE, GA 31.17 83.75 MOUNTAIN LAKE 27.93 81.60 MYAKKA RIVER STATE PARK 27.23 -82.32 NAHUNTA 3 E, GA 31.22 81.93 NAPLES 26.17 81.78 NICEVILLE 30.52 -86.50 OCALA 29.20 82.08 OKEECHOBEE HRCN GATE 6 27.22 80.80 ORLANDO WSO MCCOY 28.45 -81.32 PANAMA CITY 5 NE 30.22 85.60 PARRISH 27.57 82.43 PENSACOLA FAA ARPT 30.47 -87.20 PERRY 30.10 83.57 PLANT CITY 28.02 82.13 POMPANO BEACH 26.23 -80.15 PUNTA GORDA 4 ESE 26.92 82.00

PAGE 355

355 Table A -1. Continued Station Name Latitude (N) Longitude (E) QUINCY 3 SSW 30.60 -84.55 QUITMAN 2 NW, GA 30.80 83.58 ROBERTSDALE 5 NE, AL 30.62 87.67 SANFORD EXPERIMENT STN 28.80 -81.23 SAPELO ISLAND, GA 31.40 81.28 ST LEO 28.33 82.27 ST PETERSBURG 27.77 -82.63 STEINHATCHEE 6 ENE 29.72 83.30 STUART 1 N 27.22 80.25 TALLAHASSEE WSO AP 30.38 -84.37 TAMIAMI TRAIL 40 MI BEN 25.75 80.83 TAMPA WSCMO ARPT 27.97 82.53 TARPON SPRINGS SWG PLNT 28.15 -82.75 TAVERNIER 25.00 80.52 THOMASVILLE 3 NE, GA 30.88 83.93 TITUSVILLE 28.62 -80.82 USHER TOWER 29.42 82.82 VENICE 27.10 82.43 VERO BEACH 4 W 27.63 -80.45 WAUCHULA 2 N 27.57 81.82 WAYCROSS 4 NE, GA 31.25 82.32 WEEKI WACHEE 28.52 -82.58 WEST PALM BEACH WSO AP 26.68 80.12 WEWAHITCHKA 30.12 85.20 WINTER HAVEN 28.02 -81.75 Note: Data assembled from SERCC (2010) metadata

PAGE 356

356 LIST OF REFERENCES Alvarez, L.W., Alvarez, W., Asaro, F. & Michel, H.V. (1980) Extraterrestrial c ause for the C retaceous -T ertiary e xtinction. Science 208, 1095-1108. Ashiq, M.W., Zhao, C., Ni, J. & Akhtar, M. (2010) GIS based high resolution spatial interpolation of precipitation in mountain plain areas of Upper Pakistan for regional climate change impact studies. Theoretical and Applied Climatology 99, 239-253. Azqueta, D. & Delacmara, G. (2006) Ethics, economics and environmental management. Ecological Economics 56, 524-533. Baigorria, G.A., Jones, J.W. & O'Brien, J.J. (2007) Understanding rainfall spatial vari ability in southeast USA at different timescales. International Journal of Climatology 27, 749-760. Beaver, J.R., Crisman, T.L. & Bays, J.S. (1981) Thermal regimes of Florida lakes. Hydrobiologia, 83, 267-273. Beninca, E., Huisman, J., Heerkloss, R., Jhn k, K.D., Branco, P., Van Nes, E.H., Sheffer, M. & Ellner, S.P. (2008) Chaos in a longterm experiment with a plankton community. Nature 451, 822-825. Beyer, H.L. (2004) Hawth's Analysis Tools for ArcGIS Version 3.27. Available at : http://www.spatialecology.com/htools (last accessed 9 June 2010). Bierbaum, R.M. & Raven P.H. (2007) A t wo -pronged c limate s trategy. Science 316, 17. Boer, E.P.J. de Beurs, K.M. & Hartkamp, A.D. (2001) Kriging and thin plate splines for mapping climate variables. International Journal of Applied Earth Observation and Geoinformation, 3, 146-154. Bonan, G.B. (2008) Forests and climate change: forcings, feedbacks, and the climate benefits of forests. Scienc e 320, 1444-1449. Bowen, G.J., Clyde, W.C., Koch, P.L., Ting, S., Alroy, J., Tsubamoto, T., Wang, Y. & Wang, Y. (2002) Mammalian d ispersal at the Paleocene/Eocene boundary. Science 295, 2062-2065. Box, E.O., Crumpacker, D.W. & Hardin, D.E. (1993) A clima te model for location of plant species in Florida, U.S.A. Journal of Biogeography 20, 629-644. Box, E.O., Crumpacker, D.W. & Hardin, D.E. (1999). Predicted effects of climatic change on distribution of ecologically important native tree and shrub species in Florida. Climatic Change 41, 213-249. Brooks, T.M., Mittermeier, R.A., de Fonseca, G.A.B., Gerlach, J., Hoffmann, M., Lamoreux, J.F., Mittermeier, C.G., Pilgrim, J.D. & Rodrigues, A.S.L. (2006) Global biodiversity conservation priorities. Science 313, 58-61.

PAGE 357

357 Brown, D.P. & Comrie, A.C. (2002) Spatial modeling of winter temperature and precipitation in Arizona and New Mexico, USA. Climate Research 22, 115-128. Brown, K. (2002) Innovations for conservation and development. Geographical Journal 168, 6 17. Canadell, J.G., Le Quere, C., Raupach, M.R., Field, C.B., Buitenhuis, E.T., Ciais, P., Conway, T.J., Gillet, N.P., Houghton R.A. & Marland, G. (2007) Contributions to accelerating atmospheric CO2 growth from economic activity, carbon intensity, and efficiency of natural sinks. Proceedings of the National Academy of Sciences 104, 18866-18870. CES (2006) Practical issues relat ed to climate change in Florida: A workshop on needs assessment and responses summary report. 12-13 January 2006, Palm Beach Gardens, FL. Available at : http://www.ces.fau.edu/ClimateWorkshop2006/ClimateWorkshop 2006_SummaryReport.pdf (last accessed 9 June 2010). CES (2007) Climate change in Florida summary final report 9-11 May 2007, Tampa, FL. Available at : http://www.ces.fau. edu/ClimateConference2007/presentations/ CCC_Final_report_2007.pdf (last accessed 9 June 2010). Chen, E. & Gerber, J.F. (1990) Climate. Ecosystems of Florida (ed. by R.L. Myers & J.J. Ewel), pp 11-34. University Press of Florida, Gainesville, FL. Chuanyan, Z., Zhongren, N. & Guodong, C. (2005) Methods for modelling of temporal and spatial distribution of air temperature at landscape scale in the southern Qilian mountains, China. Ecological Modelling 189, 209-220. Coenen, D. (2005) Interannual lake temperature trends as indicators for climatic change in Florida MS Thesis, University of Florida, Gainesville, FL. Coenen, D., Porzecanski, I. & Crisman, T.L. (2008). Future directions in colservation and development: incorporating the reality of climate change. Biodiversity 9(3 -4), 106-113. Collins F.C. & Bolstad P.V. (1996). A comparison of spatial interpolation techniques in temperature estima tion. In Proceedings of the Third International Conference/Workshop on Integrating GIS and Environmental Modeling. 21-25 January 1996, Santa Barbara, CA Collins, W.D. Bitz, C.M., Blackmon, M.L., Bonan, G.B., Bretherton, C.S., Carton, J.A., Chang, P., Doney, S.C., Hack, J.A., Henderson, T.B., Kiehl, J.T., Large, W.G., McKenna, D.S., Santer, B.D. & Smith, R.D. (2006) The c ommunity c limate s ystem model version 3 (CCSM 3). Journal of Climate 19, 2122-2143. Crowley, T.J. & North, G.R. (1988) Abrupt c limate cha nge and e xtinction e vents in Earth history. Science 240, 996-1002. Crumpacker, D.W., Box, E.O. & Hardin, D.E. (2001a) Implications of climatic warming for conservation of native trees and shrubs in Florida. Conservation Biology 15, 1008-1020.

PAGE 358

358 Crumpacker, D.W., Box, E.O. & Hardin, D.E. (2001b) Potential breakup of Florida plant communities as a result of climatic warming. Florida Scientist 64, 29-43. Crumpacker, D.W., Box, E.O. & Hardin, D.E. (2002) Use of plant climatic envelopes to design a m onitoring system for early biotic effects of climatic warming. Florida Scientist 65, 159184. Damania, R. & Hatch, J. (2004) Protecting Eden: markets or government. Ecological Economics 53, 339-351. Dickens, G.R., O'Neil, J.R., Rea, D.K. & Owen, R.M. (1995) Dissociation of oceanic m ethane hydrate as a c ause of the c arbon isotope e xcursion at the e nd of the P aleocene. Paleoceanography 10, 965. Easterling, D.R., Horton, B., Jones, P.D., Peterson, T.C., Karl, T.R., Parker, D.E., Salinger, M.J., Razuvaye v, V. Plummer, N., Jamason, P. & Folland, C.K. (1997) Maximum and m inimum t emperature t rends for the globe. Science 277, 364-367. Edwards, P.J. & Abivardi, C. (1998) The value of biodiversity: where ecology and economy blend. Biological Conservation, 83, 239-246. EPA (2008) Florida greenhouse gas emissions and sinks inventory: summary. Available at: http://www.epa.gov/climatechange/emissions/downloads/FLSummary_03_08.pdf (last accessed 11 June 2010). ESRI (2002) ArcGIS 9: using ArcGIS spatial analyst ESRI Publishing, Redlands, CA. ESRI (2003) ArcGIS 9: using ArcGIS geostatistical analyst ESRI Publishing Redlands CA. Fenneman, N.M. (1917) Physiographic subdivision of the United States. Proceedings of the National Academy of Sciences 3, 17-22. Ferraro, P.J. (2001) Global habitat protection: limitations of development interventions and a role for conservation performance payments. Conservation Biology 15, 990-1000. Folke, C. (2006) The economic perspective: conservation against development versu s conservation for development. Conservation Biology 20, 686-688. Frey, D.G. (1982) Contrasting strategies of gamogenesis in northern and southern populations of Cladocera. Ecology 63, 223-241. Friedlingstein, P. (2008) A steep road to climate stabilizat ion Nature 451, 297-298. Gallagher, K.S. (2006) Limits to leapfrogging in energy technologies? Evidence from the Chinese automobile industry. Energy Policy 34, 383-394. Giles, J. (2007) Climate change 2007: how to survive a warming world. Nature 446, 716-717.

PAGE 359

359 Grieg -Gran, M.A., Porras, I. & Wunder, S. (2005) How can market mechanisms for forest environmental services help the poor? Preliminary lessons from Latin America. World Development 33, 1511-1527. Hannah, L., Midgley, G.F., Lovejoy, T., Bond, W.J., Bush, M., Lovett, J.C., Scott, D. & Woodward, F.I. (2002a) Conservation of biodiversity in a changing climate. Conservation Biology 16, 264-268. Hannah, L., Midgley, G.F. & Millar, D. (2002b) Climate change integrated conservation strategies. Global Ecology and Biogeography 11, 485-495. Harris, J.A., Hobbs, R.J., Higgs, E. & Aronson, J. (2006) Ecological r estoration and global c limate c hange. Restoration Ecology 14, 170-176. Harris, M. (2005) Nature makes them lazy: contested perceptions of place a nd knowledge in the lower Amazonian floodplain of Brazil. Conservation and Society 3, 461-478. Hays, J.D., Imbrie, J. & Shackleton, N.J. (1976) Variations in the Earth's Orbit: Pacemaker of the Ice Ages. Science 194, 1121-1132. Henry, J.A., Portier, K.M. & Coyne, J. (1994) The climate and weather of Florida. Pineapple Press Inc., Sarasota FL. Henry, J.A. (1998) Weather and climate. Water resources atlas of Florida (ed. by E.A. Fernald and E.D. Purdum), pp. 16-37. Florida State University Institute of Science and Public Affairs Tallahassee FL. Hoar, T. & Nychka, D. (2008) Statistical downscaling of the c ommunity c limate s ystem model (CCSM) : Monthly t emperature and p recipitation projections. Available at : http://www.gisclimatechange.org/Downscaling.pdf (last accessed 9 June 2010). Hoegh-Guldberg, O., Mumby, P.J., Hooten, A.J., Steneck, R.S., Greenfield, P., Gomez, E., Harvell, C.D., Sale, P.F., Edwards, A.J., Caldeira, K., Knowlton, N., Eakin, C.M., IglesiasPrieto, R., Muthiga, N., Bradbury, R.H., Dubi, A. & Hatziolos, M.E. (2007) Coral r eefs under r apid c limate c hange and ocean a cidification. Science 318, 1737-1742. Holt, F.L. (2005) The catch-22 of conservation: indigenous peoples, biologists, and cultural change. Human Ecology 33, 199-215. Hughes, L. (2000) Biological consequences of global warming: is the signal already apparent? Trends in Ecology and Evolution, 15, 56-61. Hunter, M.L. (2007) Climate c hange and moving s pecies: furthering the d ebate on a ssisted c olonization. Conservation Biology 21, 1356-1358. IPCC (2000) Emissions Scenarios Cambridge University Press Cambridge U.K

PAGE 360

360 IPCC (2005) Carbon dioxide capture and storage. Cambridge University Press, Cambridge, U.K. IPCC (2007a) Climate change 2007 t he physical s cience basis: working group I contribution to the fourth assessment report of the IPCC Cambridge University Press, Cambridge U.K. IPCC (2007b) Climate change 2007 impacts, adaptation and v ulnerability: working group II contribution to the fourth assessment report of the IPCC Cambridge University Press, Cambridge U.K. Katz, M.E., Cramer, B.S., Mountain, G S., Katz, S. & Miller, K. G. (2001) Uncorking the bottle: w hat t riggered the Paleocene / Eocene t hermal m aximum m ethane r elease? Paleoceanography 16, 549. Kleiven, H.F., Kissel, C., Laj, C., Ninnemann, U.S., Richter, T.O. & Cortijo, E. (2008) Reduced n orth Atlantic deep w ater c oeval with the glacial Lake Agassiz f reshwater o utburst. Science 319, 60-64. Koenig, R. (2008) Critical time for African rainforests. Science 320, 1439-1441. Lenton, T.M., Held, H., Kriegler, E., Hall, J.W., Lucht, W., Rahmstorf, S. & Schellnhuber, H.J. (2008) Tipping elements in the Earth's climate system. Proceedings of the National Academy of Sciences 105, 1786-1793. Lindsay, R. W. & Zhang, J. (2005) The thinning of Arctic s ea ice, 1988: h ave w e passed a tipping point? Journal of Climate 18, 4879-4894. Lovejoy, T.E. & Hannah, L.J. (2005) Climate Change and Biodiversity Yale University Press New Haven CT. Marshall, C.H., Pielke, R.A., Steyaert, L.T. & Willard, D A. (2004) The i mpact of anthropogenic l and -c over c hange on the Florida peninsula s ea b reezes and w arm s eason s ensible we ather. Monthly Weather Review 132, 28-52. Maslin, M., Owen, M., Day, S. & Long, D. (2004) Linking continental slope failures and climate change: t esting the clathrate gun hypothesis. Geology 32, 53-56. McCarty, J. P. (2001) Ecological consequences of recent climate change Conservation Biology 15, 320-331. Mc Lachlan, J.S., Hellman, J.J. & Schwartz, M.W. (2007) A f ramework for debate of a ssisted m igration in an e ra of c limate c hange. Conservation Biology 21, 297-302. Miles L. & Kapos, V. (2008). Reducing greenhouse gas emissions from deforestation and forest degradation: global land-u se implications. Science 320, 1454-1455. Moral, F.J. (2010) Comparison of different geostatistical approaches to map climate variables: app lication to precipitation. International Journal of Climatology 30, 620-631.

PAGE 361

361 Mulholland, P.J., Best, G.R., Coutant, C.C., Hornberger, G.M., Meyer, J.L., Robinson, P.J., Stenberg, G.M., Turner, R.E., VeraHerrera, F. & Wetzel, R.G. (1997). Effects of climate change in freshwater ecosystems of the south eastern United States and the gulf coast of Mexico. Hydrological Processes 11, 949-970. Mulkey, S.S. (2007) Climate change and land use in Florida: interdependencies and opportunities Report prepared for the Century Commission for a Sustainable Florida. Available at : http://snre.ufl.edu/home/files/Climate%20change% 20and%20land%20use %20in%20Florida%20V8-1s.pdf (last accessed 9 June 2010). Nalder, I.A. & Wein, R.W. (1998) Spatial interpolation of climatic n ormals: test of a new method in the Canadian boreal forest. Agricultural and Forest Meteorology 92, 211-225. Na ture Editors (2007) Climate change, here and now. Nature 446, 701. NCAR GIS Initiative (20 10a) GIS climate change scenarios Available at : http://www. gisclimatechange.org (last accessed 9 June 2010). NCAR GIS Initiative (2010b) Control runs Available at : http://www.gisclimatechange. org/runSetsHelp.html (last accessed 9 June 2010). NCDC (2010a) The annual global (land and ocean combined) anomalie s (degrees C) Available at : ftp://ftp.ncdc.noaa.gov/pub/data/anomalies/annual.land_ocean. 90S.90N.df_19012000mean.dat (last accessed 9 June 2010). NCDC (2010b) State of the c limate global analysis May 2010. Available at : http://www.ncdc. noaa.gov/sotc/?report=global&year=2010&month=5&submitted=Get+Report (last accessed 15 June 2010). Obeysekera, J., Browder, J., Hornung, L. & Harwell, M. (1999) The natural South Florida system I: climate, geology, and hydrology. Urban Ecosystems 3, 223-244. Parmesan, C. & Yohe, G. (2003) A globally coherent fingerprint of climate change impacts across natural systems. Nature 421, 37-42. Parmesan, C. & Galbraith, H. (2004) Observed impacts of global climate change in the U.S. Report prepared for the Pew Center on Global Climate Change. Available at: http://www.pewclimate.org/docUploads/final%5FObsImpact.pdf (last accessed 10 June 2010). Parmesan, C. (2006) Ecological and evolutionary responses to recent climate change. Annual Review of Ecology, Evolution, and Systematics 37, 637-669. Pearson, P. N. & Palmer, M.R. (2000) Atmospheric carbon dioxide concentrations over the past 60 million years. Nature 406, 695-699.

PAGE 362

362 Pearson, R.G. & Dawson, T.P. (2003) Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global Ecology and Biogeography 12, 361-371. Petit, R.J., Hu, F.S. & Dick, C.W. (2008) Forests of the past: a window to future changes. Science 320, 1450-1452. Pielke Sr., R.A., Marland, G., Betts, R.A., Chase, T.N., Eastman, J.L., Niles, J.O., Niyogi, D.D.S & Running, S.W. (2002) The influence of landuse change and landscape dynamics on the climate system: relevance to climate change policy beyond the radiative effect of greenhouse gases Philosophical Transactions of the Royal Society London A 360, 17051719. Pope, K.O., D'Hondt, S.L. & Marshall, C.R. (1998) Meteorite impact and the mass extinction of species at the Cretaceous/Tertiary boundary. Proceedings of the National Academy of Science s 95, 11028-11029. Pounds, J.A., Bustamante, M.R., Coloma, L.A., Consuegra, J.A., Fogden, M.P.L., Foster, P.N., La Marca, E., Masters, K.L., MerinoViteri, A ., Puschendorf, R., Ron, S.R., SnchezAzofeifa, G.A., Still, C.J. & Young, B.E. (2006) Wi despread amphibian extinctions from epidemic disease driven by global warming. Nature 439, 161-167. Rahmstorf, S. (2007) A semi-empirical approach to projecting future sealevel rise. Science 315, 368-370. Raupach, M.R., Marland, G., Ciais, P., Le Quere, C., Canadell, J.G., Klepper, G. & Field, C.B. (2007) Global and regional drivers of accelerating CO2 emissions. Proceedings of the National Academy of Sciences 104, 10288-10293. Robock, A. & Mao, J. (1995) The v olcanic s ignal in s urface t emperature obse rvations. Journal of Climate 8, 1086-1103. Roe, G.H. & Baker, M.B. (2007) Why is climate sensitivity so unpredictable? Science 318, 629632. Rosenthal, W.D., Hammer, G.L. & Butler. D. (1998) Predicting regional grain sorghum production in Australia using spatial data and crop simulation modelling. Agricultural and Forest Meteorology 91, 263-274. Ruddiman W.F. (2003) The anthropogenic g reenhouse e ra began t housands of y ears ago. Climatic Change 61, 261-293. Ruddiman, W.F. (2005) The e arly anthropogenic hypothesis a year l ater. Climatic Change 69, 427-434. Sagan, C & Mullen, G. (1972) Earth and Mars: evolution of atmospheres and surface temperatures. Science 177, 52-56.

PAGE 363

363 Saunders, M.A. & Lea, A.S. (2008). Large contribution of sea surface warming to recent increase in Atlantic hurricane activity. Nature 451, 557-560. Schellnhuber, H.J., Cramer, W., Nakicenovic, N., Wigley, T., Yohe, G. & Pachauri, R. (2006) Avoiding dangerous c limate change Cambridge University Press Cambridge U.K. Siegenthaler, U., Stocker, T.F., Monnin, E., Luthi, D., Schwander, J., Stauffer, B., Raynaud, D., Barnola, J. -M., Fischer, H. & Masson Delmotte, V. (2005) Stable c arbon c ycle -c limate r ela tionship d uring the l ate Pleistocene. Science 310, 1313-1317. SE RCC (2010) Historical c limate summaries Available at : http://www.sercc.com/climateinfo/ historical/historical.html (last accessed 9 June 2010) Stanton, E. & Ackerman, F. (2007) Florida and climate change: the costs of inaction. Report prepared for the Environmental Defense Fund. Available at : http://w ww.ase.tufts.edu/ gdae/Pubs/rp/Florida_hr.pdf (last accessed 9 June 2010). Stern, N. (2007) The economics of c limate change: the Stern r eview Cambridge University Press Cambridge, U.K. Stokstad, E. (2008) A second chance for rainforest biodiversity. Scie nce, 320, 1436-1438. Stott, P.A., Tett, S.F.B., Jones, G.S., Allen, M.R., Ingram, W.J. & Mitchell, J.F.B. (2001) Attribution of twentieth century temperature change to natural and anthropogenic causes. Climate Dynamics 17, 1-21. Tans, P. (2010) NOAA Earth System Research Laboratory trends in atmospheric carbon dioxide Available at : http://www.esrl.noaa.gov/gmd/ccgg/trends/ (last accessed 9 June 2010). Tchoukanski, I. (2007) ET Geowizards Ve rsion 9.6.1. Available at : http://www.ian -ko.com/ (last accessed 9 June 2010). Thomas, C.D, Bodsworth, E.J., Wilson, R.J., Simmons, A.D., Davies, Z.G., Musche, M. & Conradt, L. (2001) Ecological and evolutionary processes at expanding range margins. Nature 411, 577-581. Thornton, M. & Wilhelmi, O. (2008) Temperature anomaly GIS demonstration project Available at : http://www.gisclimatechange.org/Temperature%20anomaly%20GIS %20tutorial.zip (last accessed 9 June 2010). Turner, R.K., Paavola, J., Cooper, P., Farber, S., Jessamy, V. & Georgiou, S. (2003) Valuing nature: lessons learn ed and future research directions. Ecological Economics 46, 493-510. UNFCCC (1997) Kyoto Protocol to the United Nations Framework Convention on Climate Change Available at : http://unfccc.int/resource/docs/convkp/kpeng.html (last accessed 9 June 2010).

PAGE 364

364 UNFCCC (2007a) Compilation and synthesis of fourth national communications Available at : http://unfccc.int/resource/docs/2007/sbi/eng/l26a01.pdf (last accessed 9 June 2010). UNFCCC (2007b) GHG data from UNFCCC. Available at: http://unfccc.int/ghg_data/ ghg_data_unfccc/items/4146.php (last accessed 10 June 2010) Ustrnul, Z. & Czekierda, D. (2005). Application of GIS for the development of climatological air temperature maps: an example from Poland. Journal of Applied Meteorology 12, 43-50. Vecchi, G.A. & Soden, B.J. (2007) Increased tropical Atlantic wind shear in model projections of global warming. Geophysical Research Letters, 34, L08702 (5 pp). Walther, G.R., Post, E., Convey, P., Menzel, A., Parmesan, C., Beebee, T.J.C., Fromentin, J.M., Hoegh-Guldenberg, O. & Bairlein, F. (2002) Ecological responses to recent climate change. Nature 416, 389-395. Watts, W.A. (1980) The late Quaternary vegetation history of the southeastern United States. Annual Review of Ecology and Systematics 11, 387-409. Watts, W.A. & Stuiver, M. (1980) Late Wisconsin climate of northern Florida and the origin of species rich deciduous forest. Science 210, 325-327. Watts, W.A., Hansen, B.C.S. & Grimm, E.C. (1992) Camel Lake: A 40000yr record of vegetational and forest history from No rthwest Florida. Ecology 73, 1056-1066. Western, D. (2006) Human-modified ecosystems and future evolution. Proceedings of the National Academy of Sciences 98, 5458-5465. Williams, J.W. & Jackson, S.T. (2007) Novel climates, no -analog communities, and eco logical surprises. Frontiers in Ecology and the Environment 5, 475-482. Wilson, M.A. & Carpenter, S.R. (1999) Economic valuation of freshwater ecosystems in the United States: 1971-1997. Ecological Applications 9, 772-783. Winsberg, M. (1990) Florida weather University of Central Fl orida Press Orlando, FL. WMO (1967) A note on climatological normals Report of a working group of the commission for Climatology. World Meteorological Organization Technical Note No. 84, 19pp. Zachos, J., Pagani, M, Sloan, L., Thomas, E. & Billups, K. (2001) Trends, rhythms, and aberrations in global climate 65 ma to present. Science 292, 686-693. Zachos, J.C., Rhl, U., Schellenberg, S.A., Sluijs, A., Hodell, D.A., Kelly, D.C., Thomas, E ., Nicolo, M ., Raffi, I ., Lourens, L.J ., McCarren, H & Kroon, D (2005) Rapid a cidification of the ocean during the PaleoceneEocene t hermal m aximum. Science 308, 1611-1615. Zahabu, E., Skutsch, M.M., Sosovele, H. & Malimbwi, R.E. (2007). Reduced emissions from deforestation and degradation. African Journal of Ecology 45, 451-453.

PAGE 365

365 Zielinski, G.A., Mayewski, P.A., Meeker, L.D., Whitlow, S. & Twickler, M.S. (1996) A 110,000-y r r ecord of e xplosive v olcanism from the GISP2 (Greenland) i ce core. Quaternary Research 45, 109-118.

PAGE 366

366 BIOGRAPHICAL SKETCH Danny Coenen was born in Jlich, Germany in 1978. He resettled to the United States in 1997, to marry Jennifer L. Foster in 1998. He enrolled at Missouri State University in Springfield, Missouri, from 1998 until 2000, at which time he transferred to the School of Natural Resources and Environment at the University of Florida in Gainesville. After earning a Bachelor of Science degree in environmental science cum laude in May 2003, he remained at the School of Natural Resources and Environment to begin graduate studies in August 2003, earning the Mas ter of Science in interdisciplinary ecology degree in August 2005. Danny Coenen enter ed the interdisciplinary ecology Ph. D. program of the School of Natural Resources and Environment as an Alumni Fellow with the support of the Department of Environmental Engineering Sciences in August 2005. During his time as a graduate student, he was fortunate to receive a course development grant from the Center for European Studies at the University of Florida, allowing him to develop and teach the first dedicated climate change science and policy course at the University of Florida. Following graduation, Danny Coenen will join the faculty of the School of Natural Resources and Environment as a lecturer and begin a post-doctoral fellowship to continue developing the mod els introduced in this dissertation at the University of South Florida in Tampa.