Application of remotely sensed data to a geographic information system for microclimate change analysis

MISSING IMAGE

Material Information

Title:
Application of remotely sensed data to a geographic information system for microclimate change analysis
Physical Description:
xxvii, 515 leaves : ill. ; 29 cm.
Language:
English
Creator:
Jordan, Jonathan David, 1962-
Publication Date:

Subjects

Subjects / Keywords:
Agricultural Engineering thesis, Ph. D
Dissertations, Academic -- Agricultural Engineering -- UF
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1994.
Bibliography:
Includes bibliographical references (leaves 492-514).
Statement of Responsibility:
Jonathan David Jordan.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002042877
oclc - 33315000
notis - AKN0755
System ID:
AA00004722:00001

Full Text











APPLICATION OF REMOTELY SENSED DATA TO A GEOGRAPHIC
INFORMATION SYSTEM FOR MICROCLIMATE CHANGE ANALYSIS


















By


JONATHAN DAVID JORDAN


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


1994



























To my parents,

Don and Dorothy Jordan














ACKNOWLEDGMENTS


This research was performed with the help of the staff

and facilities of the Remote Sensing Application Laboratory

(RSAL) located at the Agricultural Engineering Department of

the University of Florida. The author is grateful for the

assistance provided by the RSAL director, Dr. Sun F. Shih;

RSAL manager, Orlando Lanni; departmental computer technician

Curtis Weldon; and RSAL assistants Chih-Hung Tan, Yu Rong Tan,

and Bruce E. Myhre.

Acknowledgments are also due to additional persons and

agencies for key assistance in various portions of this study.

Technical information concerning the Advanced Very High

Resolution Radiometer (AVHRR) satellite imagery used in this

research was provided by Mary Hughes, Emily Harrod, Dr. Andrew

Horvitz, Dr. Katherine Kidwell, Dr. Carolyn Ng, Richard

DeRycke, and others of the National Oceanic and Atmospheric

Administration (NOAA). Both technical information and data

tapes concerning the Heat Capacity Mapping Mission (HCMM)

imagery used in this research were provided by Dr. William L.

Barnes, Barbara Pope, Locke M. Stuart, and others of the

National Space Science Data Center (NSSDC).

Essential water-body surface temperature data were

provided by Dr. Leslie Wedderburn, Brian Turkotte, Ernest

Gallego, and others of the South Florida Water Management

iii









District (SFWMD); William L. Osburn, Gail Gallagher, and

others of the St. Johns River Water Management District

(SJRWMD); David Hornsby and others of the Suwannee River Water

Management District (SRWMD); Kenneth Romie, Mark Rials, and

others of the Southwest Florida Water Management District

(SWFWMD); Thomas Cardenel, R. Malloy, and others of the

Environmental Protection Commission of Hillsborough County

(EPCHC); Donald D. Moores of the Pinellas County Department of

Environmental Management (PCDEM); and Dr. David Gowan of the

Florida Department of Environmental Protection (DEP). Both

water-body surface temperature data and statewide land-cover

maps were supplied by John Steyes and others of the Florida

Game and Freshwater Fish Commission (FGFFC). Land-cover

information and maps of the lower Lake Wales Ridge area were

provided by the Archbold Biological Station.

Aerial photographs used in this research were made

available by Dr. Helen J. Armstrong of the University of

Florida Map Library. Statewide aquaculture information and

maps were provided by Dr. Edward P. Lincoln and Dr. C. Direlle

Baird of the University of Florida Agricultural Engineering

Department. Phosphate mine and mine reclamation information

was made available by Dr. Lawrance N. Shaw of the University

of Florida Agricultural Engineering Department. Crop

information and assistance with site visits to St. Johns River

agricultural areas were provided by Dr. Dale R. Hensel of the

Hastings Agricultural Research and Education Center (AREC).

Crop information, maps, and assistance with site visits to the









Everglades Agricultural Area (EAA) were provided by Dr. George

H. Snyder of the Everglades Research and Education Center

(EREC). Assistance with site visits to south Florida citrus

orchards, pastures, and Lake Okeechobee water-temperature

stations was given by Michael Piper, David Soballe, and others

of the SFWMD; assistance with site visits to the Lake Apopka

water-temperature station and marsh restoration project was

given by J. Palenkas and others of the SJRWMD. Land-cover

information and assistance with wetland site visits in north

and panhandle Florida were provided by Jay L. Johnson of the

NWFWMD. An infrared radiometer was made available by Dr.

Donald J. Pitts of the Immokalee AREC. Use of soil sampling

and analysis equipment was provided by Dr. Donald L. Myhre and

Joseph Nguyen of the University of Florida Soil and Water

Science Department. Thanks are also given to professors Sun

F. Shih, Jerome J. Gaffney, Dorota Z. Haman, Edward P.

Lincoln, Byron E. Ruth, and George H. Snyder of my supervisory

committee for their help and support.















TABLE OF CONTENTS


page

ACKNOWLEDGMENTS......................................... iii

LIST OF TABLES........................................ xiii

LIST OF FIGURES ................................................... xxiv

ABSTRACT ............................................... xxvi

INTRODUCTION............................................ 1

Purpose and Objectives................................. 1

Importance of Surface Temperature in Climatology....... 2

Importance of Surface Temperature in Hydrology......... 4

Importance of Surface Temperature in Agriculture and
Forestry............................................. 5

Factors Affecting Surface Temperature Patterns........ 6

Temperature Impacts of Changes in Land-Cover........ 9

Temporal Inter-Relation of Forcing Factors........... 13

Spatial Inter-Relation of Forcing Factors............ 19

Potential for Future Changes in Soil Type........... 21

REVIEW OF LITERATURE..................................... 22

Difficulties of Surface Temperature Measurement....... 22

Atmospheric Correction Techniques................... 24

Emissivity Correction Techniques.................... 26

Previous Studies..................................... 30

MATERIALS AND METHODS.................................... 39

Study Area......................... ................... 39

Panhandle Zone.......... ............................ 39









North Zone .................................... ..... 41

South Zone..................................... ....... 41

Geographic Information System........................ 42

Raster Datasets........... ........... ............. 42

Vector Datasets.................................... 42

Geographic Referencing............................... 43

GIS Analyses in Raster Environment................. 44

AVHRR Image Processing................................ 45

AVHRR Data Types.................................. 46

Calibration to At-Satellite Radiant Temperature...... 49

Geographic Correction and Registration............... 51

Conversion from Radiant to Kinetic Temperature...... 61

Accuracy Assessment of Kinetic Temperature Images... 68

Water and Cloud Masking............................ 70

Final Forms of Images in the GIS Database........... 74

HCMM Historical-Image Processing...................... 74

Geographic Correction of HCMM Images................ 77

Calibration of HCMM At-Satellite Radiant
Temperature ....................................... 78

Ground-Based DSTV/Soil-Moisture Work.................. 79

Mineral Soil Investigation.......................... 79

Organic Soil Investigation.......................... 83

Vegetated Soil Investigation........................ 85

Soil Type Data........................................ 86

Mineral Soils. ...................................... 87

Organic Soils.................................... 89

Artificial Soil Type Change........................ 90


vii









Land-Cover Data...................................... 91

Natural Land-Cover............................... 91

Agricultural Land-Cover............................. 115

Urban/Industrial Land-Cover......................... 123

Special Land-Cover Conditions....................... 129

RESULTS AND DISCUSSION................................. 136

Analyses Across Macroclimate Zones.................... 136

Analyses Within Macroclimate Zones.................... 147

Spring Afternoon Natural Land-Cover Thermal
Patterns............................................ 147

Spring Afternoon Agricultural Land-Cover Thermal
Patterns........................................... 154

Spring Afternoon Urban/Industrial Land-Cover Thermal
Patterns........................................... 163

Spring Afternoon Change Analyses--Natural to
Agricultural Land-Cover........................... 172

Spring Afternoon Change Analyses--Natural to Urban/
Industrial Land-Cover............................. 178

Spring Afternoon Change Analyses--Agricultural to
Urban/Industrial Land-Cover....................... 185

Spring Afternoon Comparison of Agricultural to
Natural Heat Islands............................. 193

Spring Afternoon Comparison of Urban/Industrial to
Natural Heat Islands............................... 196

Spring Afternoon Change Analyses--Special Factors... 199

Spring Nighttime Natural Land-Cover Thermal
Patterns........................................... 203

Spring Nighttime Agricultural Land-Cover Thermal
Patterns.......................................... 208

Spring Nighttime Urban/Industrial Land-Cover Thermal
Patterns.......... ...................... ...... 218

Spring Nighttime Change Analyses--Natural to
Agricultural Land-Cover........................... 227


viii









Spring Nighttime Change Analyses--Natural to Urban/
Industrial Land-Cover............................. 232

Spring Nighttime Change Analyses--Agricultural to
Urban/Industrial Land-Cover....................... 238

Spring Nighttime Comparison of Agricultural to
Natural Cold Islands.............................. 242

Spring Nighttime Comparison of Urban/Industrial to
Natural Cold Islands.............................. 245

Spring Nighttime Change Analyses--Special Factors... 246

Spring Diurnal Natural Land-Cover Thermal Patterns.. 252

Spring Diurnal Agricultural Land-Cover Thermal
Patterns........................................... 259

Spring Diurnal Urban/Industrial Land-Cover Thermal
Patterns............................................ 267

Spring Diurnal Change Analyses--Natural to
Agricultural Land-Cover........................... 276

Spring Diurnal Change Analyses--Natural to Urban/
Industrial Land-Cover............................. 281

Spring Diurnal Change Analyses--Agricultural to
Urban/Industrial Land-Cover....................... 286

Spring Diurnal Comparison of Agricultural to
Natural Extreme Islands........................... 294

Spring Diurnal Comparison of Urban/Industrial to
Natural Extreme Islands........................... 296

Spring Diurnal Change Analyses--Special Factors..... 299

Winter Afternoon Natural Land-Cover Thermal
Patterns.......................................... 303

Winter Afternoon Agricultural Land-Cover Thermal
Patterns.......................................... 310

Winter Afternoon Urban/Industrial Land-Cover Thermal
Patterns........................................... 317

Winter Afternoon Change Analyses--Natural to
Agricultural Land-Cover........................... 326

Winter Afternoon Change Analyses--Natural to Urban/
Industrial Land-Cover.............................. 330









Winter Afternoon Change Analyses--Agricultural to
Urban/Industrial Land-Cover....................... 330

Winter Afternoon Comparison of Agricultural to
Natural Heat Islands.............................. 335

Winter Afternoon Comparison of Urban/Industrial to
Natural Heat Islands.............................. 340

Winter Afternoon Change Analyses--Special Factors... 344

Winter Nighttime Natural Land-Cover Thermal
Patterns.................... ........... ........ 348

Winter Nighttime Agricultural Land-Cover Thermal
Patterns........................................ 353

Winter Nighttime Urban/Industrial Land-Cover Thermal
Patterns........................................... 359

Winter Nighttime Change Analyses--Natural to
Agricultural Land-Cover........................... 366

Winter Nighttime Change Analyses--Natural to Urban/
Industrial Land-Cover............................. 370

Winter Nighttime Change Analyses--Agricultural to
Urban/Industrial Land-Cover....................... 376

Winter Nighttime Comparison of Agricultural to
Natural Cold Islands.............................. 382

Winter Nighttime Comparison of Urban/Industrial to
Natural Cold Islands.............................. 384

Winter Nighttime Change Analyses--Special Factors... 387

Winter Diurnal Natural Land-Cover Thermal Patterns.. 390

Winter Diurnal Agricultural Land-Cover Thermal
Patterns........................................... 395

Winter Diurnal Urban/Industrial Land-Cover Thermal
Patterns ............ .............................. 405

Winter Diurnal Change Analyses--Natural to
Agricultural Land-Cover........................... 414

Winter Diurnal Change Analyses--Natural to Urban/
Industrial Land-Cover............................. 414

Winter Diurnal Change Analyses--Agricultural to
Urban/Industrial Land-Cover....................... 419









Winter Diurnal Comparison of Agricultural to
Natural Extreme Islands........................... 423

Winter Diurnal Comparison of Urban/Industrial to
Natural Extreme Islands .......................... 429

Winter Diurnal Change Analyses--Special Factors..... 432

Analyses of Micro-Scale Maritime Effects.............. 436

Hammock Comparisons... .............................. 436

Marsh Comparisons.................................. 437

Maritime Micro-Scale Thermal Moderation............. 438

Analyses of Seasonal Effects on Deciduous Vegetation.. 438

Historical HCMM-Based Analyses........................ 439

HCMM Analyses Across Macroclimate Zones............. 439

HCMM Historical Special Condition Change Analyses... 443

Results of Ground-Based DSTV/Soil-Moisture............ 450

Mineral Soil Results................................ 451

Organic Soil Results................................ 453

Vegetated Soil Results.............................. 453

SUMMARY AND CONCLUSIONS................................ 455

Principal Findings................................... 456

Importance of Soil Type and Land-Cover.............. 456

Differences Among Natural Land-Cover Types.......... 457

Differences Among Agricultural Land-Cover Types..... 457

Differences Among Urban/Industrial Land-Cover
Types ............................................. 458

Potential for Soil Moisture Monitoring.............. 458

Recommendations for Future Research................... 459

Ground-Based Data Collection Improvement............ 459

Satellite System Improvement........................ 460










Direction of Future Research........................ 462

GLOSSARY............................................... .. 464

APPENDIX A IMAGE DOCUMENTATION ......................... 468

APPENDIX B WATER-BODY TEMPERATURE MEASUREMENT STATIONS. 475

APPENDIX C LAND-COVER POLYGON DETAILS.................. 476

REFERENCES...................... ....................... 492

BIOGRAPHICAL SKETCH...... ............................... 515


xii














LIST OF TABLES


page

1 Spring afternoon surface temperature
across-zone differences among natural
land-cover types ........................... 137

2 Spring nighttime surface temperature
across-zone differences among natural
land-cover types .......................... 139

3 Spring diurnal surface temperature variation
across-zone differences among natural
land-cover types ........................... 141

4 Winter afternoon surface temperature
across-zone differences among natural
land-cover types ........................... 143

5 Winter nighttime surface temperature
across-zone differences among natural
land-cover types ........................... 144

6 Winter diurnal surface temperature variation
across-zone differences among natural
land-cover types .......................... 145

7 Spring afternoon surface temperature
differences among natural land-cover
types in panhandle zone..................... 148

8 Spring afternoon surface temperature
differences among natural land-cover
types in north zone......................... 150

9 Spring afternoon surface temperature
differences among natural land-cover
types in south zone......................... 152

10 Spring afternoon surface temperature
differences among agricultural land-cover
types in panhandle zone..................... 155

11 Spring afternoon surface temperature
differences among agricultural land-cover
types in north zone......................... 157


xiii









12 Spring afternoon surface temperature
differences among agricultural land-cover
types in south zone........................ 160

13 Spring afternoon surface temperature
differences among urban/industrial land-
cover types in panhandle zone............... 164

14 Spring afternoon surface temperature
differences among urban/industrial land-
cover types in north zone................... 166

15 Spring afternoon surface temperature
differences among urban/industrial land-
cover types in south zone................... 170

16 Spring afternoon surface temperature change
from natural to agricultural in panhandle
zone ........................................ 173

17 Spring afternoon surface temperature change
from natural to agricultural in north
zone......................................... 175

18 Spring afternoon surface temperature change
from natural to agricultural in south
zone......................................... 177

19 Spring afternoon surface temperature change
from natural to urban/industrial in
panhandle zone.............................. 179

20 Spring afternoon surface temperature change
from natural to urban/industrial in
north zone.................................. 181

21 Spring afternoon surface temperature change
from natural to urban/industrial in
south zone .................................. 184

22 Spring afternoon surface temperature change
from agricultural to urban/industrial in
panhandle zone.............................. 186

23 Spring afternoon surface temperature change
from agricultural to urban/industrial in
north zone.................................. 188

24 Spring afternoon surface temperature change
from agricultural to urban/industrial in
south zone .................................. 191


xiv









25 Spring afternoon surface temperature of
agricultural land-cover types vs hottest
natural land-cover......................... 194

26 Spring afternoon surface temperature of
urban/industrial land-cover types vs
hottest natural land-cover.................. 197

27 Spring afternoon surface temperature change
for special conditions..................... 200

28 Spring nighttime surface temperature
differences among natural land-cover
types in panhandle zone...................... 204

29 Spring nighttime surface temperature
differences among natural land-cover
types in north zone.......................... 206

30 Spring nighttime surface temperature
differences among natural land-cover
types in south zone......................... 209

31 Spring nighttime surface temperature
differences among agricultural land-cover
types in panhandle zone..................... 211

32 Spring nighttime surface temperature
differences among agricultural land-cover
types in north zone......................... 213

33 Spring nighttime surface temperature
differences among agricultural land-cover
types in south zone......................... 215

34 Spring nighttime surface temperature
differences among urban/industrial land-
cover types in panhandle zone............... 219

35 Spring nighttime surface temperature
differences among urban/industrial land-
cover types in north zone................... 221

36 Spring nighttime surface temperature
differences among urban/industrial land-
cover types in south zone................... 225

37 Spring nighttime surface temperature change
from natural to agricultural in panhandle
zone........................................ 228









38 Spring nighttime surface temperature change
from natural to agricultural in north
zone...................... ................. 229

39 Spring nighttime surface temperature change
from natural to agricultural in south
zone..... ......... .................. ...... 231

40 Spring nighttime surface temperature change
from natural to urban/industrial in
panhandle zone .............................. 233

41 Spring nighttime surface temperature change
from natural to urban/industrial in
north zone ........................... ...... 234

42 Spring nighttime surface temperature change
from natural to urban/industrial in
south zone.................................. 237

43 Spring nighttime surface temperature change
from agricultural to urban/industrial in
panhandle zone.............................. 239

44 Spring nighttime surface temperature change
from agricultural to urban/industrial in
north zone......... ........................ 240

45 Spring nighttime surface temperature change
from agricultural to urban/industrial in
south zone. ................................... 243

46 Spring nighttime surface temperature of
agricultural land-cover types vs coldest
natural land-cover.......................... 244

47 Spring nighttime surface temperature of
urban/industrial land-cover types vs
coldest natural land-cover................. 247

48 Spring nighttime surface temperature change
for special conditions...................... 249

49 Spring diurnal surface temperature variation
differences among natural land-cover
types in panhandle zone..................... 253

50 Spring diurnal surface temperature variation
differences among natural land-cover
types in north zone........................ 255


xvi









51 Spring diurnal surface temperature variation
differences among natural land-cover
types in south zone......................... 257

52 Spring diurnal surface temperature variation
differences among agricultural land-cover
types in panhandle zone..................... 260

53 Spring diurnal surface temperature variation
differences among agricultural land-cover
types in north zone......................... 262

54 Spring diurnal surface temperature variation
differences among agricultural land-cover
types in south zone......................... 264

55 Spring diurnal surface temperature variation
differences among urban/industrial land-
cover types in panhandle zone............... 268

56 Spring diurnal surface temperature variation
differences among urban/industrial land-
cover types in north zone................... 270

57 Spring diurnal surface temperature variation
differences among urban/industrial land-
cover types in south zone.................... 274

58 Spring diurnal surface temperature variation
change from natural to agricultural in
panhandle zone.............................. 277

59 Spring diurnal surface temperature variation
change from natural to agricultural in north
zone......................................... 279

60 Spring diurnal surface temperature variation
change from natural to agricultural in south
zone ....................................... 280

61 Spring diurnal surface temperature variation
change from natural to urban/industrial in
panhandle zone.............................. 282

62 Spring diurnal surface temperature variation
change from natural to urban/industrial in
north zone................................. 284

63 Spring diurnal surface temperature variation
change from natural to urban/industrial in
south zone................ .... ........ ...... 287


xvii









64 Spring diurnal surface temperature variation
change from agricultural to urban/industrial
in panhandle zone........................... 288

65 Spring diurnal surface temperature variation
change from agricultural to urban/industrial
in north zone................................ 290

66 Spring diurnal surface temperature variation
change from agricultural to urban/industrial
in south zone............................... 293

67 Spring diurnal surface temperature variation
of agricultural land-cover types vs highest-
DSTV natural land-cover..................... 295

68 Spring diurnal surface temperature variation
of urban/industrial land-cover types vs
highest-DSTV natural land-cover.............. 297

69 Spring diurnal surface temperature variation
change for special conditions.............. 300

70 Winter afternoon surface temperature
differences among natural land-cover
types in panhandle zone..................... 304

71 Winter afternoon surface temperature
differences among natural land-cover
types in north zone......................... 306

72 Winter afternoon surface temperature
differences among natural land-cover
types in south zone......................... 308

73 Winter afternoon surface temperature
differences among agricultural land-cover
types in panhandle zone...................... 311

74 Winter afternoon surface temperature
differences among agricultural land-cover
types in north zone......................... 313

75 Winter afternoon surface temperature
differences among agricultural land-cover
types in south zone......................... 315

76 Winter afternoon surface temperature
differences among urban/industrial land-
cover types in panhandle zone............... 319


xviii









77 Winter afternoon surface temperature
differences among urban/industrial land-
cover types in north zone................... 320

78 Winter afternoon surface temperature
differences among urban/industrial land-
cover types in south zone.................. 324

79 Winter afternoon surface temperature change
from natural to agricultural in panhandle
zone............................ ....... .. 327

80 Winter afternoon surface temperature change
from natural to agricultural in north
zone.......................... ............... 328

81 Winter afternoon surface temperature change
from natural to agricultural in south
zone..... ................................... 329

82 Winter afternoon surface temperature change
from natural to urban/industrial in
panhandle zone............................... 331

83 Winter afternoon surface temperature change
from natural to urban/industrial in
north zone .................................. 332

84 Winter afternoon surface temperature change
from natural to urban/industrial in
south zone........................... ...... ..... 334

85 Winter afternoon surface temperature change
from agricultural to urban/industrial in
panhandle zone...... ....................... 336

86 Winter afternoon surface temperature change
from agricultural to urban/industrial in
north zone....... ...... ....... .. ............ 337

87 Winter afternoon surface temperature change
from agricultural to urban/industrial in
south zone........... ..................... ... 339

88 Winter afternoon surface temperature of
agricultural land-cover types vs hottest
natural land-cover.......................... 341

89 Winter afternoon surface temperature of
urban/industrial land-cover types vs
hottest natural land-cover.................. 342


xix









90 Winter afternoon surface temperature change
for special conditions...................... 345

91 Winter nighttime surface temperature
differences among natural land-cover
types in panhandle zone..................... 349

92 Winter nighttime surface temperature
differences among natural land-cover
types in north zone.......................... 351

93 Winter nighttime surface temperature
differences among natural land-cover
types in south zone......................... 354

94 Winter nighttime surface temperature
differences among agricultural land-cover
types in panhandle zone..................... 357

95 Winter nighttime surface temperature
differences among agricultural land-cover
types in north zone......................... 358

96 Winter nighttime surface temperature
differences among agricultural land-cover
types in south zone......................... 360

97 Winter nighttime surface temperature
differences among urban/industrial land-
cover types in panhandle zone............... 363

98 Winter nighttime surface temperature
differences among urban/industrial land-
cover types in north zone.................... 364

99 Winter nighttime surface temperature
differences among urban/industrial land-
cover types in south zone.................... 367

100 Winter nighttime surface temperature change
from natural to agricultural in panhandle
zone......................................... 369

101 Winter nighttime surface temperature change
from natural to agricultural in north
zone.................. ............. ..... 371

102 Winter nighttime surface temperature change
from natural to agricultural in south
zone........................................ 372









103 Winter nighttime surface temperature change
from natural to urban/industrial in
panhandle zone.............................. 373

104 Winter nighttime surface temperature change
from natural to urban/industrial in
north zone.................................. 374

105 Winter nighttime surface temperature change
from natural to urban/industrial in
south zone .................................. 377

106 Winter nighttime surface temperature change
from agricultural to urban/industrial in
panhandle zone.............................. 378

107 Winter nighttime surface temperature change
from agricultural to urban/industrial in
north zone .................................. 379

108 Winter nighttime surface temperature change
from agricultural to urban/industrial in
south zone.................................. 381

109 Winter nighttime surface temperature of
agricultural land-cover types vs coldest
natural land-cover.......................... 383

110 Winter nighttime surface temperature of
urban/industrial land-cover types vs
coldest natural land-cover.................. 385

111 Winter nighttime surface temperature change
for special conditions...................... 388

112 Winter diurnal surface temperature variation
differences among natural land-cover
types in panhandle zone..................... 391

113 Winter diurnal surface temperature variation
differences among natural land-cover
types in north zone......................... 393

114 Winter diurnal surface temperature variation
differences among natural land-cover
types in south zone......................... 396

115 Winter diurnal surface temperature variation
differences among agricultural land-cover
types in panhandle zone..................... 399


xxi









116 Winter diurnal surface temperature variation
differences among agricultural land-cover
types in north zone......................... 400

117 Winter diurnal surface temperature variation
differences among agricultural land-cover
types in south zone......................... 403

118 Winter diurnal surface temperature variation
differences among urban/industrial land-
cover types in panhandle zone............... 406

119 Winter diurnal surface temperature variation
differences among urban/industrial land-
cover types in north zone.................... 408

120 Winter diurnal surface temperature variation
differences among urban/industrial land-
cover types in south zone.................... 412

121 Winter diurnal surface temperature variation
change from natural to agricultural in
panhandle zone............................... 415

122 Winter diurnal surface temperature variation
change from natural to agricultural in north
zone......................................... 416

123 Winter diurnal surface temperature variation
change from natural to agricultural in south
zone ........................................ 417

124 Winter diurnal surface temperature variation
change from natural to urban/industrial in
panhandle zone............................... 418

125 Winter diurnal surface temperature variation
change from natural to urban/industrial in
north zone.................................. 420

126 Winter diurnal surface temperature variation
change from natural to urban/industrial in
south zone.................... .. ........ 422

127 Winter diurnal surface temperature variation
change from agricultural to urban/industrial
in panhandle zone........................... 424

128 Winter diurnal surface temperature variation
change from agricultural to urban/industrial
in north zone ............................... 425


xxii









129 Winter diurnal surface temperature variation
change from agricultural to urban/industrial
in south zone.............................. 427

130 Winter diurnal surface temperature variation
of agricultural land-cover types vs highest-
DSTV natural land-cover ..................... 428

131 Winter diurnal surface temperature variation
of urban/industrial land-cover types vs
highest-DSTV natural land-cover.............. 430

132 Winter diurnal surface temperature variation
change for special conditions............... 433

133 HCMM Winter approximate afternoon surface
temperature across-zone differences among
natural land-cover types ................... 440

134 HCMM Winter approximate nighttime surface
temperature across-zone differences among
natural land-cover types ................... 441

135 HCMM Winter approximate diurnal surface
temperature variation across-zone
differences among natural land-cover types 442

136 HCMM Winter approximate afternoon surface
temperature change for special conditions... 444

137 HCMM Winter approximate nighttime surface
temperature change for special conditions... 447

138 HCMM Winter approximate diurnal surface
temperature variation change for special
conditions.................................. 449

139 DSTV/soil-moisture relation for soil types.... 452

140 Image accuracy evaluation details............ 469


xxiii














LIST OF FIGURES


Figure page

1 Study area with climate zones and water-body
temperature stations (see Appendix B for
details) .................................... .40

2 AVHRR image without geographic correction
(polygon outlines true position of Florida). 52

3 AVHRR image with first-stage (ELP-based)
geographic correction (polygon outlines true
position of Florida)........................ 54

4 Example of explosive extrapolation of third-
order global polynomial surface model
outside of control points.................... 56

5 AVHRR image with second-stage (GCP-based)
geographic correction (polygon outlines true
position of Florida)........................ 59

6 At-satellite radiant temperature image........ 62

7 Emissivity image.............................. 63

8 Surface temperature image..................... 64

9 Water/cloud mask (NDVI) image.................. 71

10 Daytime (spring) masked surface temperature
image........................................ 73

11 Nighttime (spring) masked surface temperature
image........................................ 75

12 DSTV (spring) image............................ 76

13 HCMM daytime (winter) at-satellite radiant
temperature image......................... 80

14 HCMM nighttime (winter) at-satellite radiant
temperature image........................ 81

15 HCMM approximate-DSTV (winter) image.......... 82


xxiv









16 Natural land-cover polygons (see Appendix C
for details)............................. 93

17 Agricultural land-cover polygons (see
Appendix C for details).................. 116

18 Urban/industrial land-cover polygons (see
Appendix C for details).................. 124


xxV














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

APPLICATION OF REMOTELY SENSED DATA TO A GEOGRAPHIC
INFORMATION SYSTEM FOR MICROCLIMATE CHANGE ANALYSIS

By

Jonathan David Jordan

December 1994

Chairman: Sun F. Shih
Major Department: Agricultural Engineering

This study demonstrates the monitoring of surface

temperature pattern impacts of meso-scale changes in land-

cover and hydrology through the use of a geographic

information system (GIS) incorporating remotely sensed data.

Advanced Very High Resolution Radiometer (AVHRR) thermal-

infrared images, land-cover maps, and soil-type maps were

assembled as GIS database layers for a 1-km spatial resolution

study of Florida. Seasonal and diurnal surface temperature

patterns of many combinations of land-cover and soil type were

analyzed quantitatively. Effects of changes in land-cover and

hydrology due to both artificial (agriculture, urbanization,

wetland disturbance, and exotic-plant introduction) and

natural factors (drought, freeze, and hurricane) were studied.

The thermal-infrared images were calibrated to at-

satellite radiant temperature and geographically corrected for

importation into the GIS, then corrected for atmospheric


xxvi









effects and surface emissivity to produce surface kinetic

temperature. Seven soil types (from maps) and 32 land-cover

types (from maps, aerial photographs, and site visits) were

imported into the GIS as digitized polygons.

Results of analyses performed using the GIS indicated

that both land-cover and soil type, as well as soil moisture

and season, were significant factors influencing surface

temperature patterns in Florida. Surface temperature effects

(daytime "heat island", nighttime "cold island", and diurnal

variation "extreme island") of several agricultural and urban

soil type/season combinations matched or exceeded those found

among natural land-cover types. The surface temperature

effects of certain agricultural soil type/season combinations

matched or exceeded those of urban counterparts. Drought,

freeze damage, hurricane damage, wetland disturbance, and

exotic-plant introduction all produced significant changes in

surface temperature.


xxvii














INTRODUCTION


Surface temperature is a parameter relevant to many

fields. It is in demand for studies and models of

climatology, hydrology, agriculture, and forestry. There are

difficulties to be overcome in obtaining and applying surface

temperature data, which might be solved with new techniques

involving remote sensing and geographic information systems

(GIS). These matters are discussed below, with emphasis on

their importance to the state of Florida.


Purpose and Objectives


The purpose of this research was to use satellite imagery

together with a geographic information system to

quantitatively investigate the kilometer-scale relationship

between changes in land cover and those in surface

temperature. The specific objectives were 1) to obtain

remotely sensed thermal-infrared data at 1-km spatial

resolution for the state of Florida, using the National

Oceanic and Atmospheric Administration (NOAA) Television

Infrared Observation Satellite (TIROS) Advanced Very High

Resolution Radiometer (AVHRR) thermal-infrared imagery; 2) to

calibrate this thermal-infrared data to at-satellite radiant

temperature and to geographically correct the resulting

temperature imagery for inclusion in a GIS; 3) to calculate







2

surface kinetic temperature from at-satellite radiant

temperature through atmospheric correction and emissivity

correction; 4) to digitize land-cover and soil-type data

(obtained from maps, aerial photography, and site visits) into

polygons for inclusion in a GIS; 5) to build a GIS database

containing as coverage layers the surface temperature images,

soil type, land-cover (natural, agricultural, urban, and

industrial types), and special condition (drought, hydrologic

disturbance, freeze-damage, storm-damage); 6) to utilize the

GIS database in performing a quantitative analysis of seasonal

and diurnal relationships between land-cover, soil type, and

surface temperature; and 7) to utilize the GIS database in

performing a quantitative analysis of seasonal and diurnal

relationships between changes in land-cover type and changes

in surface temperature.

There were two additional objectives to supplement the

AVHRR work. These were 1) to perform a historical (1979)

surface temperature study of Florida based on the at-satellite

radiant temperature data of the National Aeronautics and Space

Administration (NASA) Heat Capacity Mapping Mission (HCMM)

satellite, and 2) to perform a ground-based evaluation of the

DSTV/soil moisture relationship for both mineral and organic

soil types.


Importance of Surface Temperature in Climatology


Surface temperature is one of critical parameters of

climate at macro-scale, meso-scale (Davis and Giles, 1990;







3

Giorgi and Mearns, 1991; Barron, 1992, Lewis and Wang, 1992;

McCabe and Wolock, 1992), and micro-scale (Auer, 1978; Lewis,

1984; Rourke, 1985; French and Krajewski, 1994). Macro-scale

climate (100 km to global spatial resolution, multi-year

temporal resolution) includes the basic macroclimate, or

"average weather", with its global weather systems,

atmospheric/oceanic/topographic influences, and long-term

perturbations from factors such as the El Nifo/Southern

Oscillation (ENSO), volcanic dust plumes, and greenhouse gases

(McClain et al., 1985; Harries, 1990; Mather and Sdasyuk,

1991). Macro-scale climate is not directly concerned with

near-surface (< 2 m above surface) processes (Akin, 1991).

Meso-scale climate (1 to 100 km spatial resolution, 1 day

to 1 year temporal resolution) includes local weather and is

directly concerned with near-surface processes (Henderson-

Sellers and Robinson, 1986; Harries, 1990; Hostetler and

Giorgi, 1993; Johannessen et al., 1993). Micro-scale climate

(1 km or finer spatial resolution, 1 day or finer temporal

resolution) is directly concerned with highly localized

effects, which are nested within (and do not influence) the

meso-scale climate (Henderson-Sellers and Robinson, 1986;

Harries, 1990; Akin, 1991).

Meso-scale is the smallest scale at which surface factors

have potential to force weather patterns; land-cover and/or

surface moisture organized at this scale produces an organized

atmospheric response (Shuttleworth, 1991). This response

includes air convection and turbulence (Henderson-Sellers and







4

Robinson, 1986), and even shifts in rainfall patterns due to

urban "heat-islands", irrigated-desert "oases", drained-marsh

"heat-plateaus", etc. (Auer, 1978; Dickinson, 1988; Abtew and

Khanal, 1994). In combination with soil moisture and acidity,

surface temperature controls soil-biogenic greenhouse-gas

emissions (Schimel et al., 1988; Yienger and Levy, 1994). Due

to these effects on weather patterns and greenhouse gas

emission, surface temperature is a meso-scale parameter which

must be linked to the current macro-scale global-circulation

models for further study of the global-change/greenhouse-

effect (Bolin, 1988; Risser et al., 1988; MacCracken et al.,

1990; Mather and Sdasyuk, 1991).


Importance of Surface Temperature in Hydrology


Surface temperature is an important parameter of surface

hydrology and its components such as evapo-transpiration (ET)

and soil moisture (Soer, 1980; Haan et al., 1982; Heimberg et

al., 1982; Price, 1984; Reiniger and Seguin, 1986; Ottle et

al., 1989; Sucksdorff and Ottle, 1990; Novak, 1991; Rodriguez-

Iturbe et al., 1991a, 1991b; Brutsaert and Parlange, 1992;

Chang et al., 1992; Nikolaidis et al., 1993; Zelt and Dugan,

1994). It is also an important factor in water conservation

topics such as lake and reservoir evaporation (Miller and

Millis, 1989; Hondzo and Stefan, 1991; Steinhorn, 1991; Mahrer

and Assouline, 1993), streamflow (Cayan, 1993), and effects of

oil spills on ocean evaporation (Mather and Sdasyuk, 1991).







5

A typical application of surface temperature in hydrology

is in the computation of the Bowen ratio (Linsley et al.,

1982):


P = 0.00066 p (T, Ta) / (eo ea) [1]


where P is the ratio of sensible heat transport to latent heat

transport, R is the atmospheric pressure (mbar), T, is the

surface temperature, T, is the air temperature (C), eo is the

air saturation vapor pressure (mbar) at T,, and ea is the air

vapor pressure (mbar). The Bowen ratio is used in the study

of reservoir evaporation and vegetation evapotranspiration

(Linsley et al., 1982).


Importance of Surface Temperature in Agriculture and
Forestry


Surface temperature is important to agriculture and

forestry as a factor of water-stress and growing-region for

crops and trees (Henderson-Sellers and Robinson, 1986; Seguin,

1989). Meso-scale surface temperature is of interest to

studies involving agricultural topics such as regional crop

condition and yield prediction (Idso et al., 1979, 1981;

Reginato, 1983; Taconet et al., 1986b; Hope and Jackson,

1989). It is a major factor involved in soil conservation

issues--such as soil subsidence (Lucas, 1982) and soil

degradation (Kilmer, 1982), which are critical to long-term

agricultural planning. Surface temperature is also of







6

interest in the assessment of forest-fire risk (Waters, 1976;

Chuvieco and Martin, 1994).

Together with soil moisture, meso-scale patterns and

changes of surface temperature are a major factor in outbreaks

of pests, parasites, and diseases in agricultural crops,

forest trees, and livestock (Uvarov, 1931; Geiger, 1950; Akin,

1991). Examples of such impacts include locust swarm-

behavior, Dutch elm disease, chestnut blight, tobacco blue

mold, Japanese beetle, Colorado potato beetle, potato blight,

cotton leaf worm, seedling-scald, and liver fluke (Uvarov,

1931; Rourke, 1985; Akin, 1991).


Factors Affecting Surface Temperature Patterns


Land surface temperature patterns at meso-scale are

forced by several factors which can change spatially and

temporally (Hillel, 1980; Risser et al., 1988; Lewis and Wang,

1992). These include macro-scale climate, solar irradiation,

geothermal heat sources, maritime effects, orogenic effects,

vegetation transpiration, root-zone soil moisture, soil type,

and land cover type.

Macro-scale climate has been discussed previously; its

impact on meso-scale surface temperature takes the form of

annual cyclic changes in average values of precipitation and

air temperature, which are well-documented for most places.

The macro-scale climate effect on surface temperature can be

estimated from comparison of data from similar natural land-

cover type pairs across macro-scale climate zones (temperate-







7

zone pine forest and subtropical-zone pine forest, temperate-

zone marsh and subtropical marsh, etc.). Advection effects

(strong winds, precipitation, etc.) from transient weather

phenomena such as storms, winter frontal systems, and

especially desert-winds harmattann, etc.) can have a

substantial impact on meso-scale temperature (and relative-

humidity), but these effects are sporadic and transient

outside of continental interiors (Hillel, 1980; Haan et al.,

1982), and are not addressed in this study. Solar irradiation

influences surface temperature through daily cyclic changes,

and is a primarily a function of latitude, date, and hour.

Geothermal heat sources are common at micro-scale--such as hot

springs, subterranean steam-lines (Axelsson, 1988), and

artesian wells (Jordan and Shih, 1988), but rare at meso-scale

(volcano and geyser areas).

Maritime effects operate at micro-scale (air advection

immediately adjacent to the coast) and at meso-scale

(increased humidity further inland) (Henderson-Sellers and

Robinson, 1986; Dickinson, 1988). The positions of meso-scale

water bodies and wetlands are well documented. Florida, due

to its proximity (at meso-scale) to the sea on every side, its

near sea-level elevation, and its lack of topographic

obstructions (mountains), is free from sources of major

variation in the meso-scale maritime effect. The micro-scale

maritime effect on surface temperature can be estimated from

comparison of data from coastal/inland pairs of similar







8

natural land-cover type, such as saltmarsh and freshwater

marsh, coastal hammock and inland hammock, etc.

Orogenic effects on meso-scale surface temperature can be

very pronounced in mountainous regions (Atkinson, 1985;

Henderson-Sellers, 1986; Barros and Lettenmaier, 1994). They

are nonexistent in level-terrain regions such as Florida.

Vegetation transpiration, root-zone soil moisture, soil

type, and land cover type are forcing factors of meso-scale

surface temperature which are strongly inter-related.

Vegetation transpiration is a daily cyclic phenomenon. Root-

zone soil moisture can change over hours or days; it is

artificially controlled in urban and agricultural areas, and

is a function of the soil type and macroclimate in natural

areas. Soil type is generally constant over time, and is

documented to various degrees in most of the world; it is a

particularly important surface temperature factor for cleared

areas.

Land-cover type is subject to both annual cyclic changes

(seasonal tree leaf cover, agricultural crop seasons) and

sudden changes (agricultural/urban development, natural

disasters). Land-cover type is well-documented at meso-scale

for most of the world, although available maps are often

overly simplistic in land-cover distinction--particularly for

agricultural land-cover. Meso-scale land-cover information

for application to meso-scale surface temperature work should

contain distinctions comparable to Level-III designations in

the United States Geological Survey (USGS) land use







9

classification system (Anderson et al., 1976)--for example,

"pasture" or "cropland" instead of simply the Level-II

"cropland and pasture" or Level-I "agriculture" designations.


Temperature Impacts of Changes in Land-Cover


Land-cover is a forcing factor of surface temperature

pattern which can experience meso-scale changes that are non-

cyclic and discontinuously-distributed both spatially and

temporally. These changes can be due either to natural causes

(volcanoes, storms, floods, droughts, wildfires, pests, etc.)

or artificial causes (related to agricultural and urban/

industrial development). International attention has mounted

in recent years concerning the worldwide extent of

deforestation (Mather and Sdasyuk, 1991), compared to the very

few regions currently experiencing a significant degree of

reforestation (Ireland, Senegal, England, and Algeria as of

1984).

Urban effects. Previous research has established the

concept of the urban heat island, which is characterized by

increased surface and air temperature (by 5 to 10 C) and

decreased relative-humidity in an urbanized area relative to

its surroundings (Eagleman, 1974; Lewis, 1984; Atkinson, 1985;

Balling and Brazel, 1988; Henry et al., 1989; Akin, 1991).

The more high buildings, more smog, and fewer trees, the more

pronounced the effect (Henderson-Sellers and Robinson, 1986).

The heat island is primarily a daytime effect within the

urbanized area, and has been shown to have a cellular







10

topology, rather than a smooth dome shape, due to the peak and

canyon geometry of the urban skyline (Atkinson, 1985). There

is a similar, but lesser, nighttime heat-island effect

(Eagleman, 1974; Henry et al., 1989). The importance of urban

temperatures to human well-being has been noted (Lewis, 1984;

Henderson-Sellers and Robinson, 1986; Meerow and Black, 1988).

There is a need for more detailed investigation of the

heat island effect. The heat island of urban areas has

usually been compared simply to non-urbanized areas nearby;

the influence of soil type has often been ignored. Thus, it

may be that a well-drained soil area selected for urban use

has always had an associated heat island relative to

surrounding areas of different soil type, even under its

natural cover. In addition, there is the possibility that the

inclusion of water-bodies, as is common in certain Florida

suburbs (finger canals) and mines (tailings ponds), may

counteract the urban heat island effect. The presence of

windbreak trees, as in golf-course suburb communities,

restricts the lateral flow of near-surface air, leading to

higher daytime temperatures in the open area between trees

than would be the case for a completely open field (Geiger,

1950; Crowe, 1971; Meerow and Black, 1988a; McCarty et al.,

1990). In closed-canopy parkland, the removal of undergrowth

and low tree branches increases the lateral flow of near-

surface air, producing a moderating effect on temperature

compared to open urban areas, but a decrease in humidity

compared to a natural forest (Lewis, 1984).







11

Agricultural effects. Agricultural effects on surface

temperature patterns have received less attention than urban

effects, which is undeserved, considering the vastly greater

areal extent of agricultural land-cover compared to urban

land-cover. For irrigated areas in deserts, a measurable

daytime oasis effect characterized by decreased surface and

air temperature and increased relative-humidity compared to

the surroundings has been noted (Hillel, 1980; Haan et al.,

1982; Henderson-Sellers and Robinson, 1986). For agricultural

land-cover in forest/wetland areas, a measurable daytime heat-

plateau effect characterized by increased surface and air

temperature and decreased relative-humidity compared to the

natural surroundings has been observed (Ghuman and Lal, 1987),

together with a corresponding nighttime cold-plateau of

decreased surface and air temperature compared to the natural

surroundings (Chen, 1979). The presence of belt-planted

trees, as in agricultural field windbreaks, restricts the

lateral flow of near-surface air, leading to higher daytime

temperatures in the open area between trees than would be the

case for a completely open field (Geiger, 1950; Crowe, 1971;

Meerow and Black, 1988a; McCarty et al., 1990).

Florida land-cover change. In Florida, natural causes of

meso-scale land-cover change have consisted of storms

(hurricanes), droughts, wildfires, and freezes. Artificial

factors of land-cover change have consisted of agricultural

development, urban development, industrial development, water-







12

control projects, naturalization projects, and invasion by

exotic vegetation.

Agricultural development in previous times (pre-Columbian

to early 1900s for panhandle and north; primarily from 1910 to

1950 for south) displaced much of the natural land-cover in

Florida (Fernald and Patton, 1984); in recent times it has

consisted primarily of changes in the type of agriculture--

especially between row-crops, pasture, and citrus orchard--on

the same land. Urban/industrial development in previous times

(primarily in the form of urban centers and strip mines)

displaced much of the natural and agricultural land-cover; in

recent times this development is continuing--particularly in

the form of suburbs.

Water-control projects are located in the Everglades

Agricultural Area (EAA), Kissimmee River basin, and Upper

Suwannee River basin (north-central Florida as well as south-

central Georgia). Naturalization projects include the

restoration of marshes (Payne's Prairie, Lake Apopka,

Kissimmee River basin, Lake Jessup, and the activity of

beavers naturally re-colonizing parts of the panhandle),

saltmarshes (Indian River Lagoon, Tampa Bay), and scrub (state

parks, local parks, and Archbold Biological Station). In

addition, efforts have begun in recent decades to reclaim

mined land for naturalization and even agriculture (Blakey,

1973).

Invasion by exotic vegetation has occurred at meso-scale

in south Florida (EPPC, 1990). This vegetation consists of







13

evergreen trees which have a very fast growth rate, and

transpire enough to decrease the root-zone soil moisture--

which is precisely why some of them were introduced in earlier

decades--to dry out wetlands (EPPC, 1990). The spread of

exotic forest has been beyond effective human control since

the 1950s (Barrett, 1956; EPPC, 1990), and has now reached

proportions of serious ecological and hydrological concern to

south Florida.


Temporal Inter-Relation of Forcing Factors


The net surface energy flux, soil type, and root-zone

soil moisture are forcing factors of the surface temperature

pattern which are inter-related temporally (Geiger, 1950;

Crowe, 1971; Kahle, 1977; Hillel, 1980; Bolin, 1988; Pollak,

1992). A bare-soil, flat-terrain, conductive heat-transfer

(non-advective) surface energy balance can be modeled

temporally by the harmonic equation (Mulders, 1987):


T(z,t) = Ta, + [F0/(PWl/2)]e-z/dsin(wt-z/d-r/4) [2]


where T(z,t) is the soil temperature (K) at depth (m) and

time 1 (t = 0 s at 0000 h) expressed as local solar time

(LST), Tv,, is the average (treated as constant) soil

temperature (K) at a depth of 2 to 3 m, Fo is the amplitude (W

m"2) of the net surface energy flux F, P is the soil thermal

inertia (J m-2 K-' s-1/2), d is the damping depth (m), and w is

the Earth angular rotation frequency (7.27x10-5 s-l). The net







14

surface energy flux, modeled in the form F = Fosin(wt), is

primarily a function of solar irradiation and near-surface air

temperature (Budyko, 1974; Henderson-Sellers and Robinson,

1986; Lewis and Wang, 1992). Soil thermal inertia is

expressed by the formula (Price, 1982):


P = (Apc)1/2 [3]


where A is the soil thermal conductivity (W m-1 K-1), p is the

soil density (kg m-3), and c is the soil heat capacity (J kg-1

K-1). The components X, p, and c are directly related to the

soil moisture content for a given soil type (Lillesand and

Kiefer, 1979; Carlson et al., 1981; Price, 1984; Curran,

1985). Therefore, the higher the soil moisture content, the

warmer the soil temperature during the night and the cooler

the soil temperature during the day, as has been noted in

numerous studies (Shih et al., 1986; Taconet et al., 1986b;

Sugita and Brutsaert, 1992). The above equation can be solved

for surface temperature, T,, yielding the equation (Mulders,

1987):


Ts(t) = T.,, + [Fo/(P 1/2)]sin(Ot-7/4) [4]


where the quantities are described as before. The w/4 term

translates (by 2r = 24 h) to a 3-hour time lag between maximum

F, (1200 h LST) and maximum surface temperature (1500 h LST).

Non-conductive components of soil heat transfer (dew

evaporation) can cause the surface temperature to vary







15

somewhat from that indicated by equation 4 (Hinkel and

Outcalt, 1993), but only for a short period in the early

morning (Schmugge, 1978; Price, 1982).

Diurnal surface temperature variation. Taking the

diurnal surface temperature variation (DSTV) of equation 4

leads to the equation (Mulders, 1987):


DSTV = T. T.i, = 2 F/ (PW1/2) [5]


where T,, is the maximum surface temperature (K) at t = 1500h

LST, T.in is the minimum surface temperature (K) at t = 0300h

LST, and P, w, and Fo are defined as before. Thus, DSTV is

inversely related to the root-zone soil moisture content, and

is a strong indicator of relative root-zone soil moisture

conditions for different locations of the same soil type on

the same day (Engman and Gurney, 1991). The fact that DSTV is

an indicator of daily-average root-zone soil moisture makes it

particularly useful to hydrologic modeling studies (Parlange

et al., 1992). The presence of a clay hardpan or bedrock

within the root-zone depth of shallow soils will lead to

deviations from the predicted DSTV of equation 5 (Hillel,

1980). This influence of foreign bodies within the root-zone

soil depth on DSTV has in fact been used at micro-scale to

locate buried objects/features such as abandoned mine tunnels

and bombs (Cloud, 1992).

Estimated seasonal DSTV. Seasonal values of bare-soil

DSTV can be estimated based on soil parameters. A typical







16

Florida value of Fo in summer is 0.003941 (cal cm-2 s-1) and in

winter is 0.001433 (cal cm-2 s-1) (ASHRAE, 1981). For a

mineral soil (sand or clay with typical 40% porosity), the

value of pc is 0.3 (cal cm-3 K-1) under dry condition and 0.7

under saturated condition, and the value of A is 0.0007 (cal

cm-1 s-1 K-1) under dry condition and 0.0052 under saturated

condition (Hillel, 1980). For an organic soil (peat with

typical 80% porosity), the value of pc is 0.35 (cal cm-3 K-1)

under dry condition and 1.15 under saturated condition, and

the value of X is 0.00014 (cal cm-' s-' K-I) under dry condition

and 0.0012 under saturated condition (Hillel, 1980). Plugging

these values into equations 3 and 5 produces estimated summer

DSTV (K) ranging from 15 (saturated) to 64 (dry) for mineral

soil, and from 25 (saturated) to 132 (dry) for organic soil;

it produces estimated winter DSTV (K) ranging from 6

(saturated) to 23 (dry) for mineral soil, and from 9

(saturated) to 48 (dry) for organic soil. Agricultural land-

cover values of DSTV can be expected to lie somewhere between

those of the saturated condition and those of the totally dry

condition.

Relevant depth of surface temperature/soil moisture

relation. The depth of soil to which the soil moisture

content is relevant to the surface temperature is a function

of the damping depth, d, which is given by the equation

(Hillel, 1980):


d = [2A/pcw)]1/2







17

where A, p, w, and c are defined as before. The values of X,

p, and c are functions of both soil moisture and soil type.

For a mineral soil (sand or clay with typical 40% porosity)

under a totally-dry condition, the value of d is about 8 cm;

for an organic soil (peat with typical 80% porosity) under a

dry condition, the value of d is about 3 cm (Hillel, 1980).

The attenuation factor of equation 2 is e-z/d, so that an

attenuation (100 e-z/d)% of 95% (relevant depth limit for

estimated soil-moisture) is reached at a depth of 3d,

corresponding to 24 cm for a mineral soil and 9 cm for an

organic soil. Under saturated soil condition, the value of d

increases to about 14 cm for mineral soil and about 5 cm for

organic soil (Hillel, 1980), increasing the respective

relevant depth limits to 42 cm and 15 cm. The moisture

content of such root-zone depths is of importance to surface

hydrologic modeling (Rourke, 1985; Risser et al., 1988; Milly,

1994; Zelt and Dugan, 1994) and climatological modeling

(Gillies and Carlson, 1994; Salvucci and Entekhabi, 1994;

Smith et al., 1994).

Vegetation influence on DSTV. Vegetated land cover

affects both the maximum and minimum surface-temperature

components of DSTV (Geiger, 1950; Luval et al., 1990).

Minimum surface temperature is raised by nighttime reflection

of soil-emitted energy back to the surface. The radiation

contribution from vegetation foliage at night is negligible--

foliage quickly reaches equilibrium with air temperature

(Hillel, 1980; Chen et al., 1982; Reiniger and Seguin, 1986;







18

Seguin, 1989; van de Griend and van Boxel, 1989). The

nighttime vegetation effect will decrease during winter for

deciduous vegetation species.

Maximum surface temperature is lowered by the daytime

evapotranspiration (ET) from vegetation. ET rate is a

function of vegetation type (exact species or cultivar),

ambient water-vapor pressure deficit, air temperature,

vegetation species, and vegetation water stress (Idso et al.,

1981a, 1981b; Wetzel et al., 1984; Reiniger and Seguin, 1986;

Taconet et al., 1986a, 1986b; van de Griend and van Boxel,

1989; Doyle, 1992). It should be noted that the transpiration

component of ET is present for most vegetation only from late

morning to afternoon (Bolin, 1988). If the source of

vegetation water stress is limited to root-zone soil moisture

(rather than salinity or damage from pests, diseases, wind,

hail, etc.), and the other ET factors are measured, the soil

moisture condition can be calculated. The relevant depth in

this case is dependent on the vertical distribution of the

root system (according to plant species and maturity), not on

the d-value of equation 6 (Rubin and Or, 1993). This relation

is the basis for the Crop Water Stress Index (CWSI), which is

widely used for scheduling the irrigation of agricultural

fields (Howell et al., 1983; Reginato, 1983; SOEMC, 1987; EI,

1991). Again, the daytime vegetation effect will decrease

during winter for deciduous vegetation species.

The diurnal effect on the net heat flux of the vegetated

surface, G (W m-2) can be represented (for a non-advective







19

situation) in form (Haan, 1982):


G = Fo sin(wt) LE [7]


where LE is the latent heat flux (W m-2) caused by the

vegetation. The term LE can be approximated using the Bowen

ratio, 3, producing the equation (Haan, 1982):


Fo sin(wt)
LE= [8]
1 +


where P typically ranges from 0.1 to 0.3 for humid-climate

conditions. Assuming P = 0.2, and substituting the LE from

equation 8 into equation 7, it follows that the DSTV for the

vegetated surface, DSTVv is


DSTVv = (0.3334) Fo/(PWl/2) [9]


which is a reduced-amplitude version of equation 5. The

assumption of full canopy closure is made here; the estimation

of exact effects of partial canopy (as in many forms of

agricultural land-cover) are a complex matter for study at

very fine spatial and temporal scales (Taconet et al., 1986a,

1986b; Massman, 1992).


Spatial Inter-Relation of Forcing Factors


The soil type, root-zone soil moisture, and land-cover

type are forcing factors of the surface temperature pattern

which are inter-related spatially (Akin, 1991). Different







20

types of agriculture require the maintenance of different

levels of soil moisture (Ziegler and Wolfe, 1961; Snyder,

1978; Henley, 1983; McCarty and Cisar, 1990)--standing water

for rice, taro, and fish-farm; high water-table for sod-farm;

medium water-table for winter-vegetables, sugarcane, potato,

strawberry, blueberry, blackberry, and pasture; and relatively

low water-table for leatherleaf fern, citrus, and most other

fruit trees.

Likewise, different soil types require different

agricultural management practices--irrigation of deep sands

and loams; drainage of organic soils and marls; and both

irrigation and drainage of spodosols and rockland soils

(Jones, 1948; Stewart et al., 1963; Hochmuth and Hanlon,

1989). Soil type and seasonal soil-moisture levels dictate

the natural land-cover type (scrub, forest, swamp, marsh,

etc.) and limit the possibilities of agricultural land-cover

types (citrus primarily to mineral soil, sugarcane primarily

to organic soil, blueberry to acid soil, atemoya to sub-

alkaline soil, etc.) (Critchfield, 1960; Schimel, 1988; Akin,

1991).

Urban development, however, is less impeded by soil type.

This is particularly evident in Florida, where coastal sands,

flatwoods sands, marls, and even mucks have been urbanized, as

well as the more conventional deep sands, upland loamy sands,

and sandy rockland.







21

Potential for Future Changes in Soil Type


General soil type, which is a constant forcing factor of

meso-scale surface temperature for mineral-soil areas, can

change for organic-soil areas, or even for mineral-soil areas

where mass-wasting occurs (Risser et al., 1988). Organic

soils drained for conventional agricultural use tend to

subside and eventually disappear (Snyder, 1978; Kilmer, 1982;

Lucas, 1982; Abtew and Khanal, 1994). This effect was well

understood even at the time the EAA water-control system was

being planned and implemented (Jones, 1948). The current rate

of organic soil subsidence in the EAA is about 1 inch per year

(Snyder, 1978); the natural temporal scale of soil type change

is typically of the order of 10,000 years (Dickinson, 1988;

Lucas, 1982). The potential mesoscale surface-temperature

impact of such a soil-type change in Florida is greatest in

the EAA, where organic soil will likely be replaced in the

near future by sandy/marly rockland if the agricultural land-

cover does not change to aquatic crops or restored marsh.














REVIEW OF LITERATURE


The techniques involved in the remote sensing of land

surface temperature are described. Previous studies, their

methods, and their results are discussed.


Difficulties of Surface Temperature Measurement


Unlike air temperature, surface temperature cannot be

interpolated between point measurements on land surfaces (due

to differences in several spatially-variable factors); to

obtain it synoptically at meso-scale requires some form of

satellite-based radiometry. Thermal-infrared and passive-

microwave radiometry are the two types applicable to the

typical Earth-surface range of temperatures (Lillesand and

Kiefer, 1979; Owe and Chang, 1988; Engman and Gurney, 1991).

Both of these are available from various satellite platforms.

Passive-microwave sensor data have the advantage of cloud

penetration, but are much more limited in spatial resolution

and temperature accuracy than are thermal-infrared sensor data

at this temperature range, and are sensitive to extraneous

factors such as surface microwave-roughness and radio-

communication interference (Lillesand and Kiefer, 1979; Owe

and Chang, 1988; Harries, 1990; Engman and Gurney, 1991; Owe

et al., 1992). Passive microwave data are typically used at

meso-scale (from satellite platform) for ice/snow water-

22







23

content mapping (Mather and Sdasyuk, 1991) or atmospheric

sounding (Miller et al., 1994), or at micro-scale (requiring

an aircraft platform) for surface temperature or soil moisture

mapping (Ijjas and Rao, 1992; Appleby et al., 1993; Paloscia

et al., 1993). Due to the above considerations, thermal-

infrared radiometry was selected as the source of surface

temperature data in this study.

The extraction of land surface temperature data from

satellite thermal-infrared radiometer measurements involves

three different processes--sensor calibration, atmospheric

correction, and emissivity correction. Errors incurred in any

of these processes will degrade the accuracy of the surface

temperature data (Marlatt, 1967; Llewellen-Jones et al., 1984;

Schott, 1989; Ben-Dor et al., 1994). Each satellite-based

thermal-infrared sensor has its own level of radiometric

precision and its own standardized calibration technique,

which the user must consult (Wolfe and Zizzis, 1978; Tebo,

1994a). A typical order of magnitude of radiometric precision

for thermal-infrared radiometers is 0.1 C (Myhre et al., 1988;

NOAA, 1991).

Ignoring the atmospheric correction can result in surface

temperature underestimates of up to 20 C (Llewellyn-Jones et

al., 1984; Price, 1984; Sobrino et al., 1991); this source of

error is primarily of concern in comparison of surface

temperatures from more than one time or date (in addition,

substantial spatial variation in the atmospheric effect can be

expected for images covering continental distances). Ignoring







24

emissivity correction over land surfaces can result in

underestimates of up to 10 C (Barton and Takashima, 1986);

these errors will vary spatially within an individual image--

resulting in an inability to accurately compare the surface

temperature of one location to the next. Therefore, if

atmospheric or emissivity corrections are not used, the

remotely-sensed surface temperatures may be in error

(conservative) by at least two orders of magnitude compared to

the sensor radiometric precision.


Atmospheric Correction Techniques


Atmospheric correction is needed for remotely sensed

thermal data, due to the combined effects of absorption,

scattering, and in-path radiance by the atmosphere upon

thermal-infrared imagery, even in "atmospheric window" bands

(Lillesand and Kiefer, 1979; Ben-Dor et al., 1994). The net

effect on satellite thermal-infrared radiometry is an

attenuation. Three general types of atmospheric correction

techniques have been developed.

Atmospheric modeling. Detailed modeling of the

atmosphere has been used successfully for atmospheric

correction of thermal-infrared data (Henry et al., 1989;

Wukelic et al., 1989; Luval et al., 1990; Gillies and Carlson,

1994; Sadot et al., 1994). It requires atmospheric data,

which in most instances has been provided by radiosondes,

although laser-based techniques are currently being developed

for this purpose (Tebo, 1994b). Atmospheric sensor







25

instruments exist on most current satellite platforms, but

these have very crude spatial resolution and are used for

global atmospheric research rather than meso-scale atmospheric

modeling (Zhang, 1993; Aumann and Pagano, 1994; Ellingson et

al., 1994). The collection of atmospheric data by radiosondes

is relatively expensive, and cannot be substituted with non-

local radiosonde data, non-simultaneous radiosonde data, or

estimated atmospheric values (Kerr et al., 1992).

Split-window techniques. Empirical "split-window"

techniques, based on differential atmospheric absorption

effects in different thermal-infrared bands, have long been

used successfully in atmospheric correction of thermal-

infrared data (Price, 1984; Llewellyn-Jones et al., 1984;

McClain et al., 1985; Cornillon et al., 1987; Cooper and

Asrar, 1989; Vidal, 1991). They require a sensor that

possesses multiple thermal-infrared bands, which is

fortunately available from various satellite platforms. They

also require ground-based surface-temperature measurements to

allow the initial calculation of their coefficients, but these

have already been established and reported for the individual

techniques (McClain et al., 1985; Di and Rundquist, 1994).

There are inherent difficulties in applying these techniques

to land surface studies (Becker, 1987; Sobrino et al., 1991),

the most important of which are the assumptions that the

surface emissivity is homogeneous within the instantaneous

field of view (IFOV) of the sensor, and that it is equal to 1.

Therefore, split-window techniques are primarily restricted to







26

sea-surface studies, where these assumptions are usually

valid.

Water-body reference technique. The water-body reference

technique requires one or more large water bodies, with near-

surface water temperature measurements taken simultaneously

with the thermal-infrared image (Lillesand and Kiefer, 1979).

These measurements must be taken for each image date and time,

which can pose a logistical problem for studies involving a

temporal series of images. Fortunately, such water-body data

are collected and made readily available in Florida by various

environmental agencies.

This technique includes the assumption that within the

sensor IFOV the water-body surface temperature and emissivity

(not necessarily equal to 1) are homogeneous. This assumption

is generally valid. It also assumes that the water body is

not under conditions such as very hot and dry (desert-climate)

ambient air or strong winds, which would produce a difference

between skin and near-surface bulk temperature of the water.

This assumption is valid for non-desert water bodies under low

wind conditions (Cornillon et al., 1987).


Emissivitv Correction Techniques


Radiant (blackbody) temperature values can be converted

to kinetic (surface) temperature values if the emissivity (in

the sensor bandwidth) is known with sufficient accuracy. The

physical foundation for kinetic temperature calculation by







27

radiometry is described by the Planck function (Saito et al.,

1992):


e C A-5
E(A, T) = [10]
exp [C2/(A T)] 1


where E is the measured energy (W m-2 m-m1), X is wavelength

(Im), Ci is a constant (3.74x108 W m-2 Am4), C2 is the Boltzmann

constant (14,388 Am K), T is the kinetic temperature (K), and

e is the emissivity. A commonly encountered formula for

kinetic temperature calculation is the Stefan-Boltzmann

equation (Lillesand and Kiefer, 1979):


Tkin = b1/4 Trad [11

where Tki, is the kinetic temperature (K), Trad is the radiant

temperature (K), and eb is the broadband emissivity (0 to 1).

However, this formula was designed for application to

laboratory situations of broadband (all wavelengths)

radiometry, rather than the "atmospheric window" band

radiometry performed by satellite sensors (midwave thermal-

infrared window at 3-5 im, longwave thermal-infrared window at

8-14 Am). To account for this, some researchers (Davies et

al., 1971; Price, 1983) have used a version of the Stefan-

Boltzmann equation modified for use with longwave thermal-

infrared sensor bands (Price, 1983):


Tkin = 6w1/4.5 Trad


[12]







28

where Tkl, and Trad are defined as before, and el, is the

longwave emissivity. This modified Stefan-Boltzmann formula

still contains simplifying approximations which limit its

accuracy in application to remotely sensed thermal data. The

integration of equation 10 over the bandwidth of given sensor

provides a better formula for quantitative longwave thermal-

infrared radiometry (Singh, 1985; Driggers et al., 1992; Saito

et al., 1992). It is given by the equation (Singh, 1985):


k CWNi
Tkin = [13]
In [1 E1, + E,1 exp(k CWN, / Tradi)


where Tkin is defined as before, k is the Boltzmann constant

(1.43883 cm K), Trad,i is the radiant temperature (K) in band i,

CWNi is the central wave number of band i (cm-1), and e is the

longwave emissivity. The central wave number is generally

documented for each band of a given sensor.

Emissivity presents the difficulty of being a property

that is calculable, rather than directly measurable (Fuchs and

Tanner, 1968; Friedman, 1969; Hejazi et al., 1992). Two

different methodologies have been employed to obtain

emissivity for processing remotely sensed thermal image data.

Emissivity by assignment. This commonly used technique

is based on longwave emissivity values determined in the field

or laboratory from close-range, longwave thermal-infrared

radiometry of small samples (Buettner and Kern, 1965; Fuchs

and Tanner, 1968; Friedman, 1969; Taylor, 1979; Barton and







29

Takashima, 1986; Rees, 1990; van de Griend et al., 1991;

Vidal, 1991; Salisbury and D'Aria, 1992). An emissivity value

is then assigned to a particular sensor IFOV according to

land-cover information or the normalized-difference vegetation

index (NDVI) (Gervin et al., 1985; Henry et al., 1989; Kerr et

al., 1992; Brown et al., 1993; van de Griend and Owe, 1993).

This method is primarily limited to aircraft or ground

radiometry, since the laboratory-determined emissivity values

are fine-resolution quantities bearing little relation to the

pixel-average emissivity value corresponding to a satellite

sensor IFOV (Curran, 1985; Jupp et al., 1988; Masuda et al.,

1988). An additional problem with this technique in

agricultural areas is that bare sand emissivity is controlled

by the moisture content of a very thin surface layer, and can

change over a short period of time (Fuchs and Tanner, 1968).

Twin-band technique. The physically-based "twin-band"

technique allows the calculation of pixel-average emissivity

from a thermal-infrared image (Artis and Carnahan, 1982):


,ij = exp {k (Tradi-Tradj) / [Trad.i Tad,j (Xi- j)]) [14]


where eij is the emissivity over the wavelengths from band i

to band j, Trad,i and Trad.j are the radiant temperatures (K) in

bands i and j, k is the Boltzmann constant (1.43883x10-2 m K),

and Ai and Xj are the respective central wavelengths (m) of

bands i and j. This method requires a sensor possessing two

thermal-infrared bands that are synoptic, spectrally close,







30

and spectrally narrow. This makes it inapplicable to the

single-band thermal-infrared data acquired by most of the

current satellite sensors (GOES-VISSR, Nimbus-7, Landsat-TM,

Meteor). Fortunately, the twin-band requirement is met by the

NOAA TIROS satellite series AVHRR sensor longwave bands 4 and

5. More sophisticated emissivity correction techniques

involving three thermal-infrared bands have been developed

(Hejazi et al., 1992), but few current satellites produce

imagery containing triplet thermal bands (AVHRR band 3 does

not form a triplet with bands 4 and 5, since it is a midwave

thermal-infrared band). Such multi-band techniques will

likely be the methods of choice for emissivity correction of

satellite-based surface temperatures from the more advanced

sensors aboard satellites of the future international Earth

Observing System (EOS) program.


Previous Studies


Previous studies involving remotely sensed land surface

temperature have been hampered by various factors, including

sensor limitations, lack of adequate processing techniques,

lack of consideration for one or more of the surface

temperature forcing factors, and lack of an adequate

manipulation technique for large quantities of spatial and

temporal data. In particular, there has been a longstanding

difficulty in the mixing of raster (remotely sensed) data with

vector (map) data in climatological studies (Mather and

Sdasyuk, 1991). Consistency of procedural attention to the







31

various components of meso-scale land-surface temperature

research needs to be improved so that individual project

databases can be made mutually compatible.

Chen (1979. 1980). Chen (1979, 1980) used NOAA

Geostationary Operational Environmental Satellite (GOES)

Vertical Infrared Spin-Scan Radiometer (VISSR) thermal-

infrared, winter, nighttime images of peninsular Florida in a

multi-year, meso-scale study of the feasibility of monitoring

agricultural areas for potentially crop-damaging cold

temperatures. Geographic correction was performed by NOAA

using satellite orbital telemetry; the resulting images

required manually-fitted offsets of up to 3 pixels (VISSR

thermal-infrared has 8 km nominal resolution at nadir).

Emissivity by assignment was used for both agricultural and

natural land-cover. Comparison of calibrated at-satellite

temperatures to ground-measured surface temperatures (from

hand-held thermal-infrared radiometer) and near-ground air

temperatures (from thermometry) indicated a range of error of

up to 5 C (for both sets of data), which nonetheless allowed

for determination of statistically significant differences in

temperature between broad categories of land-cover

(agriculture, marsh) and soil (organic). This study indicated

the potential for detailed research into meso-scale,

satellite-based, land-surface temperature impacts of land-

cover type and its relevance to agriculture, as well as the

difficulties of image registration, and especially the







32

importance of full correction for both atmospheric and

emissivity effects.

Cornillon et al. (1987). Cornillon et al. (1987) used

NOAA TIROS AVHRR thermal-infrared images to construct an

archive of quality-assured meso-scale sea-surface (Atlantic

ocean) temperature maps. Geographic correction was performed

based on TIROS satellite orbital telemetry data and ground-

control points; accuracy of pixel registration was to the

nearest 1.5 km (AVHRR has 1.1 km nominal resolution at nadir).

Because entire AVHRR scans (including image extremities or

"limbs") across continental distances were used, the images

were corrected for scan-angle effects. Ordinarily, AVHRR

longwave thermal-infrared images do not require correction for

scan-angle effects (Masuda et al., 1988; Kerr et al., 1992).

The water-body atmospheric correction technique was used, and

verification datasets indicated an accuracy to the nearest

0.51 C for sea-surface temperatures. This study illustrated

the value of accuracy assessment for both image registration

and surface temperature, and the surface temperature accuracy

attainable through atmospheric correction by the water-body

method.

Balling and Brazel (1988). Balling and Brazel (1988)

used NOAA TIROS AVHRR thermal-infrared images in a meso-scale

study of the urban heat-island effect in Phoenix, Arizona.

Images were geographically corrected based on satellite

orbital telemetry data. No atmospheric correction procedure

was reported. Emissivity correction was performed by







33

assignment (several values), and the Stefan-Boltzmann equation

was used to compute surface temperature. A lack of GIS

capability led to analyses based on linear transects of image

features, rather than on areal extractions. No accuracy

evaluation for geographic registration or surface temperature

was reported, but statistically significant differences in

urban center and suburb heat island effects were observed.

This study demonstrated the feasibility of measuring the

surface temperature impact of a purely meso-scale land-cover

type with AVHRR images, and the need for both accuracy

evaluation and a GIS-based analytical technique.

Cooper and Asrar (1989). Cooper and Asrar (1989) used

NOAA TIROS AVHRR thermal-infrared images in a meso-scale study

of land surface temperature in Kansas. Geographic correction

was performed by NOAA using satellite orbital telemetry. No

registration accuracy analysis was reported. Lack of GIS

overlay-analysis capability required the use of triangulation

between image features (lakes) to delineate study areas.

Atmospheric correction was performed by several techniques--

including both atmospheric modeling methods (with radiosonde

data) and split-window techniques. Ground-based radiometer

measurements of temperature were used to evaluate the accuracy

of the satellite-based surface temperature values. These

ground-based measurements themselves had a variance of about

6 (C2), due to the high variability of surface temperature at

their very micro-scale (1 m), even though the land-cover was

uniform (prairie grassland). Emissivity correction was







34

performed by assignment of a single value for the entire land

surface. The atmospheric modeling methods were found to

produce acceptable surface temperature accuracies (to nearest

3 C); all but one of the split-window techniques produced

unacceptable surface temperature accuracies. This study

indicated the difficulties associated with non-GIS

manipulation of remote sensing data, as well as the

fundamental unsuitability of ground-based point measurements

of land-surface temperature as a basis for evaluating

satellite-based meso-scale average land-surface temperature.

Henry et al. (1989). Henry et al. (1989) used NASA HCMM

satellite thermal-infrared images in a meso-scale, GIS-based

study of the urban heat-island effect of Gainesville, Florida.

Detailed land-cover information corresponding to the United

States Geological Survey (USGS) classification system

(Anderson et al., 1976) came from maps and aerial photographs.

Geographic correction was performed based on ground control

points and a first-order polynomial surface model;

registration accuracy was estimated solely by the root-mean-

square (rms) error of the fitted points ( 0.6 pixel).

Atmospheric correction was performed by the atmospheric

modeling method, with radiosonde data. Emissivity correction

was performed by assignment of a single value for all

urban/suburban land-cover. Quality evaluation was very

limited--based on near-surface air temperature measurements

collected non-simultaneously (different year) from the

satellite data; the authors acknowledged that even







35

simultaneous measurements of surface and near-surface air

temperature could vary on the order of 10 C. Despite this

problem, general heat-island impact differences were noted for

several urban and rural land-cover types through GIS-based

analyses. This study showed the power of GIS as an analytical

tool for linking satellite image data and map data, the value

of geographic registration accuracy evaluation, the necessity

of temperature accuracy evaluation, and the unsuitability of

using near-surface air temperature to evaluate surface

temperature accuracy.

Luvall et al. (1990). Luvall et al. (1990) used airborne

thermal-infrared sensor images in a micro-scale study of Costa

Rican rainforest canopy temperature and ET. Atmospheric

correction was performed by the atmospheric modeling method,

with radiosonde data. No emissivity correction was reported,

but the surface-temperature study was limited to full-canopy

vegetation surfaces having emissivity near unity. Lack of

both geographic correction and GIS capability led to eyeball

estimates of image subsets corresponding to polygons on aerial

photographs. Verification data in the form of thermocouple

leaf-temperature measurements at the top of the canopy

indicated an average remotely-sensed surface temperature

accuracy to the nearest 1.1 C. This study indicated the

potential for remote measurement of surface temperature over

forest canopy, the suitability of vegetation surface

temperature measurements for evaluating the accuracy of

remotely-sensed vegetation surface temperature, the need for







36

GIS-based analysis, and the need for geographic registration

and accuracy assessment.

Sucksdorff and Ottle (1990). Sucksdorff and Ottle (1990)

used NOAA TIROS AVHRR thermal-infrared images in a meso-scale

study of ET in Finland. Geographic correction was performed

by registration to a base map. Atmospheric correction was

performed by the atmospheric modeling technique, using

radiosonde data. No accuracy evaluation was reported for the

geographic correction or the temperature data. This study

demonstrated the use of base-map image registration to

construct a raster GIS database, as well as the need for

evaluation of geographic and temperature accuracy.

Leshkevich et al. (1993). Leshkevich et al. (1993) used

NOAA TIROS AVHRR images to construct an archive of quality-

assurred Great Lakes water-surface images. Geographic

correction was performed based on satellite orbital telemetry

data; manual offsets of up to 10 km were required for image

registration to allow construction of a raster GIS. No

further evaluation of geographic accuracy was reported.

Atmospheric correction was performed by a split-window

technique. Lake surface temperatures (day and night) were

determined to be accurate to the nearest 1 C, based on

comparisons with near-surface water-body temperature

verification data. This study illustrated the value of GIS as

an analytical tool, and the suitability of water-body

temperature measurements in accuracy assessment of remotely

sensed water surface temperature values.







37

Gillies and Carlson (1994). Gillies and Carlson (1994)

used NOAA TIROS AVHRR images in a study to estimate meso-scale

surface moisture-availability (ratio of soil moisture content

to that at field capacity) in northeast England. Four

afternoon spring and summer images from 1989 to 1990 were

calibrated to at-satellite radiant temperature. Atmospheric

correction was performed by the atmospheric modeling method,

using radiosonde data. No emissivity correction was reported.

Geographic correction was performed based on ground control

points, and the positional accuracy was determined to be

acceptable (0.8 km maximum rms error). Moisture-availability

estimation was evaluated based on point measured data, and was

found to be accurate to the nearest 5 to 7%. This study

indicated the potential for quantitative analyses of surface

parameters obtained from satellite images, and the increased

geographic accuracy resulting from the use of ground control

points, rather than relying solely on satellite orbital

telemetry for image geographic registration.

Current directions in land surface temperature research.

A methodology for detailed, meso-scale, quantitative study of

land-surface temperature patterns involving a GIS containing

the full set of forcing factors has been called for by several

researchers (Taconet et al., 1986a; Henderson-Sellers and

McGuffie, 1987; Dickinson, 1988; MacCracken et al., 1990;

Lagouarde, 1991; Mather and Sdasyuk, 1991; Sobrino et al.,

1991; Dozier, 1992; Kerr et al., 1992; Brown et al., 1993;

Lindsey et al., 1993; Wheeler, 1993). This methodology would







38

necessarily involve careful attention to the matters of

atmospheric correction, emissivity correction, geographic

correction, and quality analysis of both temperature and pixel

registration (Henderson-Sellers and Robinson, 1986; Harries,

1990; Mather and Sdasyuk, 1991; Peters et al., 1992). These

recommendations were put forth in simplest form by Heimburg et

al. (1982, p. 128):

An estimation procedure can be no more accurate than
allowed by its weakest part. From this perspective, the
most important area for future research is development of
operational methods to determine surface temperature and
net radiation from satellite data. This development
includes solutions to the problems of image registration
and atmospheric absorption corrections. The ability to
accurately overlay visible and [thermal] infrared data
collected at different times from the same area on the
earth's surface is critical to all remote-sensing methods
[for evapotranspiration estimation], as is the ability to
correct temperature and net radiation estimates for
atmospheric effects.

Satellite data suitable for such work now exist (Sader et al.,

1990; NOAA, 1991; Di and Rundquist, 1994; Gillies and Carlson,

1994), as do geographic information systems for performing

sophisticated analyses (ESRI, 1990; Lo and Shipman, 1990; Tan

and Shih, 1990; ERDAS, 1991; Wood, 1991; Rutchey and Vilcheck,

1994; Srinivasan and Engel, 1994; Wong, 1994).














MATERIALS AND METHODS


This research was performed on hardware consisting of a

personal computer (PC), high-resolution monitor, and

digitizing tablet. Software used included the PC versions of

the Earth Resources Laboratory Application Software (ELAS),

Earth Resources Data Acquisition System (ERDAS), and ARC/INFO.

The ground-based DSTV/soil-moisture work utilized soil augers,

hand-held radiometer, thermistors, and the gravimetric soil-

moisture analysis equipment of the University of Florida Soil

and Water Science Department.


Study Area


The Florida study area is shown in Figure 1. There are

three macroclimate zones--panhandle, north, and south (Fernald

and Patton, 1984; Schmidt, 1992).


Panhandle Zone


The panhandle zone includes the Florida Panhandle, which

for the purpose of this study is defined as the region to the

west of the St. Marks river. Its macroclimate is warm-

temperate, with a relatively wet winter (Fernald and Patton,

1984). Vegetation is limited to warm-temperate species

(Clewell, 1985). The natural forest trees include evergreen

broadleaf types and palms, as well as deciduous broadleaf








40

















Panhwdle







North

4













0 40kmrn
























Figure 1. Study area with climate zones and water-body
temperature stations (see Appendix B for details).







41

types and both evergreen and deciduous conifers. The

agricultural vegetation types include deciduous orchards and

nearly year-round pasture and crops.


North Zone


The north zone, for the purpose of this study, includes

the region to the east of the St. Marks river and to the north

of Lake Okeechobee. It is a transition zone, with a

macroclimate that is warm-temperate to sub-tropical, with a

winter that is relatively drier than that of the panhandle,

but not as dry as that of the south (Fernald and Patton, 1984;

Schmidt, 1992). Vegetation is limited to warm-temperate and

sub-tropical species. The natural forest trees include

evergreen broadleaf types and palms, as well as deciduous

broadleaf types and both evergreen and deciduous conifers.

The agricultural vegetation types include both evergreen and

deciduous orchards, and nearly year-round pasture and crops.


South Zone


The south zone, for the purpose of this study, includes

the region from Lake Okeechobee southwards. Its macroclimate

is sub-tropical, with a distinctly dry winter (Fernald and

Patton, 1984; Schmidt, 1992). Vegetation includes warm-

temperate, sub-tropical, and tropical species (Barrett, 1956;

Elias, 1980; Morton, 1982; FDNR, 1990). The natural forest

trees include evergreen broadleaf types and palms, deciduous

broadleaf types (some tropical), and both evergreen and







42

deciduous conifers. The agricultural vegetation types include

both evergreen and deciduous (some tropical) orchards, year-

round pasture and crops, and multi-season field crops.


Geographic Information System


The GIS database used in this research was assembled from

both raster and vector components, or "layers". The raster

layers included satellite images; the vector layers included

land-cover and soil type data.


Raster Datasets


A raster dataset consists of lines and elements ("rows"

and "columns"). The number of elements per line is a

constant, forming in concept a rectangular grid, each

identically-sized unit (picture element or "pixel") of which

is assigned data in the form of digital numbers (DNs). There

is one DN for each data type (sensor band, etc.) included in

the raster. ELAS and ERDAS were used to store and manipulate

the image raster files. Further information about raster

datasets can be found in ERDAS (1991).


Vector Datasets


A vector dataset consists of nodes and arcs which make up

individual polygons. One or more polygons may be included

within a "coverage" of a particular geographic region. For

example, a pasture-on-muck-soil coverage might consist of

several individual polygons distributed across Florida. Each







43

polygon can be assigned an attribute file, which contains one

or more types of data (such as ownership, water-quality

parameters, etc.). In this research, attribute files were not

constructed for the polygons, since the polygons were intended

for importation into the raster environment. ARC/INFO

(modules "ADS", "CREATE", and "TABLES") was used to record the

digitized map polygons as vector files. Further information

about vector datasets can be found in ESRI (1990).


Geographic Referencing


Both raster and vector datasets must be geographically

referenced in order to be included in a GIS (ESRI, 1990;

Connin, 1994; Wong, 1994). Raster datasets are geographically

referenced by knowledge of pixel size and the position

(measured either from pixel center or a pixel corner) of one

raster-corner pixel (typically the upper-left). Vector

datasets are geographically referenced by knowledge of the

position of each node. A consistent geographic coordinate

system must be used for all of the datasets in the GIS. The

Universal Transverse Mercator (UTM) system was selected for

use throughout this study. It is commonly used on maps having

conformal projection (suitable for navigation) published by

the United States Geological Survey (USGS) and other agencies.

Further information about geographic referencing and map

projections can be found in ERDAS (1990).







44

GIS Analyses in Raster Environment


Analyses in this study were performed in a raster

environment. Polygon files in ARC/INFO vector format were

converted to ERDAS raster format ("DIG" file) equivalents.

This conversion was performed using ARC/INFO modules

"TRANSFORM" and "UUNGEN", and ERDAS module "DXIN".

Statistical data (mean and standard deviation) were then

extracted from the raster data (temperature values)

corresponding to each polygon. This extraction process was

performed using the ERDAS modules "CUTTER", "STITCH" (for

assembling coverages of more than one polygon), and "BSTATS".

Where a given land-cover polygon contained more than one soil

type, it was subdivided (using ERDAS module "DIGSCRN") into

two or more final polygons having a single land-cover and a

single soil type.

Analyses of surface temperature patterns were performed

using the mean, standard deviation, and sample-size data

extracted from the GIS. Separate within-zone analyses were

performed for the three macroclimate zones. T-tests were run

at a = 0.05 and a = 0.01. The test statistic was given by the

separate-variance formula (Ott, 1988):


Xi Xz
T = [15]
(s12/n1 + s22/n2)0.5

where X& and xz are the respective means of samples 1 and 2,

si and s2 are the respective standard deviations of samples 1







45

and 2, and nj and n2 are the respective sample sizes of samples

1 and 2. The difference of two sample means was considered

statistically significant if T < -t /2 or T > t,/2. In the

analyses reported in this study, nj + n2 >> 30, so that the

critical values were t0.025 = 1.960 and to.005 = 2.576 (Walpole

and Myers, 1978).


AVHRR Image Processing


NOAA TIROS-AVHRR images of surface temperature were

obtained for two seasons and two times-of-day. Two day/night

pairs of images (14 December 1989 and 12 December 1992) were

required for complete winter coverage of Florida, due to

partial cloud contamination. A single day/night pair of

images (11 April 1993) was adequate for spring coverage of the

state. Details of individual images are provided in Appendix

A.

Winter (December) images allowed the analysis of

differences in surface temperature patterns due to natural

defoliation (for deciduous vegetation types) and agricultural

management practices (winter-crop season). Deciduous

vegetation in Florida includes both temperate and tropical

species, so that seasonal effects could be studied in all

three climate zones.

Spring (April) images allowed the analysis of differences

in surface temperature patterns due to growth flush (for

natural and many cultivated vegetation types) and agricultural







46

management practices (spring-crop season). It is a season of

particularly high irrigation demand in agricultural and

suburban areas in all three climate zones.

Time-of-day for the images included nighttime (late

night/early morning) and daytime (afternoon). Repeat coverage

of a given spot at nearly the same LST occurs every 9 days

(Kerr et al., 1992) for TIROS satellites, so that a minor

variation in coverage time occurs from day to day within this

period. There are also long-term changes in repeat-coverage

time for TIROS satellites; these are very gradual (years),

compared to changes for other weather-satellites such as the

Russian Meteor series (weeks). Nighttime (c. 0300 h LST),

surface temperature images represented the minimum values in

the diurnal cycle. Daytime (c. 1500 h LST) surface

temperature images represented the maximum values in the

diurnal cycle. The difference between these two values was

the DSTV, which indicated relative values of daily-average

root-zone soil moisture for a given land-cover/soil type

combination.


AVHRR Data Types


The NOAA TIROS-satellite AVHRR images were obtained from

the National Environmental Satellite Data and Information

Service (NESDIS) in the form of local area coverage (LAC)

level lb packed format data on computer-compatible tapes

(CCTs). Each tape was down-loaded, and then one raster file

of image data, one tabular file of Earth Location Points







47

(ELPs), and one tabular file of calibration coefficients were

extracted.

LAC-format AVHRR. AVHRR LAC images have the full spatial

resolution (1.1 km nominal at nadir) of the AVHRR; the

thermal-infrared bands have the full AVHRR radiometric

precision of 0.1 C, stored in 10-bit (0-1024 DN) data

precision (NOAA, 1991). The AVHRR bands are described (NOAA,

1988) as follows: band 1 (red) at 0.58 to 0.68 Am, band 2

(near-infrared) at 0.725 to 11.1 pm, band 3 (midwave thermal-

infrared) at 3.55 to 3.93 Am, band 4 (longwave thermal-

infrared) at 10.3 to 11.3 Am, and band 5 (longwave thermal-

infrared) at 11.5 to 12.5 pm. For typical earth surfaces (not

hot lava flows, fires, etc.), AVHRR bands 1 and 2 measure

reflected energy (daytime only), band 3 measures both

reflected (in daytime) and emitted energy, and bands 4 and 5

measure emitted energy (daytime or nighttime).

Other AVHRR formats. There are other forms of AVHRR

image which do not retain full spatial resolution nor full

radiometric precision, but are made available with greater

frequency than LAC. For each TIROS satellite, a single 10-

minute AVHRR High Resolution Picture Transmission (HRPT) image

per 102-minute orbit can be stored on-board for later

transmission to a NOAA ground reception station (NOAA, 1991),

but there are usually only two orbital coverages of a given

location per day, and not every transmitted HRPT image is

selected for inclusion in the NOAA LAC archive. More frequent

availability is provided by the NOAA-archived global area







48

coverage (GAC) format AVHRR images, which have reduced spatial

resolution (4 km nominal at nadir), but keep the original data

precision (10 bits); a complete orbital path of GAC data can

be stored on-board per orbit for later transmission to a NOAA

station (NOAA, 1991). Users with their own digital-signal

ground station can receive HRPT images directly, and achieve

a coverage frequency of at least twice per day from each

operational satellite, but they also have to calculate their

own ELPs and calibration coefficients from the HRPT telemetry

(Brush, 1985; Emery et al., 1989; Klaes and Georg, 1992).

Users with their own analog-signal ground station can

receive automatic picture transmission (APT) format AVHRR

images, with reduced spatial resolution (4 km nominal at

nadir) and reduced data precision (8 most significant bits

pre-analog), and achieve a coverage frequency of at least

twice per day from each operational satellite every day (NOAA,

1982b). APT images contain their own calibration information

(NOAA, 1982b, 1988; Olivier, 1990), but have no ELPs nor the

telemetry information to calculate them (NOAA, 1988). They

are limited to two NOAA-selected AVHRR bands--typically bands

2 and 4 in daytime, and bands 3 and 4 at nighttime.

Use of TIROS/Meteor APT archive. During the course of

this research, the APT ground station located at the Remote

Sensing Application Laboratory (RSAL) of the University of

Florida Agricultural Engineering Department was utilized to

obtain APT images for purposes of building a browse file for

selecting dates and times for ordering NOAA LAC images. These







49

APT images included multiple daily coverages by the four

current NOAA TIROS satellites (NOAA-9, -10, -11, -12) and the

short-lived but productive NOAA-13, and also the Meteor-APT of

various Russian Meteor-series weather satellites (Meteor 2-21,

3-3, 3-4, and 3-5). The Meteor APT consists of daytime

panchromatic (0.5 to 0.7 jm) images at somewhat finer spatial

resolution (2 km nominal at nadir). It should be noted by

users of weather-satellite data that imagery from the NOAA

TIROS-series satellites (as well as the Russian Meteor-series

satellites) is subject to temporary suspension on rare

occasions due to participation in the international Search and

Rescue Satellite (SARSAT) program in cases of emergencies at

sea (WMO, 1989; NOAA, 1991).


Calibration to At-Satellite Radiant Temperature


The thermal-infrared, 10-bit, image data of AVHRR bands

4 and 5 were calibrated to at-satellite radiant temperature by

the method of NOAA (1991). One pair of calibration

coefficients (scaled slope and intercept of the sensor

internal calibration) was extracted from the level lb LAC CCT

for each scan line of the image. At-satellite radiant

temperature was then calculated for the pixels of each line by

the equation (NOAA, 1991):


C2 CWNi
Tradsati = [16]
In [1 + C, CWNi3 / (Si DN + Ii)]


where Trad,sat,i is the at-satellite radiant temperature (K) in







50

band j, Ci is a constant = 1.1910659x10-5 (mW m-2 ster-1 cm4), C2

is a constant = 1.438833 (cm K), CWNi is the central wave

number (cm-1) for band I in one of three discrete target-

temperature ranges, Si is the scaled calibration slope (mW m-2

ster-1 cm) for band i, and Ii is the scaled calibration

intercept (mW m-2 ster-1 cm) for band i, and DN is the 10-bit

digital number (0 to 1023).

This calibration technique is based on a linear fit of

AVHRR sensor response to target temperature within three

discrete target-temperature ranges (180 to 225 K, 225-275 K,

275-320 K). It greatly reduces the at-satellite radiant

temperature errors (up to 4.3 K at extremities) that would

result from a simple two-point calibration over the entire

target-temperature range (180 to 320 K) of AVHRR data (NOAA,

1988). It should be noted that a although a single target-

temperature range (270-310 K) is generally employed in AVHRR

calibration for sea-surface work (NOAA, 1991), all three

standard target-temperature ranges have to be addressed in

daytime land surface work. In addition, there is an upper

limit of 320 K (47 C) for target temperature; the AVHRR

longwave thermal-infrared band sensors saturate (DN = maximum)

at this limit, and higher at-satellite radiant temperatures

cannot be recorded (NOAA, 1988; Chuvieco and Martin, 1994).

Fortunately, none of the images used in this study contained

at-satellite radiant temperature data reaching this saturation

limit.







51

Scaling of the output radiant temperature values was

performed in order to retain as high a level of radiometric

precision as possible (0.2 C) in the 8-bit (0 to 255 DN) data

storage format to be used in the GIS of this research. The

scaling was given by the equations:


DN, = 5 (Trad,,t,i + 10) [17]

and

DNi = 6 (Tradsa.t,i 15) [18]


where Trad,sat, is the at-satellite radiant temperature (C) in

band i, and DN, is the 8-bit scaled digital number output for

band i. Equation 17 was used for nighttime and winter

afternoon images; equation 18 was used for spring afternoon

images.


Geographic Correction and Registration


A base-map image was constructed to allow registration of

the AVHRR images prior to their importation into the GIS.

Weather-satellite images without any geographic correction are

completely unusable in a GIS (Figure 2). Base-map

construction, and subsequent AVHRR image registration, was

performed using a two-stage geographic correction process, as

is recommended (Thomas et al., 1987; Chen and Lee, 1992;

Peters et al., 1992) for highly warped (containing distortions

requiring a second-order or higher global polynomial surface

model) imagery.
























































Figure 2. AVHRR image without geographic correction (polygon
outlines true position of Florida).







53

The first stage involved the use of the extracted ELP

data from the AVHRR LAC image. These ELPs, which are provided

in the form of latitude/longitude coordinates, are imbedded in

the raw image at every 40th pixel along each scan line. They

form an evenly-distributed network of known geographic

coordinates throughout the entire image, which is the most

desirable situation for application of geographic correction

techniques. The accuracy of these coordinates is dependent

upon the accuracy of the satellite orbital telemetry data used

by NOAA to calculate them (NOAA, 1991). Early in the course

of this study, it was found that geographic correction based

solely on the ELPs produced output images with positional

errors of up to 10 km (Figure 3); this problem has been

reported in several AVHRR-based studies (Cornillon et al.,

1987; Nelson, 1989; Leshkevich et al., 1993). It should be

noted that the positional error shown in Figure 3 is not a

simple offset; there are still second-order distortions

present in the image.

In order to keep output positional errors closer to the

nominal spatial resolution (1.1 km) of the raw AVHRR LAC

images, a second stage of refined geographic correction was

performed based on ground control points (GCPs). The picking

of these GCPs simultaneously from monitor displays of the

AVHRR image (band 2 in daytime, band 4 at night) and from maps

was greatly facilitated by the first stage of correction,

which had removed most of the Earth-curvature and view-angle

distortions. Picking of GCPs directly from the raw image is
























































Figure 3. AVHRR image with first-stage (ELP-based) geographic
correction (polygon outlines true position of Florida).







55

not advisable, since the application of the high-order global

polynomial surface model required for such a distorted image,

combined with the relatively poor distribution of GCP network

obtainable from most images, can easily result in instability

of pixel fits interpolated between the GCPs (Thomas et al.,

1987), and "explosion" of pixel fits extrapolated outside of

the GCP network (Figure 4).

First-stage geographic correction of base-map image.

First-stage geographic correction was performed using the ELP

data extracted from the raw AVHRR LAC image. These ELP data

formed a 240-ELP grid (consisting of 16 grid lines with 15

ELPs each) having a 40 x 40 pixel spacing. The latitude/

longitude values in this ELP grid were converted to Universal

Transverse Mercator (UTM) northing/easting values (UTM zone 17

format) and entered into the ERDAS module "GCP". A second-

order global polynomial surface model was fitted to the ELPs.

The Nearest-Neighbor resampling technique was then applied to

the image; this is the only resampling technique that does not

corrupt (smooth or average) image data values (Lillesand and

Kiefer, 1979; Peters et al., 1992). These two operations were

performed using ERDAS modules "COORDN" and "NRECTIFY." The

polynomial fit (mapping equation) was as follows:

E' = 1000 ao + a1 E + a2 L + 0.001 a3 E2 + 0.001 a4 E L +

0.001 a5 L2
and
L' = 1000 bo + bi E + b2 L + 0.001 b3 E2 + 0.001 b4 E L +

0.001 b5 L2 [19]


























































Figure 4. Example of explosive extrapolation of third-order
global polynomial surface model outside of control
points.







57

where E was the original element number, L was the original

line number, E' was the fitted element number, L' was the

fitted line number, and the values of the coefficients were ao

= 1.371406, a, = -0.1608168E-2, a2 = -0.3716236E-3, a3 =

0.4853050E-6, a4 = 0.1728803E-6, a, = 0.2920270E-7, bo =

-2.333373, b, = -0.1690915E-3, b2 = 0.8806317E-3, b3 =

-0.2704551E-8, b4 = 0.2105952E-7, and b, = 0.3205131E-8, which

are unitless. Resampling was done to an output pixel size of

1 km. Details concerning geographic correction techniques for

satellite images can be found in remote sensing literature

(Lillesand and Kiefer, 1979; Gonzalez and Wintz, 1987; Thomas

et al., 1987; ERDAS, 1990; Novak, 1992; Peters et al., 1992;

Di and Rundquist, 1994).

Second-stage geographic correction of base-map image.

The first-stage output image was then imported into ELAS. A

set of 115 well-distributed GCPs (coastal features and lakes

of size appropriate to the spatial resolution of the image)

was picked simultaneously from a monitor display of the image

and from 1:500,000 scale UTM maps of Florida (DMAAC, 1987).

ELAS module "CPPP" and a digitizing tablet were the tools used

for this process. Panhandle GCP coordinates, located within

UTM zone 16, were converted to their equivalents in UTM zone

17 format. The first-stage output image was then re-imported

into ERDAS along with the GCP set mentioned above. A second-

order global polynomial surface model was fitted to the GCPs.

The Nearest-Neighbor resampling technique was then applied to







58

the raster. The polynomial fit (mapping equation) was as

follows:


E' = 1000 ao + a, E + a2 L + 0.001 a3 E2 + 0.001 a4 E L +

0.001 a5 L2

and

L' = 1000 bo + b, E + b2 L + 0.001 b3 E2 + 0.001 b4 E L +

0.001 b5 L2 [20]


where E was the original element number, L was the original

line number, E' was the fitted element number, L' was the

fitted line number, and the values of the coefficients were

a0 = 0.1952093, a, = 0.9931500E-3, a2 = -0.3316682E-5, a3 =

0.1140539E-7, a4 = 0.3552246E-8, a5 = 0.8012727E-9, b0 =

3.655660, bi = -0.5229729E-4, b2 = -0.1038144E-2, b3 =

0.1049909E-7, b4 = 0.1479114E-7, and b5 = 0.5142951E-8,

which are unitless. Second-stage resampling was done to an

output pixel size of 1 km, to form a base-map image of 1000

elements by 1000 lines covering the entire Florida study area

and small portions of southern Alabama and Georgia (Figure 5).

Location of the first pixel (element 1, line 1) was at

-187,000 m east and 3,584,000 m north (UTM zone 17 format) in

the ERDAS raster GIS reference system.

Accuracy assessment of base-map image. The positional

accuracy of the base-map image was carefully evaluated, since

all future images would be registered to it. ERDAS module

"COORDN" reported that the root-mean-square (rms) error for

























































Figure 5. AVHRR image with second-stage (GCP-based)
geographic correction (polygon outlines true position of
Florida).







60

the second-stage fit in the element direction was 0.648 km,

and in the line direction was 0.597 km; the overall rms error

for the fit was 0.881 km. These figures apply only to the GCP

pixels, not to other resampled pixels; they represent a form

of validation check. The potential for the user to be misled

by these GCP-based rms values can be seen in the

extrapolation-exploded image of Figure 4, which had a GCP-

based rms error of only 1.2 km, even though the image was

clearly rendered useless. In order to verify the positional

accuracy of the entire base-map image, a set of 91 well-

distributed GCPs (different from the 115 used to build the

base-map image) were digitized in the manner described above.

The resulting rms error was found to be 0.495 km in the

element direction, and 0.566 km in the line direction; the

overall rms error was 0.752 km. Considering both the overall

rms error of the fit (0.881 km) and that of the verification

(0.752 km), the base map was demonstrated to be spatially

accurate to within 1 km.

Registration of subsequent images to base-map image.

Subsequent AVHRR LAC images were registered to the base-map

image by a two-stage process with ELP-based first-stage

geographic correction similar to that described above. The

only change was in the manner of picking the GCPs for the

second-stage geographic correction; they were picked from

simultaneous monitor displays of the first-stage corrected

image and the base-map image (rather than maps). The ELAS

module "OCON" was used to simultaneously display these images







61

and pick the GCPs. These GCPs and the AVHRR LAC image being

registered were then imported into ERDAS for the second-stage

geographic correction. Second-order global polynomial surface

models were used whenever possible, but third-order models

were used if second-order models were insufficient to produce

registration rms errors below 1.2 km. Registered AVHRR images

had rms errors ranging from 1.09 to 1.2 km. Registration

details of individual images are given in Table 140 of

Appendix A.


Conversion from Radiant to Kinetic Temperature


In order to produce kinetic surface temperature images

from the at-satellite radiant temperature images, a three-

stage process was implemented. First, atmospheric correction

of an at-satellite radiant temperature image (Figure 6) was

performed by the water-body calibration technique. Second, an

instantaneous pixel-average emissivity image (Figure 7) was

constructed by the twin-band method. Third, the atmospheric-

corrected radiant temperature image and emissivity image were

used to produce a kinetic temperature image (Figure 8) based

on the analytical solution of the Planck equation. The

overall conservative bias due to the atmospheric effect, and

the "hiding away" of high surface temperature in urban and

agricultural areas due to the emissivity effect, are both

evident in the uncorrected image of Figure 6, when it is

compared to the fully corrected image of Figure 8.

























































Figure 6. At-satellite radiant temperature image.

























































Figure 7. Emissivity image.


























































Figure 8. Surface temperature image.









Atmospheric correction. The water-body calibration

technique described by Lillesand and Kiefer (1979) was

performed on the at-satellite radiant temperature images from

bands 4 and 5. This technique is based on the near-uniform

emissivity and relatively stable temperature (over the

satellite overpass time) of water-body pixels. Hourly near-

surface (0.5 m) water temperature data from three permanent

instrument stations located within Lake Okeechobee were

obtained from the South Florida Water Management District

(SFWMD) "DBHYDRO" database (Figure 1). These kinetic

temperatures (thermistor-based values reported to nearest 0.1

C) were converted to AVHRR band 4 and 5 radiant temperature

equivalents by rearranging equation 13 and plugging in the

water temperatures:


k CWNi
Trad.i = [21]
In (c.-1 [exp (k CWN, / Tkin) + E6 1]}


where Trad,i is the radiant temperature (K) calculated in band

i for a station pixel, Tkin is the measured water-surface

kinetic temperature (K), k is the Boltzmann constant (1.438883

cm K), CWNi is the central wave number for band i (cm-1), and

E, is the longwave emissivity of the lake water. A well-

established value of 0.99 was used for e, (Buettner and Kern,

1965; Barton and Takashima, 1986; Saunders, 1986; Masuda et

al., 1988; Wukelic et al., 1989; Salisbury and D'Aria, 1992).

Central wave numbers were obtained from the tabular values







66

published by NOAA (1991) for each TIROS satellite AVHRR

sensor, each band i, and each surface-temperature range. The

highest surface-temperature range was used to find the CWNi

for afternoon LAC images, and the sea-surface range was used

to find the CWN1 for nighttime and early-morning images. The

differences between the at-satellite radiant temperatures and

the atmospherically-corrected radiant temperatures were

averaged (from up to 3 values, according to station data

availability) to obtain atmospheric correction factors which

were applied to the at-satellite radiant temperature data from

bands 4 and 5 of each LAC image.

Emissivity correction. The atmospheric-corrected radiant

temperature images from bands 4 and 5 were used to produce a

longwave thermal-infrared emissivity image for each AVHRR

image by the twin-band technique of Artis and Carnahan (1982).

Instantaneous pixel-average emissivity calculation was

performed by plugging appropriate AVHRR values into equation

14:


ee = exp (k (Tr.d5-Trad,) / [Trad,5 Trad4 (X5-_4)] [22]


where el is the pixel-average longwave thermal-infrared

emissivity, k is the Boltzmann constant (1.43883 cm K), Trad,4

is the atmospheric-corrected band 4 radiant temperature (K),

Trad,5 is the atmospheric-corrected band 5 radiant temperature

(K), and A4 and A5 are the respective central wavelengths

(inverse of central wavelength number, m) of AVHRR bands 4 and







67

5. The last two values come from NOAA (1991) look-up tables

for the TIROS satellite, band, and temperature range, as

described previously. An 8-bit scaling of the calculated

emissivity values preserved the precision to the nearest 0.01:


DN = 100 e, [23]

where the parameters are described as before. The assumption

in this method that surface emissivity values in AVHRR bands

4 and 5 are identical is a simplification (Price, 1984);

slight differences in emissivity values between these bands

(up to 0.01) will result in an uncertainty of 1 C when the

calculated longwave-band emissivity is applied to kinetic

temperature calculation.

Kinetic temperature calculation. The emissivity image

was used with the atmospheric-corrected radiant temperature

image from band 4 to calculate the kinetic temperature image

for each AVHRR image. This computation was performed by

plugging appropriate AVHRR band 4 values and the pixel-average

emissivity values into equation 13:


k CWN4
Tkln = [24]
In [1 E1, + el, exp (k CWN4 / Trad,)]


where Tknf is the pixel-average kinetic temperature (K); Trad.4

is the atmosphere-corrected, pixel-average, band 4 radiant

temperature (K); k is the Boltzmann constant (1.43883 cm K),

CWN4 is the central wave number for band4 (cm-1), and e1 is the

pixel-average longwave emissivity. Band 4 central wave







68

numbers were obtained from the tabular values published by

NOAA (1991) for each TIROS AVHRR sensor and each surface-

temperature range. The highest surface-temperature range was

used to find the CWN4 for afternoon LAC images, and the sea-

surface range was used to find the CWN4 for nighttime and

early-morning images. Scaling of the output Tkin images to an

8-bit format was performed using the formula


S5 (Tin + 10), for low Tkin images
DN = 1[25]
6 (Tkin 15), for high Tkin images


where Tkin is the kinetic temperature (C), low-temperature

images were defined as those having a range of land surface

temperatures between -10.0 and 41.0 C, and high-temperature

images were defined as those having a range of land surface

temperatures between 15.0 and 57.0 C. In either case, a

precision of 0.2 C was kept by the scaling. It should be

noted that the AVHRR sensor saturation limit of 47 C applies

only to at-satellite radiant temperatures; a surface such as

a cleared sandy field or an urban area can have an afternoon

kinetic surface temperature well above 47 C and yet, due to

emissivity and atmospheric effects, it can easily have an at-

satellite radiant temperature under 47 C.


Accuracy Assessment of Kinetic Temperature Images


The water-body temperature data used to calculate the

atmospheric correction served as a validation data set. The







69

water-body temperature data from other stations served as a

verification data set. Procedures and results of validation

and verification are described below.

Validation of kinetic temperature. Validation assessment

was performed using the previously mentioned near-surface

water temperature data from the three permanent instrument

stations located within Lake Okeechobee (Figure 1). This was

done by comparing the processed image kinetic temperature

values with the corresponding water-body temperature data.

Errors indicated by the validation data set for the AVHRR

kinetic temperature images ranged from 0.0 to 1.1 C; the

average error was 0.5 C. This range of single-pixel basis

validation error for lake surface temperature values is even

lower than expected from the uncertainty associated with the

twin-band emissivity correction method, and indicates that the

station water-temperature (thermistor-based) data were

themselves very well calibrated. Validation details of

individual images are given in Table 140 of Appendix A.

Verification of kinetic temperature. Verification

assessment was performed using near-surface (0.5 m) water

temperature data from the permanent instrument station located

within Lake Apopka, and from periodic sampling by boat in Lake

Sampson and Tampa Bay (Figure 1). The Lake Apopka data were

obtained from the St. Johns River Water Management District

(SJRWMD); the Lake Sampson data were obtained from the

Suwannee River Water Management District (SRWMD); the Tampa

Bay data were obtained from the Environmental Protection







70
Commission of Hillsborough County (EPCHC). The processed

image kinetic temperature value for each of these verification

sites was compared with the corresponding near-surface water

temperature data (thermistor-based values reported to nearest

0.1 C). Errors indicated by the verification data set for the

AVHRR kinetic temperature images ranged from 0.4 to 3.4 C; the

average error was 1.9 C. This range of single-pixel basis

verification error values for atmospheric-corrected lake-

surface temperature is slightly higher than that (0.2 to 0.8

C) reported for atmospheric-corrected sea-surface temperature

under ideal clear-sky conditions (Llewellyn-Jones et al.,

1984), but far lower than would be the case (up to 20 C)

without atmospheric correction. Because the verification

stations are located in a different zone (north) from the

validation stations (south), they provide for each image a

conservative check on the error due to statewide spatial

variation in the atmospheric correction factor. Verification

details for individual images are given in Table 140 of

Appendix A.


Water and Cloud Masking


A masking procedure was used to exclude kilometer scale

water bodies and clouds from the surface temperature analyses

in this study. The mask was prepared from a normalized

difference vegetation index (NDVI) image (Figure 9). While

NDVI is commonly applied to vegetation vigor studies (Fischer,

1994; Teillet and Fedosejevs, 1994; Wade et al., 1994), it


























































Figure 9. Water/cloud mask (NDVI) image.







72

also distinguishes water and cloud surfaces (NOAA, 1990). The

NDVI was calculated from bands 1 and 2 (red and near-infrared)

of the AVHRR image (NOAA, 1990):


NIR R
NDVI = [26]
NIR + R


where NIR is the band 2 DN, and R is the band 1 DN. This NDVI

was scaled for 8-bit storage according to the standard global

vegetation index (GVI) procedure (NOAA, 1990):


255, for NDVI < -0.05

NDVIs = 0, for NDVI > 0.60 [27]

240-350(NDVI+0.05), otherwise


where NDVI, is the scaled NDVI. Values of NDVI, above 219 were

found by inspection to indicate water and cloud surfaces. A

separate NDVI image was prepared for each diurnal image pair,

using the daytime image band 1 and 2 data. This accounted for

any changes in kilometer-scale cloud or water body extent--

such as drifting clouds and floods. Each pixel of a surface

temperature image which corresponded with a water/cloud pixel

was then assigned a value of zero, creating a masked surface

temperature image (Figure 10). When each GIS coverage was

later analyzed using the ERDAS module "BSTATS," the option to

exclude zero values from statistical computations was

selected.























































Figure 10. Daytime (spring) masked surface temperature image.