Group Title: Satellite-derived surface temperatures and their relationships to land cover, land use, soils and physiography of North-Central Florida /
Title: Satellite-derived surface temperatures and their relationships to land cover, land use, soils and physiography of North-Central Florida
CITATION PDF VIEWER THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00099576/00001
 Material Information
Title: Satellite-derived surface temperatures and their relationships to land cover, land use, soils and physiography of North-Central Florida
Physical Description: x, 181 leaves : ill., maps ; 28 cm.
Language: English
Creator: Dicks, Steven E., 1957-
Publication Date: 1986
Copyright Date: 1986
 Subjects
Subject: Temperature   ( lcsh )
Physical geography -- Florida   ( lcsh )
Aerial photography in geomorphology   ( lcsh )
Geography thesis Ph. D
Dissertations, Academic -- Geography -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
 Notes
Thesis: Thesis (Ph. D.)--University of Florida, 1986.
Bibliography: Bibliography: leaves 169-180.
General Note: Typescript.
General Note: Vita.
Statement of Responsibility: Steven E. Dicks.
 Record Information
Bibliographic ID: UF00099576
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000879788
notis - AEH7573
oclc - 014937647

Downloads

This item has the following downloads:

PDF ( 6 MBs ) ( PDF )


Full Text











SATELLITE-DERIVED SURFACE TEMPERATURES AND THEIR
RELATIONSHIPS TO LAND COVER, LAND USE, SOILS
AND PHYSIOGRAPHY OF NORTH-CENTRAL FLORIDA






BY


STEVEN E. DICKS


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


1986














ACKNOWLEDGEMENTS


I wish to thank my supervisory committee, Drs. James A.

Henry, Cesar Caviedes, Robert Marcus and Robert Lindquist,

for their assistance in the completion of my doctoral

program. Thanks must also be extended to Douglas Jordon and

Ray Seyfarth for their assistance in helping me learn and

operate the Earth Resources Laboratory Applications Software

(ELAS) software, Dr. Virginia Hetrick for assistance with

the facilities of the Northeast Regional Data Center, John

Price for supplying his radiative transfer model and helpful

conversations on its use, and Paul Stayert for putting up

with me in the darkroom.

On a more personal note, I need to thank my parents and

family for their support; now they will finally have a

doctor in the family. Last, but certainly not the least, I

must thank Mikel Renner for her assistance the last three

years in getting me through this program.















TABLE OF CONTENTS



PAGE

ACKNOWLEDGEMENTS .... . . . . . . .ii

LIST OF TABLES . .............. . .. .vi

LIST OF FIGURES . . . . . . . ... . . vii

ABSTRACT .... ....... . . .... .viii


CHAPTER

I. INTRODUCTION AND OBJECTIVES . . . . . ... .1

II. PHYSICAL BASIS OF THERMAL REMOTE SENSING . . . 4

Factors Governing Surface Temperatures . . 4
Satellite Measurement of Surface Temperature .. 9
Atmospheric Attenuation and Emission . .. 10
Sensor-Related Inaccuracies ........ 12
Surface Characteristics Influencing
Surface Temperatures . . . .. .13
Canopy structure . . . . . .14
Terrain characteristics . . . .. .17

III. PREVIOUS INVESTIGATIONS USING THERMAL DATA ... .19

Urban Climate and Meteorological Studies . . 19
Agricultural Applications of Thermal Data . 24
Crop Moisture Stress .... . . .25
Evapotranspiration .... . . . .27
Soil Moisture ............... 30
Cold-Prone Area Mapping . . . . ... .32
Geologic Applications of Thermal Data ... .34
Land Cover Mapping .. . . . . . .37

IV. STUDY AREA AND METHODOLOGY . . . . . ... .39

Study Area ...... . . . . .. .39
Soils and Vegetation . . . . ... .41
Physiography .............. .45
Sea Island District . . . . .. .46
Ocala Uplift District . . . . .. .48








Central Lake District . . . ... 51
Eastern Flatwoods District . . . .. .51
Data ... . . . . . . . . 53
Heat Capacity Mapping Mission . . . .. .53
Landsat MSS . . . . . . .. . 55
Meteorological and Ancillary Data . . .. .58
Image Processing and Analysis . . . .. .58
Geometric Correction Procedures . . .. .59
Production of Output for Interpretation . 62
Overlaid Landsat MSS and HCMM images 62
HCMM temperature maps . . . ... 63
Greyscale HCMM IR images . . . ... .66
Image Analysis . . . . . ... .67

V. IMAGE INTERPRETATION RESULTS . . . . . ... .70

Thermal Regions . . . . . . ... 70
Daytime Surface Temperature Patterns . . .. .77
November 5, 1978 . . . . . .77
Meteorological conditions . . ... .77
Gulf Coast Region . . . . .78
Suwannee Agricultural Region . . .. .82
Interlachen Karst Region . . . .. .84
Hastings Agricultural Region . . ... .85
Lake George Region . . .. . . .86
Atlantic Coastal Region . . . ... .87
Central Region ... . . . . .88
December 17, 1978 . . . . . .90
Meteorological conditions . . ... .90
Gulf Coast Region . . . . ... .91
Suwannee Agricultural Region . . .. .92
Interlachen Karst Region . . . .. .96
Hastings Agricultural Region . . ... .97
Lake George Region . . . . .97
Atlantic Coast Region . . . ... .98
Central Region .. . . . ... .99
March 28, 1979 .. . . . . . 101
Meteorological conditions . . .. 101
Gulf Coast Region . . . . ... .102
Suwannee Agricultural Region . . .. .105
Interlachen Karst Region . . . .. .106
Hastings Agricultural Region . . .. .107
Lake George Region . . . ... .107
Atlantic Coastal Region . . ... .108
Central Region . . . . .109
October 7, 1979 .... . . . 110
Meteorological conditions . . .. 110
Gulf Coast Region .. . . . ... .111
Suwannee Agricultural Region . . .. .115
Interlachen Karst Region . . . .. .116
Atlantic Coast Region . . . ... .117
Central Region . . . . . ... 117
Nighttime Surface Temperature Patterns . 119
May 21, 1978 . . ... . . . . 119








Meteorological conditions . . ... .119
Gulf Coast Region . . . . ... .119
Suwannee Agricultural Region . . .124
Interlachen Karst Region . . . .. .125
Hastings Agricultural Region . . .. .126
Lake George Region . . . . ... .126
Atlantic Coast Region . . ... .127
Central Region . . . . . .127
November 3, 1978 ........... . 129
Meteorological conditions . . . 129
Gulf Coast Region . . . . ... .133
Suwannee Agricultural Region . . .. .134
Central Region . . . . . .135
February 1, 1979 ... . . . : 135
Meteorological conditions . . ... .135
Gulf Coast Region . . . . ... .136
Suwannee Agricultural Region . . .. .140
Interlachen Karst Region . . . .. .140
Hastings Agricultural Region . . .. .141
Lake George Region .. . . . . 141
Atlantic Coast Region . . . ... .142
Central Region .. . . . . .143

VI. DISCUSSION AND COMPARISON TO PREVIOUS
INVESTIGATIONS . . .. . . 145

Consistency of Surface Temperature Patterns 145
Surface Temperatures and Natural Land Covers 148
Surface Temperature and Cultural Land Covers 153
Agricultural Land Covers . . . .. .153
Forestry Practices . . . . ... 155
Urban and Industrial Land Covers ... .156
Influence of Soils on Surface Temperatures . 158
Physiography and Surface Temperatures . . 160

VII. SUMMARY AND CONCLUSIONS . . ... . ... 163

Summary and Conclusions . . . . ... .163
Future Research .... . . . . . 166

BIBLIOGRAPHY . . . . . .. . . . .. 169

BIOGRAPHICAL SKETCH ........ .. .. . . 181
















LIST OF TABLES


TABLE

1. North-central Florida soils . . . .

2. Sensor characteristics..... . . .

3. HCMM and Landsat data . . .

4. November 5, 1978 meteorological conditions.

5. December 17, 1978 meteorological conditions.

6. March 28, 1979 meteorological conditions.

7. October 7, 1979 meteorological conditions.

8. May 21, 1978 meteorological conditions. .

9. November 3, 1978 meteorological conditions.

10. February 1, 1979 meteorological conditions.

11. Correlation coefficients... . . . .


PAGE


. . 42

. . 54

. . 57

. . 81

. . 94

. . 104

- . 114

. . 122

. 132

. . 139

. . 146
















LIST OF FIGURES


PAGE


RE

Study area. .. . . . . . .

North-central Florida soils . . . .

Vegetation of north-central Florida . .

Physiographic divisions of north-central
Florida. .... .. . . .

November 5, 1978 Landsat MSS TVI image.

Thermal Regions . . . .

November 5, 1978 HCMM IR image. . . .

December 17, 1978 HCMM IR image. . . .

March 28, 1979 HCMM IR image. . . .

October 7, 1979 HCMM IR image . . .

May 21, 1978 HCMM IR image. . . . .

November 3, 1978 HCMM IR image. . . .

February 1, 1979 HCMM IR image. . . .


FIGUF

1.

2.

3.

4.


5.

6.

7.

8.

9.

10.

11.

12.

13.


. . . 40

. . . 43

. . . 44



. . . 47

. . . 64

. . . 73

. . . 80

. . . 93

. . 103

. . 113

. . 121

. . 131

. . 138














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



SATELLITE-DERIVED SURFACE TEMPERATURES AND THEIR
RELATIONSHIPS TO LAND COVER, LAND USE, SOILS
AND PHYSIOGRAPHY OF NORTH-CENTRAL FLORIDA


By


STEVEN E. DICKS


May 1986


Chairman: James A. Henry
Major Department: Geography



Relationships between satellite-derived surface

temperatures and surface materials in north-central Florida

were examined using subsets of four day and three night Heat

Capacity Mapping Mission (HCMM) thermal infrared scenes.

The images were mapped to the Universal Transverse Mercator

coordinate system and corrected for atmospheric attenuation

and thermal emission. The images were selected to provide

day and night coverage for the fall, winter and spring

seasons. From these data, overlayed HCMM and Landsat

multispectral scanner (MSS) images, isotherm maps, color-

coded temperature maps and enhanced greyscale images were


viii








produced and used in conjuction with black and white and

color infrared air photos, Landsat MSS and Thematic Mapper

images, and land use, soils and physiographic maps to

examine the influence of surface features on surface

temperatures.

These data show that surface temperatures are largely

controlled by a combination of surface moisture availability

and vegetation density. Natural land covers show a daytime

temperature gradation reflecting the importance of these

factors, with wettest land covers such as marshes and

wetlands displaying the coolest temperatures and sparsely

vegetated forests on well drained soils registering the

highest surface temperatures. The nighttime data reveal

that land cover has less of an influence on surface

temperatures, with surface moisture being the primary factor

determining temperatures.

Modifications of natural land cover exert a significant

influence on surface temperature patterns. Agricultural

lands, because of their areal extent and high surface

temperatures, have the most profound influence on

temperature patterns. Commercial forestry practices,

particularly clearcutting, lead to increased day surface

temperatures, as do urbanization and surface mining. Night

surface temperatures are less influenced by cultural land

covers. Bare agricultural lands and clearcuts generally

exhibit the lowest surface temperatures, and urban areas

tend to have a localized warming influence.








Soil drainage classes show a strong correspondence to

surface temperatures, particularly in the daytime data.

Well drained soils exhibit the highest day and lowest night

surface temperatures. The results also indicate that the

sensitivity of thermal data to moisture, vegetation cover

and soil drainage characteristics make it a potentially

useful data source for coastal plain physiographic studies.














CHAPTER I
INTRODUCTION AND OBJECTIVES


All objects having temperatures above absolute zero emit

electromagnetic radiation- (EMR) at a rate determined by the

temperature and emissivity of the object. This emitted EMR

offers a valuable source of information that can be remotely

sensed and used to measure or infer surface properties not

easily obtained from other data sources. Of particular

interest to researchers in several disciplines is EMR

emitted in the thermal infrared (IR) wavelengths.

Geographers and climatologists have used IR data to examine

energy balances over urban areas to better understand the

influence of urbanization on climate. Measurement of

temperature variations determined from IR data have shown

great potential for improving crop management practices such

as irrigation and pesticide application by allowing the

identification of stressed vegetation and regions of low

soil moisture. IR data have also been employed by

agricultural meteorologists for frost forecasting. The

ability to estimate regional evapotranspiration (ET) rates

using remotely sensed surface temperatures has potential

benefits to agronomists, soil scientists and hydrologists.

Geologists have found that thermal data are complementary to





2

the visible and reflective infrared data that have been in

use since the beginnings of aerial photography. Geothermal

mapping, rock type differentiation, structural mapping and

physiographic mapping all potentially benefit from the use

of IR data.

This study utilizes-IR data obtained by the Heat Capacity

Mapping Mission (HCMM) satellite to examine a study area in

north-central Florida. The primary purpose of this study is

to examine relationships between satellite-observed

temperature patterns and surface features in the study area,

and variations in these relationships through time.

To date, temperature pattern studies of the Florida

peninsula have concentrated on the central and southern

portions of the state and were intended to aid in detecting

cold-prone agricultural regions (Allen et al., 1983; Chen et

al., 1979; 1982; 1983; Shih and Chen, 1984). Most of these

studies utilized coarse resolution geostationary satellite

(GOES) data for nocturnal winter time periods. This study

extends previous investigations to the northern part of the

Florida peninsula, using higher resolution data and day and

night imagery from various times of the year. The results

of this investigation will provide an increased

understanding of the influence of land cover and land use

patterns on the distribution of surface temperatures in

north Florida. This information will be helpful in

evaluating the potential of IR data for mapping land cover,





3

physiography and soils in the region. It will also provide

insight into inadvertent climatological changes produced by

the alteration of the natural landscape by human activities.

The following aspects of surface temperature patterns

will be examined in detail:

Relationships between vegetative land covers and

surface temperature patterns. Of particular interest

is the ability to differentiate between various land

covers on the basis of surface temperatures.

Modification of surface temperature patterns by human

activities. Specifically examined are modifications

related to urban and industrial land uses and

agricultural and forestry practices.

Relationships between satellite-derived temperatures

and soil type and physiographic features.














CHAPTER II
PHYSICAL BASIS OF THERMAL REMOTE SENSING


The temperature measured by hand-held, aircraft or

satellite thermal radiometers is the apparent radiant

temperature of the object. The apparent radiant temperature

differs from the kinetic temperature, or that measured by a

thermometer, by its dependence on a number of factors,

including kinetic surface temperature, atmospheric

conditions, sensor characteristics and the nature of the

earth's surface that is being examined. These factors will

be briefly discussed below.



Factors Governing Surface Temperatures

The temperature of the earth's surface is governed by

balances between incoming and outgoing radiative fluxes

(Rosenberg, 1974; Oke, 1978; Price, 1982). These fluxes can

be described by the following energy balance equation:


Rn = H + LE + G


where Rn is the net radiative flux at the earth's surface, H

is sensible heat flux, LE is the latent heat flux, and G is

the heat flux into the ground. Net radiative flux is

dependent on the incoming and outgoing shortwave and






5

longwave radiation balances. The shortwave balance can be

represented by the following equation:


Rswb = Rsw (1 a),


and the longwave balance by


Rlwb = Rlwd Rlwu


where the Rswb is the shortwave radiation balance, Rsw is

incoming shortwave (both solar and diffuse) radiation, a is

shortwave albedo, Rlwb is the longwave radiation balance,

and Rlwd and Rlwu are the downward and upward longwave

radiation fluxes. These equations can be combined to give

Rn:


Rn = (1-a) Rsw + Rlwb.



The right side of the energy balance equation contains

the sensible, latent and soil heat flux parameters.

Sensible heat flux at the surface is governed by transfer

first through the laminar boundary layer via molecular

conduction, and then through the overlying surface layer via

turbulent transfer (Oke, 1978). Transfer through the

laminar boundary layer is described by the following

equation:


H = -pc(Mh)dT/dz,


and transfer through the turbulent surface layer by








H = -pc(Kh)(dT/dz + DALR)


where p is the density of air (gm/cm3), c is the specific

heat of air at constant temperature and pressure (J/Kg/K),

dT/dz is the temperature gradient (K/cm), Mh is molecular

diffusivity of air (approximately 0.21 m2/s), Kh is the eddy

conductivity (m2/sec), and DALR is the dry adiabatic lapse

rate.

Transfers of sensible heat can also be described using a

resistance approach analogous to Ohm's Law (Rosenberg,

1974):


H = -60pc (Ts Ta)/ra


where p and d are the density and specific heat of air, Ts

and Ta are the surface and air temperatures and ra is the

atmospheric resistance factor. Resistance will increase

with increasing windspeed or turbulence as the height of the

laminar boundary layer is decreased and mixing in the

overlying turbulent layer increases.

The second term on the right side of the energy balance

equation is latent heat transfer. Latent heat fluxes are

the result of evaporation (EV) and transpiration (TR) with

evaporation being described by


EV = -pL(Kw)dq/dz


where p is as defined above, L is the latent heat of

vaporization (J/kg), Kw is the eddy diffusivity for water








vapor (mi/s) and dq/dz is the vapor pressure gradient.

Latent heat exchanges from transpiration can be estimated

using a resistance approach:


TR = -pL(0.622)/P((es-ea)/(ra+rs))


where p and L are as defined above, P is photosynthetic

fixation of energy (W/cm2), es and ea are stomatal and

atmospheric vapor pressures and ra and rs are atmospheric

and stomatal resistances. Atmospheric resistance depends on

windspeed as described above. Because water vapor from

plants must pass through leaf stomata, the rs term depends

on stomatal characteristics.

The last term in the radiation balance equation is ground

heat flux, and is a function of the mean temperature

gradient and the ability of a soil to transmit heat:


G = -k (dT/dz)avg


where k is thermal conductivity (W/m/K), and (dT/dz)avg is

the mean temperature gradient. The value of k varies with

both depth and time for a given soil, although if bulk

averages are required, k varies with soil particle

conductivities, soil porosity and soil moisture. The

dependence on porosity and soil moisture is the result of

differences in the thermal properties of air and water in

pore spaces. Temperature change resulting from the addition

of heat to a volume of material is related to the volumetric







heat capacity (C) (J/m3/Kelvin). Heat capacity is the

amount of heat necessary to raise the temperature of a cubic

meter of a material by 1 K. The total temperature response

of a soil is given by the ratio between thermal conductivity

and volumetric heat capacity, and is termed thermal

diffusivity (m2/s). Thermal diffusivity gives the amount of

time required for temperature changes to travel through a

soil. Thermal diffusivity is increased by the initial

addition- of moisture to a soil because of increased thermal

contact between particles and the expelling of air (a poor

conductor of heat). Beyond an approximately 20% soil

moisture content (by volume) thermal diffusivity declines as

thermal conductivity levels off and heat capacity continues

to rise (Oke, 1978).

The above thermal characteristics determine the diurnal

temperature fluctuations of soil, and can be combined to

provide another thermal descriptor of the earth's surface,

thermal inertia. Thermal inertia is defined as follows:


P = SQRT(Kcp)


where K is thermal diffusivity, c is specific heat and p is

density. Thermal inertia is the resistance of a material to

temperature change; materials with large thermal inertias,

such as wet soils or water bodies, will display smaller

diurnal temperature ranges than will those with small

thermal inertias.








Satellite Measurement of Surface Temperature

As previously mentioned, all materials with a temperature

above absolute zero emit EMR. The amount of EMR emitted by

a perfect emitter, or blackbody, is given by Planck's

equation:


L = C1(W) S/(exp(C2/WT) 1)


where L is the spectral exitance for a given wavelength

(W/m2/micron), Cl is 3.74 X 10 16 W/m2, W is the wavelength

of the EM energy in meters, C2 is 1.44 X 102 m K and T is

temperature in degrees K. This formula shows that the

amount of energy emitted by a perfect blackbody is

temperature and wavelength dependent. The maximum

wavelength (W) at which energy is emitted is given by the

Wien Displacement Law:


W = C/T


where C is 2.989 X 10 3 m K, and T is temperature in degrees

Kelvin. Total exitance of a blackbody is given by the

Stefan-Boltzmann Law:


M(total) = a T4


where a is 5.669 X 10 8 W/m2/K. Because natural objects are

not perfect blackbodies the above equations must be

corrected for the varying emissivity characteristics of

different materials. The emissivity of a material is given





10

by the ratio between the spectral exitance of the material

and a true blackbody. The closer the emissivity is to 1.0,

the closer the apparent radiant temperature of an object

will be to the true kinetic temperature.

In most thermal remote sensing applications, the quantity

measured by the sensor is spectral radiance, and the Planck

equation is solved for temperature. The temperature derived

from the inverted Planck equation, when not corrected for

emissivity or atmospheric attenuation, is termed the

apparent temperature. Correcting for emissivity variations

is difficult when aircraft- or satellite-based sensors are

used because a wide variety of surface features with varying

emissivities contribute to the measured radiance. Because

most vegetated and urban surfaces have emissivities greater

than 0.95, emissivity variations will be relatively minor

over such land covers (Taylor, 1979; Artis and Carnahan,

1982). For sensors with large instantaneous fields of view

the radiance is an average of several emitters on the

earth's surface, a fact that tends to reduce the

significance of emissivity differences.



Atmospheric Attenuation and Emission

If emissivity variations can be assumed to be equal,

relative temperature differences can be readily obtained by

thermal sensors, but atmospheric attenuation and emission of

thermal infrared EM radiation make the determination of








actual surface temperatures difficult, and can lead to

greater errors than do emissivity variations. Attenuation

of infrared EM radiation is wavelength dependent and

primarily caused by line absorption by water vapor, carbon

dioxide and ozone, and by continuum absorption between

spectral lines by carbon dioxide and water vapor (Chahine,

1983). Most sensors are designed to utilize wavelengths

that avoid the main spectral absorption lines. Continuum

absorption varies mainly in response to changes in

atmospheric moisture, with the carbon dioxide component

being almost constant spatially and temporally. Thermal

emission is related to the amount of water vapor present and

the temperature profile of the atmosphere.

Radiation received by the satellite can be determined by

use of the radiative transfer equation. There is a linear

relationship between radiation emitted at the surface and

radiation received at the satellite (Schott and Volchok,

1985):


L(h) = t(h)eL(T) + t(h)rLd + Lu(h)


where L(h) is the radiance reaching the sensor at altitude h

(W/cm2/sr), t(h) is the transmission to altitude h, e is the

emissivity of the surface, L(T) is the blackbody radiance

associated with an object on the ground at temperature T

(W/cm2/sr), r is the reflectivity of the surface (r = 1-e),

Ld is the downwelling sky radiance (W/cm2/sr) and Lu(h) is








the upwelled radiance from the air column between the

surface and the sensor at altitude h (W/cm2/sr). These

values are integrated over the range of wavelengths

measured. Corrections for atmospheric effects are generally

accomplished using a numerical approximation of the

radiative transfer equation, with atmospheric soundings

providing the necessary temperature and humidity data

(Schott, 1979; Leckie, 1982; Price, 1983b; Suits, 1983;

Vukovich 1983; 1984).



Sensor-Related Inaccuracies

Several types of satellite-based imaging sensors are used

to measure IR. Commonly they utilize an oscillating mirror

to direct incoming radiation to a detector that converts

incoming watts of EMR to output voltages. A well-designed

detector has a linear or near-linear calibration curve

relating incoming radiation to output volts. The voltages

are then sampled during ground processing to integer values

that can be related to specific measured radiances or

temperatures.

Modern sensors have stated accuracies of approximately

0.1 K, though noise from the sensor itself and associated

electronics can degrade this accuracy somewhat. A common

measure of the signal-to-noise ratio of a sensor is noise

equivalent temperature difference (NETD). NETD is a measure

of the temperature change required to produce a voltage





13

change equal to the noise, and in a well designed sensor it

should be constant over the full range of possible

temperatures (Short and Stuart, 1982). Inaccuracies in

measured temperatures or radiances may be caused by post-

launch changes in sensor calibration. These calibration

changes may be the result of sensor contamination or

degradation over time. It may be possible to correct for

calibration changes using data collected by the sensor or by

comparing satellite temperatures to those collected on the

ground, but these changes are often poorly understood and

accurately compensating for them is difficult (Barnes and

Price, 1980).

Though not an inaccuracy per se, spatial resolution is

also important in determining the temperature measured by a

radiometer. The larger the field of view of the sensor, the

more complex will be the mixture of objects viewed by the

instrument. The result is that small features with very

high or low temperatures may be masked by the temperature of

surrounding features because of the averaging effect of the

sensor.



Surface Characteristics Influencing Surface Temperatures

As discussed above, the radiant temperature measured by

thermal sensors is a function of the kinetic temperature and

emissivity of a surface, atmospheric temperature and

humidity characteristics, sensor calibration and surface





14

energy fluxes. The radiant temperature measured by a sensor

is the composite surface temperature of all objects within

the sensor's field of view, and therefore represents a

complex mix of surface features. Vegetation canopy

structure and terrain characteristics exert important

influences on remotely-sensed surface temperatures.


Canopy structure. Canopy structure can be described in

terms of vegetation geometry, leaf area and vegetation

distribution and density. The structure of natural

vegetation canopies is generally complex, and radiation

budgets over vegetated areas are likewise more complex than

those over non-vegetated surfaces. As incoming solar

radiation is reflected, transmitted and absorbed by multiple

leaf layers, trunks, stems, ground litter, bare soil and

undergrowth, it is changed in spectral content and amount.

Reflectance of EMR is controlled by a number of factors

including leaf area, shape and orientation, proportion of

diffuse versus direct sunlight and solar illumination angle

(Ahmad and Lockwood, 1979). Reflectances are spectrally

variable over vegetated surfaces. In the visible

wavelengths blue and red light tend to be absorbed by

chlorophylls and caratenoids more than is green light. Over

the entire visible spectrum absorption is generally between

80% and 90%. Between 0.7 and 1.0 microns, reflectance is

generally 40-45% for most plants. Beyond the visible range

the reflectance is controlled largely by water content, with








absorption increasing with increasing leaf moisture.

Average reflectances for the entire EMR spectrum are on the

order of 12-18% for forests and 17-26% for croplands

(Monteith, 1973).

Where soils are visible through the vegetation

reflectance is also dependent on soil spectral

characteristics. Maximum reflectances are found over

smooth, fine-grained, light-colored soils. Most dry soils

exhibit a smooth, gradually increasing reflectance from the

visible to reflective infrared wavelengths (Swain and Davis,

1978). Changes in spectral characteristics are primarily

related to organic content, mineral composition and soil

moisture. Reflectivity, particularly in the IR, will

usually decrease with increases in these factors.

Longwave radiation is also absorbed, reemitted and

reabsorbed within the canopy. Sutherland and Bartholic

(1976) examined the effect of crop geometry on emissivities

for crops. Crop geometry affects the total emissivity of

areas sensed by aircraft or satellite radiometers because of

differing amounts of soil and vegetation in the field of

view of the sensor. The authors found that total emissivity

was insensitive to crop geometry when crop height-to-spacing

ratios were greater than 1, and therefore errors in

radiometrically measured temperatures were small.

Kimes (1980; 1983) and Kimes et al. (1980) found that for

row crops, the amount of soil visible through vegetation






16

influenced the measurement of canopy temperatures. Because

soil temperatures are often much higher than vegetation

temperatures, a composite temperature of an area may not

accurately reflect true canopy temperature. Over a soybean

canopy with a 35% ground cover, it was found that composite

nadir temperatures differed by as much as 11 C from canopy

temperatures. Ground temperatures for the same canopy were

as much as 15 C higher than air temperatures (Kimes, 1980).

High soil temperatures also contribute to steep vertical

temperature gradients within the canopy, particularly when

wind speeds are low (Kimes et al., 1980). These steep

temperature gradients over bare soil occur because incoming

solar radiation raises soil temperatures significantly over

those of the overlying air.

The influence of irregular tree canopies on radiometer-

derived imagery has been examined by Balick and Wilson

(1980) and Fritschen et al. (1982). Balick and Wilson

utilized high resolution imagery and temperature profile

data for Alamos Canyon, New Mexico, to examine night canopy

temperatures. Most notable in the imagery was that trees

displayed significantly higher temperatures than the

underlying bare ground. It was found that tree-crown

temperature remained very close to that of the surrounding

air, while being 3 to 5 C warmer than the ground. This is

the inverse of the daytime situation of ground temperatures

being warmer than canopy temperatures.








Terrain characteristics. Terrain characteristics

important in determining surface temperatures include

elevation, slope and proximity to large water bodies.

Elevation and slope are important in controlling cold air

drainage, particularly under stable winter nocturnal

conditions. The relationship between decreasing temperature

and increasing elevation in mountainous regions is well

known, and must be corrected for in thermal modelling of ET

or thermal inertia (Price, 1983a; 1985; Kahle, et. al,

1984).

Fritschen et al. (1982) investigated temperature

characteristics of a forested valley in Washington using

nighttime infrared imagery and found that forest structure

and elevation as well as local meteorological conditions

were responsible for observed temperature patterns. As in

Balick and Wilson (1980), tree tops were found to be warmer

than the underlying ground. In the Washington study cold

air drainage was found to be the controlling factor in tree-

crown temperature variations. Near the center of the valley

where the cold air layer was thicker only tree tops

protruded through the cold air and appeared warmer than the

surrounding area. In elevated areas such as on small hills

or valley slopes more of the tree extends above the cold air

layer and is evident on the imagery. Additionally, areas

with older, taller trees were highlighted on the imagery

because they too were able to protrude through the cold air

layer.





18

Mahrt and Heald (1983) found that terrain features were

important in controlling surface temperatures even in areas

of low relief. Using aircraft IR data over agricultural

areas in eastern Colorado and western Oklahoma they found

that terrain curvature as well as elevation was important in

controlling surface temperature. It was suggested that

large concave curvatures lead to more effective trapping of

cold air than do convex curvatures. The effect of terrain

variables on temperature was significant even during periods

of relatively high windspeeds.

Location relative to water bodies is also important in

controlling surface temperatures. The moderating influence

of water bodies on temperatures is well known, and often

taken advantage of by farmers in frost-prone areas. This

moderating influence is caused by the large heat capacity of

water bodies. Advection of moisture and heat downwind of

large water bodies is important in influencing temperature

patterns, particularly under cold weather conditions. Bill

et al. (1977) examined the moderating influence of Lake

Apopka, Florida on downwind surface temperatures under

winter nocturnal conditions. With windspeeds less than 1

m/s no downwind influences were observed. Windspeeds on the

order of 4 m/s lead to temperature increases as much as 5 C

over surrounding areas.














CHAPTER III
PREVIOUS INVESTIGATIONS USING THERMAL DATA



Urban Climate and Meteorological Studies

Growing recognition of the influence of urban areas on

climate and meteorology, particularly on the radiation

balance of urbanized areas, has led to the use of remotely

sensed thermal data for studying urban climate. Remotely

sensed information on reflected shortwave and emitted

longwave radiation offers a number of advantages over ground

level measurements of these phenomena. These advantages

include the synoptic view and coverage of large areas by

aircraft and satellite sensors as well as the spatial

integration of surface temperatures within the sensor's

field of view. Also valuable is the ability to use remotely

sensed IR data along with meteorological data to calculate

various surface characteristics such as thermal inertia,

moisture availability and sensible heat loss.

The heat islands of a number of large cities have been

examined using aircraft and satellite thermal data. Matson

et al. (1978) utilized NOAA 5 satellite data (1.1 km

resolution) to examine heat islands over fifty cities in the

eastern United States. The authors found urban-rural

temperature differences ranging from 2.6 to 6.5 C. A






20

detailed examination of St. Louis revealed that the highest

temperatures were experienced in areas in which the greatest

building density and industrialization were found.

Temperatures were up to 3.1 C higher in these areas than in

the surrounding rural areas. Examination of Baltimore and

Washington, D.C., showed that core areas displayed the

highest temperatures, with differences from surrounding

rural areas being as high as 5.2 C. Several urban heat

corridors corresponding to major traffic arteries were also

visible in the area.

Price (1979) used higher resolution HCMM (0.6 km) data to

detect urban-rural daytime temperature differences in the

northeastern United States ranging from 16.5 C for New York

City to 6.8 C for Montpelier, Vermont. The fact that these

temperatures are much higher than those found by Matson et

al. (1978) is partially because the HCMM imaging time is

closer to the period of the day when surface heating is at a

maximum, while the NOAA 5 sensor acquired mid-morning and

early evening imagery when temperature differences are not

at their peak.

More detailed investigations of urban heat islands using

remotely sensed data have been carried out for the cities of

Baltimore, Maryland, Los Angeles, California, and St. Louis,

Missouri. Pease et al. (1976) utilized high resolution

aircraft IR scanner data to study the urban area of

Baltimore. The authors used isoline maps of surface albedo,








emitted IR energy, absorbed energy, net radiation and

temperature of the urban area along with numerical modeling

to determine the relative importance of surface moisture,

surface material thermal properties, surface aerodynamic

roughness and albedo in controlling the urban heat island.

Model results indicated that surface wetness variations in

the summer and absorption of radiation by vertical surfaces

in the winter were most responsible for modelled urban

temperature variations. Both modelling and remotely sensed

temperatures indicated that the urban heat island does not

display a monolithic nature, but instead consists of several

relatively warm and cool areas throughout the city.

Examination of the imagery showed that during the day

vegetated areas within the city were the coolest areas

(except for water surfaces), while commercial and industrial

land exhibited much higher temperatures. The higher

reflectivity of vegetated areas along with increased latent

heat exchanges appeared to be responsible for the lower

temperatures observed in these parts of the city.

Residential areas, because of their higher reflectance, also

were generally cooler than heavily built-up areas.

Carlson et al. (1981) examined the cities of Los Angeles

and St. Louis, using HCMM data as input into numerical

models to study the relationship between urban temperature

distributions and surface energy balance, moisture

availability and thermal inertia. As in other studies,





22

highest day temperatures were found over the most heavily

industrialized or commercialized areas where vegetation

cover was minimal. Vegetated areas, mostly because of ET

potential, were the coolest areas (excepting water bodies)

within the Los Angeles and St. Louis urban areas. Nighttime

temperatures in Los Angeles displayed less contrast than did

day temperatures. Areas close to the Pacific Ocean were

generally cooler at night than areas 10-15 km inland, and a

weak correspondence was found between areas that were warm

in the day and those that were warm at night.

In both St. Louis and Los Angeles a strong correspondence

between the moisture availability and temperature pattern

was observed. Areas of high moisture availability were

generally cooler during the day because more of the energy

budget was partitioned as latent rather than sensible heat

flux, resulting in lower radiant temperatures. The authors

noted a striking lack of detail in the thermal inertia maps

for St. Louis and Los Angeles. Thermal inertia has often

been considered as one of the more important thermal

properties responsible for the distribution of the nocturnal

urban heat island. That this does not appear to be

important in influencing Los Angeles and St. Louis heat

island patterns appears to follow current thinking on the

causes behind the formation of urban heat islands (Goward,

1981). Because thermal inertia did not appear to be

important, Carlson et al. considered moisture availability








to be the most important factor in determining the

distribution of urban temperature patterns.

On the basis of findings of Willis and Deardorff (1978)

and Lamb (1978), Carlson and DiCristofaro (1981) proposed

that satellite-derived heat flux maps be used to aid in

modelling plume dispersion from industrial smokestacks.

Willis and Deardorff (1978) and Lamb (1978) found that plume

dispersal was related to boundary layer turbulence that in

turn is related to surface heat flux. They found that the

greater the turbulence, the more a plume is dispersed both

horizontally and vertically, producing a smaller

concentration downwind of the source. Carslon and

DiCristofaro suggested that heat flux maps calculated from

remotely sensed data would provide a good source of data for

predicting plume concentrations and dispersal over a terrain

with a complex heat flux pattern, as is normally found.

Vukovich (1983) utilized HCMM data to examine temperature

and reflectivity patterns over St. Louis. As found

elsewhere, the lowest temperatures were over water bodies,

in the case of St. Louis, along the Mississippi River and

Horseshoe Lake. Low temperatures were also evident in

vegetated areas because of increased evapotranspiration over

these surfaces. Highest temperatures were found in heavily

built-up commercial or industrial districts, areas that also

displayed low reflectances. It was also noted that day

surface ground temperatures measured by the HCMM satellite








were more readily influenced by small scale land use

patterns than were nighttime temperatures.



Agricultural Applications of Thermal Data

Satellite-derived thermal data have been utilized for a

number of non-urban studies, particularly in agricultural

regions. Thermal data have shown potential for use in

mapping vegetation stress, soil moisture, ET and frost

potential in agricultural regions. Important in most

agricultural applications of IR data is the influence of

surface and subsurface moisture on radiant temperature

patterns. Water affects surface temperatures in two ways.

First, moisture changes the thermal properties of a soil

(ie., heat capacity, thermal diffusivity, thermal inertia,

and thermal conductivity). The changing of thermal

properties of a material can be particularly influential in

determining nighttime temperature patterns. For example, at

night, wet soils will usually be warmer than their drier

counterparts because of their higher thermal inertia.

Second, surface water availability controls the amount of ET

occurring at the surface. Over vegetated surfaces there is

greater potential for increased ET when soil moisture is

high, resulting in decreased temperatures as more energy is

used in latent heat exchanges as opposed to radiative or

sensible transfers.








Crop Moisture Stress

One of the most promising uses of thermal data is for the

detection of moisture stress in crops. The availability of

moisture to vegetation has been shown to be important in

determining the surface temperature of crop canopies. The

temperature of a plant leaf is controlled largely by latent

heat releases during transpiration. As water availability

decreases, leaves lose their turgidity and begin to wilt.

To avoid dehydration, leaf stomata begin to close either

partially or fully. As this closure of the stomata occurs,

evaporative water losses decrease, as do latent heat

exchanges, and sensible heat exchanges become more important

in controlling leaf temperatures (Byrne et al., 1979;

Sumayao et al., 1980; Keener and Kircher, 1983). The result

is higher leaf temperatures for vegetation undergoing

moisture depletion.

Pinter et al. (1979) established that midday radiant leaf

temperatures of diseased cotton and sugarbeet plants were 3

to 5 C higher than those of adjacent healthy plants.

Temperature increases were related to a root-rot disease

that affected water uptake and therefore transpiration

rates. These temperature differences were observable over a

wide variety of soil moisture conditions, even when plants

were wilting. Gardner et al. (1981) examined crop

temperatures and their relationships to plant phenology and

yield for a differentially irrigated corn crop. It was





26

revealed that optimal yield decreases with moisture stress,

and therefore canopy temperatures could be used to aid in

predicting crop yields. Additionally it was noted that

after crop cover is complete crop temperature data could be

used to monitor phenological development throughout the

growing season.

Idso et al. (1977) outlined the potential for the use of

thermal remote sensing for agricultural water management,

soil moisture surveillance, evaporation measurement, crop

yield prediction and irrigation scheduling. Soil moisture

was found to be correlated with midafternoon-presunrise

canopy temperature differences for a number of different

crops. An empirically-based equation for calculating

24-hour evaporation using incoming and outgoing thermal

radiation was developed and found to be useful for a wide

variety of soil and crop types. Using a combination of a

moisture stress index termed the stress degree day (SSD)

along with the standard growing degree day (GDD), it was

discovered that grain yields could be predicted for a

variety of moisture stress conditions. The combination of

the SSD and GDD measurements could also be used to aid in

efficient irrigation scheduling.

Paloscia and Pampaloni (1984) examined the use of surface

temperatures of corn and wheat crops derived from

measurements of emitted microwave EMR. The authors found

that microwave-derived temperatures could be used in two








ways to evaluate crop moisture stress. The first method

used a normalized microwave radiometer-measured temperature

along with the air vapor pressure deficit to calculate a

moisture stress index. The second utilized the difference

between vertical and horizontal components of the emitted

microwave radiation to define a polarization index. Both

methods proved to be sensitive to moisture stress in crops.



Evapotranspiration

The measurement of ET rates is important in water budget

studies, particularly for irrigated cropland. Because of

the difficulty in obtaining regional estimates of ET rates

using standard in situ methods, there have been several

attempts to apply remotely sensed thermal data to this

problem.

Heilman et al. (1976) used crop temperatures obtained

from aircraft scanner IR data in energy balance equations to

estimate ET rates over soybean, sorghum and millet crops.

Latent heat exchanges derived from the aircraft data were

between 62.5% and -43.6% of lysimeter-derived measurements.

A primary cause of differences between modeled and

lysimeter-derived measurements was in errors in the

measurement of remotely-sensed surface temperature because

of atmospheric effects.

Soer (1980) used aircraft IR scanner and meteorological

data as input into energy balance and aerodynamic equations





28

to estimate regional ET over grasslands in the Netherlands.

Momentary ET calculated from satellite data and

meteorological parameters was compared with modelled 24-hour

ET to relate daily to momentary ET over the study area.

Estimates of ET obtained using remote sensing techniques

were within 30% of measurements made using water balance

estimates in the study area. The accuracy of calculating ET

was mainly dependent on obtaining accurate crop surface

temperatures and emissivities, and on using accurate surface

roughness coefficients in the aerodynamic equations.

Price (1980; 1983a) discussed the physical basis behind

the modelling of ET using HCMM data and derived equations

for obtaining a 24-hour average using day-night paired

temperature images and meteorological data. Flux rates

calculated using HCMM data covering a southwest Idaho study

area produced results comparable to those obtained by

numerical simulations. It was established that farmed areas

exhibited a high variability in evaporative flux, most

likely because of the mixture of irrigated and non-irrigated

land in the region.

Reginato et al. (1985) demonstrated that in areas where

screen-height air temperature, incoming solar radiation,

windspeed and vapor pressure data are collected, it should

be possible to calculate ET rates on a field-by-field basis

using aircraft scanner data. Comparisons of the remotely

sensed ET measurements with those obtained using lysimeters








showed a high correlation (r=0.9) between the two methods.

The ability to extend ET measurements over large areas was

found to be dependent on clear sky conditions and the

ability to accurately extrapolate wind speed and air

temperature measurements beyond the locality of the

meteorological station.

Klaassen and van den Berg (1985) used an energy balance

method for calculating ET over grasslands in the Netherlands

using NOAA Advanced Very High Resolution Radiometer (AVHRR)

data. A split-window technique to correct for atmospheric

attenuation that is commonly used to derive sea-surface

temperatures was used to calculate radiant surface

temperatures. ET measurement inaccuracies were found to be

related mainly to inaccurate windspeed and air temperature

measurements. To avoid these problems air temperature and

wind speed at the 50 m level were modeled. This provided

better predictions of mesoscale windspeed and temperature by

avoiding the near-surface atmospheric layer where complex

surface energy and momentum fluxes dominate. The models

allowed calculations of ET that were on the average of 7

W/m2 less than surface observations and with RMS errors of

34 W/m2, a result that is within the range of surface

measurement inaccuracies of ET. These results indicate that

satellite-derived ET fluxes can be used not only to find ET

differences but to calculate actual ET rates.








Soil Moisture

Differences in the thermal characteristics of wet versus

dry soils provide a means for mapping soil moisture using

temperatures obtained from passive microwave or IR sensors

(Idso et al., 1975; Schmugge, 1978). The large heat

capacity and thermal conductivity of water mean that moist

soils will have higher thermal inertias than will dry soils.

These differences are detectable by the use of remotely

sensed diurnal temperature measurements, particularly when

there is minimal vegetative cover (Price, 1977; Kahle, 1977;

Pratt and Ellyett, 1979; Price, 1985).

Schmugge et al., (1977) discovered that coarse resolution

(25 km) Nimbus-5 microwave radiometer data are sensitive to

near-surface soil moisture. Examination of agricultural

regions in Illinois, Indiana, Texas and Oklahoma showed that

an inverse relationship exists between soil moisture

expressed as percent field capacity and satellite brightness

temperature. These relationships hold true only under

conditions of minimal ground cover, because vegetation

absorbs most of the emitted microwave EMR.

Heilman and Moore (1980) examined-relationships between

surface temperatures derived from hand-held and aircraft

radiometers and soil moisture. Soil water content

correlated highly (r=0.9) with day-night temperature

differences measured using a hand-held radiometer with

percent covers from 30-90%. Aircraft scanner data were used





31

with equations derived from hand-held radiometers to test

the potential for inferring soil moisture over wider areas.

Maximum differences between predicted and observed soil

moisture (measured as percent of field capacity rather than

soil water content) were -24.5% and 5.3%, with the average

being 1.6% for a wide variety of soil types. Maximum errors

occurred where high-percent ground covers existed because

soil temperature measurements were less accurate over these

areas.

Heilman and Moore (1982a; 1982b) used HCMM data to

estimate soil moisture and depth to groundwater in

southeastern South Dakota. Surface temperatures, after

correction for variations in percent vegetative cover,

correlated well with percent of field capacity.

Correlations between HCMM-derived surface temperatures and

depth to groundwater were made using 5 dates from June to

September, with highest correlations found using the

September data. It was noted that, since correlations

between surface temperature and both groundwater and soil

moisture were high, it was not possible to separate the

influences of the two different factors.

The thermal structure of an agricultural region in the

Beauce Plateau in France was examined by Cheevasuvit et al.

(1985) using NOAA polar orbiter thermal data. In this

region the thermal structure was divided into at least three

distinct sections. Regions of homogeneous temperature








correlated well with regions of equal soil water status,

indicating that the thermal structure of a region can

provide an indication of moisture status on a regional

scale.



Cold-Prone Area Mapping

Satellite-derived IR data have been used to examine

nocturnal temperature patterns over agricultural regions to

determine areas that are particularly prone to cold weather

damage. Chen et al. (1983) established that under clear

skies, winter nocturnal temperatures measured by GOES were

highly correlated (r>0.80) with shelter-height temperatures.

Maximum differences between satellite-derived and shelter-

height temperatures were greatest in the early evening and

decreased as the night progressed, with satellite

temperatures being consistently lower.

Chen et al. (1979) examined nocturnal temperature

patterns over the Everglades agricultural area south of Lake

Okeechobee in Florida using GOES IR data. Satellite

temperatures were generally equal to shelter-height

temperatures or lower by about 1.2 C. The warming influence

of Lake Okeechobee was evidenced by the fact that for a one-

pixel distance around the lake temperatures remained above

freezing while surrounding areas did not. That the

agricultural area was colder than surrounding areas was

attributed primarily to the strong radiative cooling of the








drained organic soils in the region. These soils also

exhibit higher emissivities than sandy soils, and therefore

should release energy at higher rates.

Chen et al. (1982) examined the influences of soils and

water on cold-prone areas in peninsular Florida using

nighttime GOES data. Coldest areas were most often found in

regions of low soil moisture content. These areas were

generally areas of well drained to excessively drained sandy

soils along the Central Florida Ridge. Warmer sites were

usually in wetter areas dominated by lakes, swamps or poorly

drained soils. The influence of moisture on surface

temperatures was particularly evident on one night, when the

northern part of the state exhibited higher temperatures

than did the southern end of the state because of frontal

rainfall in the north.

Allen et al. (1983) compared HCMM and GOES data for

mapping surface temperatures, and used HCMM apparent thermal

inertia (ATI) imagery to predict nocturnal cold-prone areas

in peninsular Florida. The greater resolution of the HCMM

data (0.6 by 0.6 km) as opposed to the GOES (6 by 8 km)

proved to be useful in providing a more detailed picture of

surface temperature distributions in the peninsula. HCMM

data were however limited for operational use by the

relative infrequency of the satellite's repeat cycle. HCMM

ATI imagery corresponded well with the general soils map of

the state. Areas containing well drained, sandy soils





34

exhibited higher thermal inertias than did less well drained

areas, and likewise these areas tended to be have the lowest

temperatures at night.

Regional frost mapping in Southern Victoria, Australia,

was attempted using HCMM data by Kalma et al. (1983). Five

winter images were obtained under conditions of low wind

speed and cloud cover. It was determined that realistic

distributions of temperature could be obtained from the HCMM

data, with lowest nighttime temperatures found in narrow

valleys, basins and depressions, and over fallow lands,

pastures and orchards. Higher temperatures were observed

over urban and built-up areas, forested areas, water bodies

and swampy areas. The authors concluded that HCMM data did

not have sufficient spatial resolution to be useful for

local frost mapping, though future systems with higher

resolutions would be useful.



Geologic Applications of Thermal Data

IR data have been widely used by geologists for mineral

exploration, geologic mapping, structural mapping and

geomorphology. The primary use of thermal data is in the

calculation of thermal inertia and temperature-difference

images, or as single-date imagery. The resulting images are

then interpreted manually. Several models have been

developed for calculating thermal inertia using as input

diurnal radiative temperatures, albedo and surface





35

meteorological parameters, but these models are generally

applicable only to areas where there is little or no

vegetation cover or evaporating water and therefore will not

be discussed here (Price, 1977; Kahle, 1977; Pratt and

Ellyett, 1979; Kahle et al., 1984; Price, 1985).

Single-date IR imagery has shown potential for detecting

mineralogical and chemical differences because of the

differing emissivities of various surface materials (Goetz

and Rowan, 1981). Lyon (1972) used an aircraft-mounted non-

imaging infrared spectrometer to show that mineralogical and

chemical differences on the surface can be detected using IR

data. Vincent and Thomson (1972) likewise demonstrated that

rock types could be discriminated using ratios of the IR

radiance from two thermal bands. Abrams et al. (1984)

demonstrated that color composite images produced using the

HCMM day visible-reflected infrared and day and night IR

data were valuable aids for geologic mapping.

The primary difficulty in discriminating rock types and

mineralogy using remotely sensed data is that the surface

should be largely unobscured by vegetation to obtain good

spectral signatures. In areas where vegetation cover is

found, thermal data-are useful for structural and geomorphic

mapping. Sabins (1969) evaluated the usefulness of IR

imagery for structural mapping in southern California. The

area was arid, with most of the vegetation being found in

irrigated fields. When compared to standard panchromatic





36

aerial photographs of the area, the IR imagery displayed the

greater contrast of the two forms of data, a fact attributed

to greater nighttime thermal emissivity variations than

daytime visible reflectance variations. The utility of the

IR imagery to structural mapping was demonstrated by the

fact that anticlines not visible in the photographs could be

recognized because of alternating beds of warm sandstones

and cool siltstones.

Structural mapping of the Front Range and adjacent plains

in Colorado was carried out by Offield (1975) using aircraft

IR scanner data. Several circular and linear topographic

features in the Front Range were clearly visible on the

thermal data, often more so than on visible photography.

The visibility of these topographic features is caused by

thermal shading (temperature contrasts) related to

cumulative heating effects between dawn and the imaging

time. In the plains areas, temperature contrasts were best

displayed in the pre-sunrise imagery and were related to

varying agricultural practices, drainage patterns and

moisture patterns along structural features.

Though Sabins (1969) and Offield (1975) used high

resolution aircraft scanner data for structural mapping,the

potential of coarser resolution satellite IR data has been

demonstrated by Schneider et al. (1979). Enhanced nighttime

NOAA AVHRR data were used to study the regional

geomorphology of an area covering North and South Dakota and








parts of Minnesota, Montana and Wyoming. Visible in the

imagery were the Missouri Escarpment, Coteau des Prairies,

several rivers and recessional moraines. The authors

considered moisture variations to be the most important

factor in determining temperature differences. North facing

slopes receive less incoming sunlight than do south facing

slopes and therefore retain moisture longer. This moisture

in turn increases the thermal inertia of materials on the

northern slopes, leading to a shaded-relief effect.

Elevation was also an important determining factor in

temperature. Topographic profiles and corresponding

temperature profiles displayed high negative correlations,

with major escarpments and valleys being clearly visible.



Land Cover Mapping

Thermal infrared data collected by satellite and aircraft

scanners can be used as input into pattern recognition

algorithms for land cover mapping. Thermal data potentially

offer additional information when combined with the normally

used visible and reflective infrared data, and may result in

classification accuracy increases. Though current uses of

thermal data for land cover classification are limited by

poor spatial or radiometric resolution or by spatial

resolution differing from simultaneously collected visible

and reflected infrared data, a few attempts to utilize

thermal data in classification schemes have been carried

out.





38

Ormsby (1982) found that inclusion of the Landsat 3 IR

band along with MSS4, 5, 6 and 7 improved classification

accuracies. IR data were particularly useful for aiding in

differentiating between urban and bare ground classes.

Price (1981) used principal component analysis to show that

additional information is provided by the Landsat MSS IR

data. Price however urged caution in the use of IR data for

multispectral classification because of its dependency on

slope, aspect and the thermal characteristics of the

surface. These dependencies may vary within as well as

between land covers and therefore may lead to spurious

classification results.

Byrne et al., (1981) found that HCMM visible and near

infrared and IR data could be used to monitor intermittently

flooding marshes in Australia. HCMM data were used to map

free water, woodland, damp grass and soil and dry areas.

The authors proposed that thermal data could be used to

monitor intermittently flooded marshlands because of the

sensitivity of the thermal band to moisture changes.














CHAPTER IV
STUDY AREA AND METHODOLOGY



Study Area

The area examined in this study includes approximately

45,600 km2 of north-central peninsular Florida (Fig. 1) and

was selected because of the wide variety of natural and

cultural land covers found within its boundaries. The

largest urban areas within the region include Jacksonville,

St. Augustine, Daytona Beach, Palatka, Starke, Lake City and

Gainesville. The other main cultural land covers include a

phosphate mining operation near White Springs, two heavy

mineral mines to the east of Starke and extensive

agricultural regions in the western portion of the study

area and in the Hastings area east of the St. Johns River.

The remainder of the study area consists primarily of

disturbed and undisturbed woodlands.





















































Figure 1: Study area.








Soils and Vegetation

Three major soil groups are found within the study area,

the North Florida Flatwoods, Central Florida Ridge and

Central and South Florida Flatwoods Soils (Table 1 and Fig.

2). North Flatwoods soils are primarily poorly drained

spodosols and entisols. Central Florida Ridge soils include

entisols, ultisols and alfisols, and are generally well to

excessively drained, with the exception of the Eureka-

Emeralda-Terra Ceia association found in Alachua and Marion

counties. The Central and South Florida Flatwoods soils are

primarily histosols and entisols. Most of these

associations are poorly drained except for the coastal

sands.

A wide variety of vegetation communities typical of

northern Florida are found within the study area (Fig. 3).

The dominant vegetation community in the region is pine

flatwoods. This community consists of longleaf, slash and

pond pines with an undergrowth of herbs, palmetto, shrubs

and small trees. Many flatwoods contain small hardwood

forests, prairies, swamps and cypress in poorly drained

areas. Longleaf pine-turkey oak forests are common on well

drained uplands. Undergrowth is often minimal, consisting

primarily of wire grass. The excessively well drained areas

of the Ocala National Forest contain sand pine communities.

Old dunefields in Marion and Levy County are also covered by

this community. Mixed hardwood forests are found in Levy








TABLE 1

North-central Florida soils.



Soils of the North Florida Flatwoods

6 Centenary-Leon-Plummer Association
7 Chipley-Kureb-Lakeland Association
8 Coastal Beach and Dunes Association
9 Coxville-Ocilla-Portsmouth Association
10 Ichetucknee-Chaires-Chiefland Association
11 Leon-Pelham-Mascotte Association
12 Plummer-Rutledge Association
13 Tidal Marsh and Tidal Swamp Association

Soils of the Central Florida Ridge

14 Adamsville-Lochloosa-Sparr Association
15 Alpin-Blanton-Chipley Association
16 Arredondo-Kendrick-Millhopper Association
17 Astatula Association
18 Blanton-Susquehanna-Fuquay Association
19 Blichton-Flemington-Kanapaha Association
20 Candler-Apopka-Astatula
21 Eureka-Emeralda-Terra Ceia Association
22 Jonesville-Pedro Association

Soils of the Central and South Florida Flatwoods.

24 Bushnell-Boca Association
25 Coastal and Beach Dunes Association
26 Istokpoga-Samsula Association
32 Paola-St. Lucie-Daytona Association
34 Pomona-Wauchula-Placid Association
35 Riviera-Winder Association
36 Tidal Marsh and Tidal Swamp Association
37 Wabasso-Felda-Pompano Association

See Figure 2 for accompanying map.






















































LEGEND

] SOILS OF THE NORTH FLORA FLATWOODS

SSOOLS OF THE CENTRAL AND SOUTH FLORIDA FLAruoo00

W SOLS OF THE CENTRAL FLORA RIDGE


Figure 2: North-central Florida soils.
After Caldwell and Johnson, 1982.


0Amnu* l~)*

























































LEGEND
1 COASTAL STRAIN 4 LONGLEAF PINE TURKEY OAK 7 MANGOVE & COASTAL MARSH
2 pME FULTWOOS 5 CYPRESS a MIXE HARDWOOD FOREST 0 aro Sa 5S
3 SAND PIE SWA FOREST FRESH WATER MARSH


Figure 3: Vegetation of north-central Florida.
After Davis, 1980.





45

and western Alachua and Marion counties. These forests are

mostly located on uplands with clayey soils and contain

oaks, magnolias, hickories, sweetgums and maples.

Swamp forests are found along the St. Johns, Suwanee,

Oklawaha and Santa Fe rivers. Maple, bay, cypress and gum

trees are common in swamp forest communities. Fresh water

marshes are located in some of the lower, poorly drained

areas. Mangrove swamp forests and coastal marshes are

common along the Gulf coast in saline or brackish waters.

Smaller extents of this community are found along the east

coast inland waterway from Daytona to Jacksonville. The

east coast barrier islands are dominated by coastal strand

vegetation. These communities usually consist of pioneering

grasses and shrubs near the shore with forests increasing

towards the lagoon side of the barriers.



Physiography

The physiography of the region is quite diverse,

primarily being controlled by solution of the limestone

bedrock and by the presence of Plio-Pleistocene shoreline

features. This region has been described and mapped in some

detail by several authors, most notably White (1958; 1970)

and Brooks (1981b) (Fig. 4). The following descriptions are

taken primarily from a physiographic map and accompanying

pamphlet by Brooks (1981b). More detailed information on

the physiography and geology of the region can be found in








Cooke (1945); Pirkle (1956); Puri (1957); Pirkle et al.

(1963); Bermes et al. (1963); Clark et al. (1964); Pirkle et

al. (1965); Williams et al. (1977). The study area includes

four physiographic regions of the state, the Sea Island

District in the northeast, the Ocala Uplift District in the

west, the Central Lake District in the center-south, and the

Eastern Flatwoods District in the east (Fig. 4). In the

following discussion the numbers and letters in parentheses

refer to the sections found on the physiographic map.


Sea Island District. The Sea Island District is divided

into three subsections. The Okefenokee Upland is an

undissected upland with poorly organized drainage. Included

in this subsection is the Okefenokee Basin (la), with

elevations ranging from 36 to 46 m. Vegetation in this

subsection is primarily marsh and cypress, grading into

poorly drained flatwoods. Two subsections are dominated by

two depositional marine features, the Lake City Ridge (lb)

and Trail Ridge (Ic). Elevations on the Lake City Ridge

reach 62 m, and the primary vegetation community is pine

flatwoods. Trail Ridge is one of the more notable

physiographic features found in northern Florida, being a

relict barrier with elevations from 46 m to a maximum of 73

m in the uplifted southern portion. Pine flatwoods are

common on the northern parts of the ridge, and longleaf

pine-turkey oak and sand pine are found on the southern end

of the ridge. The remaining area of the Okefenokee
































































0 ~ 0
hi fl_


Figure 4: Physiographic divisions of north-central Florida.
After Brooks, 1981b.








subsection is the High Flatwoods (Id), and consists of a

poorly drained area with elevations between 43 and 58 m.

The vegetation consists mainly of pine flatwoods and

extensive areas of riverine swamps.

The second subsection of the Sea Island District is the

Duval Upland and includes the St. Mary's Upland (2a), Black

Creek Basin (2b) and Penney Farms Upland (2c). This area

has elevations generally between 8 and 30 m and includes

several subdued paleo-beach ridges and marine terraces.

Vegetation communities include pine flatwoods on poorly

drained marine terraces, swamps along river basins and

longleaf pine-turkey oak communities .in the better drained

southern sand hills.

The third subsection is the Northern Coastal Strip

(3a-3g). This area is primarily the result of Pleistocene

and Holocene sea level fluctuations, with most elevations

being below 10 m, though some beach ridges can reach as high

as 22 m. Numerous beach ridges and dune fields are found

within this subsection. Vegetation communities include salt

water marsh, pine flatwoods, and longleaf pine-turkey oak.


Ocala Uplift District. The western portion of the study

area falls within the Ocala Uplift District. The region is

one in which Tertiary limestones are at or near the surface

and solution is a dominating force in shaping landforms.

The Big Bend Karst subsection is a low-relief surface

generally less than 6 m in elevation. Dune fields are found





49

in the Keaton Beach Coastal Strip (5b4) and longleaf pine-

turkey oak communities are dominant in this area. Horseshoe

Beach Coastal Strip (5b5) is a low limestone plain with very

swampy pine flatwoods. The Cedar Keys Coastal Strip (5b6)

contains a number of drowned relic dunes, with a maximum

elevation of 16 m found on one dune. The Waccasassa Coastal

Strip (5b7) is another low limestone plain vegetated with

hardwood forest and mixed flatwoods and swamps.

Inland of the coastal strip lies an area of poorly

drained terraces including the Waccasassa Flats (5c4),

Mallory Swamp (5c2) and San Pedro Bay (5cl). Elevations

range from 30 m in the north to below 17 m in the southern

Waccasassa Flats. Vegetation communities include pine

flatwoods in the Waccasassa Flats and grade into swamps and

low pine flatwoods in San Pedro Bay and Mallory Swamp.

Proceeding further inland one encounters the Suwannee River

Valley subdistrict. The Upper section (5dl) is a youthful

valley characterized by high bluffs, rock shoals and rapids.

The Lower section (5d2) displays less relief and is

characterized by swamp forest bordered by well drained

plains.

The Northern Peninsula Plains (5el-5e4) are found inland

of the Suwannee River Valley and are karst plains generally

between 18 and 30 m elevation. Vegetation types include

pine flatwoods, mesic hammock and freshwater marsh. Notable

in the southern part of this region are several prairies and






50

lakes resulting from solution of the limestone bedrock at

the water table. Examples are Paynes, Sanchez and Levy

Prairies, and Orange Lake and Lochloosa Lake. The Wellborn

Uplands (5fl-5f3) are the westward dissected extremity of

the Lake City Ridge. Maximum elevations here exceed 70 m.

Bordering the Sea Island District are the Northern

Peninsula Slopes (5gl-5g4). This area is a transitional

zone from the plateau region to the east and includes

numerous karst features. Vegetation communities include

mesic hammock and pine forests, with longleaf pine-turkey

oak found in the better drained areas. Overall this area is

well drained via surface or subsurface streams or sinks, and

elevations vary from 60 m at the eastern edge to lows of 27

m in karst depressions. Included in this subsection are

Newnans Lake and San Felasco Hammock.

The east-central portion of the Ocala Uplift district

includes the Marion Hills subsection (5hl-5h5). The area is

characterized by hill systems ranging from 24 to over 60 m

in elevation. Karst landforms are also common within this

subsection. Areas of sandy soils support longleaf pine-

turkey oak communities, while hardwood forests are found

elsewhere, particularly in the Fairfield Hills region east

of Orange Lake. The Oklawaha Valley (5i) is characterized

by river swamp bordered by poorly drained flatwoods. The

Newberry Sand Hills (5j) are dominated by a large forest of

longleaf pine-turkey oak forests with elevations mostly

between 24 and 45 m.






51

Central Lake District. The Central Lakes District is an

area of active sinkhole development that is part of the

central Florida ridge system. The northern part is an area

of perched lakes and prairies and includes Lake Santa Fe

(4a). Vegetation primarily consists of pine flatwoods and

swamp forest in this part of the District. The Interlachen

Sand Hills (4b) region is located to the east of Lake Santa

Fe in Putnam and Bradford counties. Elevations here reach

67 m, and numerous sinkhole lakes are present. The primary

vegetation community in this well-drained region is longleaf

pine-turkey oak.

South of the Interlachen area is the St. Johns Offset

(4c). The St. Johns River jogs to the west here, and

numerous springs are found in this area, as well as Lake

George, the largest lake in northern Florida. The primary

vegetation communities include pine flatwoods and river

swamp forest with many cabbage palms. West of the St. Johns

Offset is the Ocala Scrub (4d), much of which is within the

Ocala National Forest. The area is a paleo-dune field

covered by sand pine, longleaf pine and turkey oak.

Elevations in the Ocala Scrub range from 50 m in the west to

25 m in the east. The Crescent City-Deland Ridge (4d) is an

area of sand hills between Lake George and Crescent Lake.


Eastern Flatwoods District. The eastern coastal side of

the study area falls within the Eastern Flatwoods District.

Bordering the St. Johns River valley is the Palatka





52

Anomalies (lal-la6) subsection. This area is characterized

by limestone solution, stream diversion and possible

faulting. Elevations throughout this subsection are

generally below 12 m, except in the area of the Palatka

Relic Hills, where heights of 26 m are found. The primary

vegetation communities are pine flatwoods and swamp forest.

East of the city of St. Augustine is found the St.

Augustine Ridge Sets subsection (Ib). Elevations are

between 9 and 15 m, and the area is characterized by a

series of barrier island deposits. The subdued ridges in

the area are covered with pine flatwoods, and cypress is

found in the intervening swales. To the south of the St.

Augustine Ridge Sets lies the Volusia Ridge Sets subsection

(Ic). This subsection includes four distinct parts: the

Talbot Terrace at about 12 m, an eastern boundary ridge at

about 14 m, the Pamlico Terrace at about 8-10 m and the

Atlantic Coastal Ridge with maximum elevations of 17 m.

Several sets of beach ridges are found on the terraces and

the vegetation primarily consists of pine flatwoods.

The Atlantic coast is included in the Central Atlantic

Coastal Strip (lel). The principal feature in the

subsection is a coquina ridge forming the major portion of

the barrier island along the coast. Inside of this barrier

island is a lagoon system that is increasingly vegetated by

salt water marshes as one travels north.








Data

This study uses data obtained from two satellite-based

sensor systems, the Heat Capacity Mapping Mission (HCMM)

radiometer and the Landsat Multispectral Scanner (MSS).



Heat Capacity Mapping Mission

HCMM data were selected for this study because of the

closeness of the satellite's overpass to times of maximum

and minimum surface temperatures and their increased

resolution as compared to NOAA AHVRR or GOES data. The HCMM

satellite was the first of a planned series of Applications

Explorer Mission satellites to be placed in orbit. The

satellite collected data from its April 26, 1978 launch

until failure in September of 1980. The HCMM satellite was

intended to provide data for the study of thermal properties

of the earth's surface, and as such it carried instruments

for measuring reflectivity in the visible and reflective

infrared wavelengths (DAYVIS band) and for measuring emitted

IR (Table 2). The IR channel had a NETD of 0.4 K at 280 K.

The instantaneous field of view for the DAYVIS band was 500

m at nadir, and for the IR band it was 600 m at nadir,

though both were resampled during ground processing to

481.5-m cells. The satellite orbit was designed to allow

for day and night coverage of an area with 12-or-36-hour

separation (depending on latitude) with mid-latitude imaging

times of about 0230 and 1330 Local Sun Time (LST). The








orbit allowed for repeat coverage over an area every 16

days, though overlapping passes can reduce this time in some

instances. Types of data collected by or derived from the

HCMM satellite include day and night infrared radiances and

surface temperatures, temperature difference and

reflectivity images, and apparent thermal inertia images.

These data are available in image and/or digital form (HCMM,

1980).



TABLE 2

Sensor characteristics.


HCMM MSS

Band Bandwidth Band Bandwidth
(micrometers) (micrometers)

DAYVIS 0.55 1.1 MSS4 0.5 0.6
IR 10.50 12.5 MSS5 0.6 0.7
MSS6 0.7 0.8
MSS7 0.8 1.1






Calibration problems with the HCMM thermal radiometer

were noted after launch, and it was decided on the basis of

ground measurements that satellite temperatures were 5.2 K

too warm (Barnes and Price, 1980). As a result of this

finding, all HCMM data processed after June 1978 had 5.2 K

subtracted from them. Vukovich (1984) suggested that after

June 1978 the satellite calibration was corrected and that






55

data from July 1978 to at least September 1979 were 5.2 K

too low. Vukovich did find that calibrated and

atmospherically corrected HCMM data were within 1 K of

ground-measured surface temperatures.



Landsat MSS

The Landsat MSS has been in operation onboard five

different satellites since 1972. The primary purpose of the

Landsat satellites is to supply data for the study of

geologic, hydrologic, vegetative and cultural features on

the earth's surface. This system measures reflected visible

and infrared radiation in four spectral bands (hereafter

referred to as MSS4, MSS5, MSS6 and MSS7) between 0.5 and

1.1 micrometer wavelengths (Table 2). Additionally, Landsat

3 contained an IR band, though the data provided by this

sensor were of poor quality. The instantaneous field of

view of the sensor is 79 by 79 m, but because of the

sampling rate of the sensor and processing of the data done

on the ground, the area represented by each pixel is 56 by

79 m. Each MSS scene covers an area on the surface of the

earth approximately 185 by 185 km, and every area between

82 north and south latitude was covered every 9 or 18 days

depending on the number of satellites in operation at a time

for Landsats 1, 2 and 3. The imaging time was at

approximately 0930 LST so as to maximize topographic

shadowing for geologic purposes and to minimize afternoon






56

convective cloud cover. These data are available in either

digital or photographic image formats.

The satellite data used in this study are all in digital

form (Table 3). The HCMM data sets were chosen to provide a

representation of seasonal and day-night temperature

variations. Unfortunately, because of cloud coverage no

summer daytime HCMM data were available and many of the

nighttime images were of reduced quality, as will be

discussed in Chapter 5. The Landsat MSS data were collected

on November 5, 1978, and corresponded to HCMM data collected

on the same day.








TABLE 3

HCMM and Landsat data.


Data Date Scene Id.

HCMM 11-05-78 Day AA193-185-201,2

HCMM 12-17-78 Day AA235-183-611,2

HCMM 03-28-79 Day -AA336-181-901,2

HCMM 10-07-79 Day AA529-181-101,2


HCMM 05-21-78 Night AA025-074-003

HCMM 11-03-78 Night AA191-072-703

HCMM 02-01-79 Night AA281-070-603


Landsat MSS


11-05-78 30245-15230








Meteorological and Ancillary Data

Atmospheric sounding data for dates corresponding to the

HCMM imagery were obtained from the National Climatic Center

at Asheville, North Carolina. These data were used as input

into a radiative transfer model for correcting satellite-

derived temperatures for atmospheric attenuation.

Meteorological conditions on the imaging dates were obtained

from National Oceanographic and Atmospheric Administration

(NOAA) daily and three-hourly weather data and from 0700

surface weather maps. Remotely-sensed data used for land

cover identification included 1983 and 1984 National High

Altitude Mapping Program (NHAP) false color transparencies

at 1:58,000 scale, black and white aerial photographs

obtained between 1974 and 1979, Landsat MSS data collected

in April of 1973 and 1974 and December 18, 1982 1:500,000

Landsat Thematic Mapper (TM) imagery. Additional data

sources include soils, vegetation and physiographic maps and

field inspections.



Image Processing and Analysis

The processing and analysis of the data can be divided

into three stages: geometric correction, generation of

output for interpretation and image interpretation. These

steps will be discussed separately below.








Geometric Correction Procedures

All digital image processing was done using the Earth

Resources Laboratory Applications Software (ELAS) at the

University of Florida and Institute of Food and Agricultural

Sciences Remote Sensing and Image Processing Laboratory and

the IBM 7350 system running the High Level Image Processing

System (HLIPS) software at the Northeast Regional Data

Center. The ELAS modules used here are designated by their

four letter acronyms and are described in the ELAS Users

Guide (Graham et al., 1984).

The Landsat MSS data were used to provide information on

land covers within the study area that could be then related

to temperature variations as measured by the HCMM IR data.

Several preprocessing steps were required before such

information could be obtained from the satellite data.

Landsat MSS data are collected by a series of 24

detectors, 6 for each band, and because calibration

differences occur between the detectors a striping effect is

sometimes observed in the data. To correct for this

calibration problem a destriping procedure (DSTB and DSTR)

was used to minimize this detector variability. The MSS4

data contained a large number of missing scanlines and were

therefore excluded from any further analyses.

As with all satellite imagery, platform instability and

sensor characteristics produce geometric inaccuracies in the

data. These inaccuracies were corrected by mapping the





60

satellite data to the Universal Transverse Mercator (UTM)

projection. This mapping procedure involved selecting

points in the satellite data and corresponding UTM

coordinates from 1:24,000 quadrangle sheets. Such points-

were usually road intersections or small lakes. From these

points a least squares technique was used to calculate a

piece-wise linear mapping function that allowed Landsat data

to be mapped to the UTM coordinate system. Fifty control

points were selected from a 1865-pixel-by-1585 line Landsat

subscene covering the central portion of the study area.

These points were used to calculate a mapping function

(OGCN) with a RMS error of 51 m, indicating that the

corrected pixel locations are within 51 m of their actual

UTM coordinates. A bilinear interpolation procedure was

then used to resample the pixels from the original 56-by-79

m size to 79-by-79 m and to map the data to the UTM

projection (OGEO).

The HCMM subscene, which is 512 pixels by 384 lines, was

also geometrically transformed to the UTM coordinate system.

It was first necessary to register the November 5, 1978 HCMM

data to the corrected Landsat data of the same date to allow

comparison of the two data sets. This registration was

accomplished by first subdividing the 481.5-m HCMM pixels so

they were approximately the same size as the corrected

Landsat pixels. Next, 48 matching control points were

chosen in each of the two images and a third-order








polynomial mapping function with a 137-m RMS error was

calculated using the HLIPS registration procedure. The HCMM

data were then registered to the Landsat MSS data using a

cubic convolution interpolation algorithm. Ater completion

of the registration procedure a counterclockwise skewing of

approximately 7 of the resampled HCMM data was noted. This

skewing did not notably affect the analysis of the imagery.

Since only the November 5, 1978 HCMM data were compared

directly to the Landsat MSS data, the following approach was

chosen for registering the remaining HCMM images to UTM

coordinates and to each other. The non-subdivided November

5, 1978 data were chosen as the base map and registered to

the UTM coordinate system using 51 control points found on

1:250,000 topographic maps. A linear mapping function with

a 283-m RMS error was calculated, and bilinear interpolation

was used to resample the data to the UTM projection (PMGC,

PMGE). The remaining HCMM data sets were then overlayed to

the corrected November 5, 1978 data (OCON, OVLA).

Though all HCMM data sets were overlayed using mapping

functions with sub-pixel accuracy, visual inspection of the

imagery showed that some areas were misregistered by one or

two pixels. This misregistration was on the order of that

obtained by Watson et al. (1982), and was caused by the

difficulty in finding adequate control points in certain

portions of the imagery. Fortunately most misregistered

areas were not within the primary area of interest and

therefore did not pose a serious problem for image analysis.








Production of Output for Interpretation

Several types of data display techniques were used for

image interpretation in this study. These are discussed

below and include overlaid Landsat MSS and HCMM images,

isotherm and color-coded temperature maps and enhanced

greyscale images.


Overlaid Landsat MSS and HCMM images. Overlayed Landsat

MSS and HCMM images were used in conjunction with air photos

to aid in relating individual HCMM pixels to specific land

covers. Because the study area was larger than available

digital Landsat MSS data, overlayed HCMM-Landsat data were

only used over the central portion of the study area. Two

methods of displaying the co-registered Landsat MSS-HCMM IR

data were tested, contrast-stretched red-green-blue images

(RGB) and intensity-hue-saturation images (IHS). RGB images

are simply made by displaying three different image bands

using the red, green and blue guns of the color output

device. In an IHS display three bands are displayed as

intensity (brightness), hue (color) and saturation (color

purity) (Siegal and Gillespie, 1980; Schowengerdt, 1983).

Landsat MSS5, MSS7, a linear combination of MSS5 and MSS7

known as the Transformed Vegetation Index (TVI) and the HCMM

IR were displayed in different RGB and IHS combinations on

the color display. An RGB image with the TVI displayed in

red and the HCMM IR in green and blue proved to be the most

useful for relating HCMM temperatures to the various land

covers.





63

The TVI is one of several linear combinations of Landsat

MSS bands that have been shown to be sensitive to green-leaf

biomass and productivity over a wide variety of natural and

agricultural land covers (Rouse et al., 1974; Tucker, 1979;

Tucker et al., 1981; Curran, 1982; Myers, 1983; Jackson,

1983; Huete et al., 1984; 1985; Jensen and Hodgson, 1985).

The TVI is calculated using the following formula:


TVI = SQRT (((MSS7 MSS5)/(MSS7 + MSS5)) + 1.0)


This results in an index between 0.0 and 1.33 that is then

multiplied by 100.0 for display purposes (Fig. 5). The TVI

offered better vegetation discrimination than did MSS5 or

MSS7, while also allowing identification of water and urban

areas. Color plots and slides of the resulting images were

made for interpretation purposes.


HCMM temperature maps. Isoline and color-coded,

atmospherically corrected temperature maps were produced

from the geometrically corrected HCMM data. To obtain

surface temperatures that were as close as possible to true

surface temperatures, corrections for atmospheric

attenuation and emissivity due to water vapor were done.

Atmospheric corrections were made using Price's (1983b)

radiative transfer model. A spatially constant surface

emissivity of 0.97 was chosen based on emissivity values for

vegetation and soils given by Taylor (1979) and Smith

(1983). This model has been shown by Price (1983b) to




















































Figure 5: November 5, 1978 Landsat MSS TVI image.








correct apparent surface temperatures to within plus or

minus 2.0 to 3.0 C of actual surface temperatures.

Input data for the model included atmospheric pressure,

temperature and dew point at standard and critical

radiosonde levels. The radiosonde data were collected by a

NOAA meteorological balloon launched from Waycross, Georgia,

approximately 180 km north of the center of the study area.

Day HCMM data were corrected using soundings obtained at

0000 GMT the day after the satellite overpass. These

soundings correspond to approximately 1830 LST on the day of

the satellite overpass. Night HCMM data were corrected

using soundings obtained at 1200 GMT the day of the

satellite overpass. These data correspond to approximately

0630 LST on the day of the satellite overpass.

After atmospheric correction equations were derived from

the model, lookup tables for converting HCMM digital numbers

(DNs) to temperature were calculated using the following

equation (HCMM, 1980):


T(I) = Kl/ln(K2/(I-K3) + 1.0)


where T is temperature in degrees K, I is the HCMM DN, K1 is

14421.587, K2 is 1251.1591 and K3 is -118.21378,

Additionally, 5.2 K were added to the raw HCMM temperatures

to correct for radiometer calibration errors (Vukovich,

1984). New lookup tables of corrected temperatures were

made using equations provided by Price's model. Based on








the assumption that canopy temperatures approximate air

temperatures (Smith et al. 1981), comparisons of radiometric

temperature of forested areas to shelter-height temperatures

throughout the study showed the corrected HCMM temperatures

to be accurate within plus or minus 2.0 to 3.0 C for most

images.

Atmospherically corrected temperatures were displayed

using isotherm maps obtained from the Surface II mapping

program (Sampson, 1975) and as color-coded temperature maps.

Contour maps offer the advantage over color-coded

temperature maps of being a smoothed representation of the

data, and therefore easier to interpret on a general level.

The maps were generated at a scale of 1:500,000, allowing

them to be readily overlaid on available maps and Landsat

MSS imagery.

The color-coded temperature maps were produced by

assigning colors to ranges of temperatures obtained from the

satellite imagery and. corrected-temperature lookup tables.

For day imagery each color represented approximately three

DNs, or a range of approximately 0.7 to 0.9 C. Night

images, because of the decreased temperature range, were

generally assigned two DNs (approximately 0.5 to 0.7 C

range) per color.


Grevscale HCMM IR images. Geometrically corrected

greyscale HCMM and IR images allowed the most-detailed

analysis of surface temperature patterns in this study.





67

These data were used for statistical correlation analyses

and visual interpretations. Correlations between DNs of the

different dates were calculated for the entire study area

(excluding the Atlantic Ocean and Gulf of Mexico) as well as

for subsets of the study area. The results of the

correlation analysis are given in Chapter 5.

The atmospherically uncorrected HCMM data were also

examined visually to avoid the loss of detail that occurs

when DNs are converted to temperatures and merged into color

classes or displayed as isotherm maps. The data were

enhanced using various linear and histogram equalization

contrast stretches (Schowengerdt, 1983). These enhancements

simply expand the limited range of image DNs to fully

utilize the range of the display device.



Image Analysis

As stated above, the object of image analysis was to

relate HCMM-derived surface temperatures to land cover and

land use, soils and physiographic features in north-central

Florida. A hierarchical approach to image analysis was

used, proceeding from the interpretation of general to more

specific temperature-surface feature relationships. First,

geometrically and atmospherically corrected 1:500,000 HCMM

isotherm maps were overlayed on Landsat MSS photographic

images. This allowed the identification of general surface

temperature patterns and comparison of the relative








complexity of these patterns for different parts of the

study area.

In the second step, color-coded temperature maps and

enhanced greyscale imagery were examined in conjunction with

Landsat MSS, TM and black and white air photo mosaics. This

allowed the interpretation of greater detail than that

obtainable from isoline maps because actual pixels were

visible in the data. These images were interactively

displayed using the ELAS software, allowing various contrast

stretches, enlargements and statistics to be generated. The

ELAS DGTZ module was particularly useful at this stage of

interpretation. This program allows the user to digitize

areas on maps and highlight corresponding areas in the HCMM

data. This capability was valuable for identifying the

location of specific surface features in the relatively

coarse scale HCMM data.

The final step involved the examination of land cover-

temperature relationships for a subset of the HCMM data for

which color-infrared NHAP aerial photographs were available.

This area extends from Kingsley Lake in the north to the

Oklawaha River in the south and from San Felasco Hammock in

the west to the town of Interlachen in the east. Each of

the 38 NHAP photos covering this area were examined along

with the corresponding HCMM data. Individual DN values for

various land covers were extracted from the HCMM data and

converted to temperatures using the atmospherically





69

corrected lookup tables discussed previously. These data

provided the most detailed information on surface

temperature characteristics of the study areas.














CHAPTER V
IMAGE INTERPRETATION RESULTS


Temperature patterns for each imaging date are described

separately below, first addressing the day and then the

night data. For each date the meteorological conditions

under which the data were obtained are given as well as a

description of temperature patterns. A common format for

the description of the data is based on the fact that the

study area can be subdivided into several relatively

homogeneous regions, which often correspond with surface

features. These thermal regions will be discussed first,

followed by a discussion of the individual imaging dates.



Thermal Regions

Seven thermal regions are discernable in all of the day

images (Fig. 6). The Gulf Coast Region includes all or

parts of Taylor, Lafayette, Dixie and Levy counties.

Daytime temperatures in this region are generally low

relative to those found over less heavily vegetated regions.

The eastern boundary of this region closely follows that

between the Central Florida Ridge and North Florida

Flatwoods Soils (see Figure 2 in Chapter 4). Soils within

the Gulf Coast Region include Tidal Marsh and Tidal Swamp,





71

Centenary-Leon-Plummer, Plummer-Rutledge and Itchetucknee-

Chaires-Chiefland associations. All are poorly drained

soils. Vegetation is primarily pine flatwoods with numerous

cypress stands and marshes. Hardwood swamp forests are

found along lakes and rivers and the Gulf coast is bordered

by salt marshes.

The Suwannee Agricultural Region runs north and south the

length of the study area, with daytime temperatures in this

region among the warmest in the study area. The area is

dominated by agricultural land covers, with Suwannee and

Gilchrist counties having more than 50% of their land in

farms, and Alachua, Levy and Marion having more than 30% in

farms (U. S. Department of Commerce, 1981). Common crops

include corn, sorghum, wheat, peanuts, soybeans and various

hay and silage crops. Cultivated fields and pastures are

interspersed with smaller forested areas consisting of mixed

hardwoods, planted pine or longleaf pine-turkey oak

communities. The boundaries of the Suwannee Agricultural

Region closely follow those of the soils of the Central

Florida Ridge. Included are the Alpin-Blanton-Chipley,

Blanton-Susquehanna-Fuquay, Astatula, Jonesville-Pedro and

Candler-Apopka-Astatula associations. All of these

associations contain well drained sandy soils.

The Interlachen Karst Region extends from Kingsley Lake

in Bradford county south to the Oklawaha River in northern

Marion county, and from Lake Santa Fe eastward into Putnam


























Figure 6: Thermal Regions.
Lighter tones indicate warmer temperatures.























Y4 4 es





THRALRG OSON1HCMDA
Sr.S








GC TH I
S.iEjIL





74

county. This region of elevated temperatures includes soils

from all three major soil groups, but little correspondence

is found between soils found on the 1982 general soils map

(Caldwell and Johnson, 1982) and temperature patterns

observable in the HCMM imagery. One exception occurs in the

south where the excessively drained Candler-Apopka-Astatula

association is found. Also located in the region are the

Pomona-Wauchula-Placid, Istokpoga-Samsula, and Leon-Pelham-

Mascotte associations. The soils included in these

associations are all poorly drained. It should be noted

that an earlier soil map (Beckenbach and Hammett, 1962)

shows a greater percentage of the Interlachen Karst Region

to contain well drained soils, and surface temperature

patterns correspond well with this map. The vegetation in

this region includes pine flatwoods in the north, longleaf

pine-turkey oak on well drained sand hills, and marsh and

cypress in karst depressions. The region is notable for the

large number of small karst-related lakes and sinkholes.

East of the St. Johns River in Flagler, Putnam and St.

Johns counties is the Hastings Agricultural Region. Common

crops are potatoes, vegetables, cabbage and various hay

crops. This region falls entirely within the Soils of the

Central and South Florida Flatwoods and includes only the

poorly drained soils of the Pomona-Wauchula-Placid

association. The eastern edge of this region is bounded for

the most part by the Riviera-Winder association. Other than








this boundary, no clear relationship between temperature

patterns and soil associations can be identified in the

region.

The Lake George Region is a region of slightly warmer

temperatures bordered on the north and west by the Oklawaha

River and extends to the southern edge of the study area and

east to Crescent Lake. The boundaries of the Lake George

Region correspond closely to two soil associations of the

Central Florida Ridge, the Candler-Apopka-Astatula and

Astatula associations. These associations are composed of

well drained sandy soils. West of Lake George in the Ocala

National Forest the primary vegetation communities are sand

pine and longleaf pine-turkey oak forest. Sand pine forests

are actively logged within the national forest. Between

Lake George and Crescent Lake pine flatwoods and longleaf

pine-turkey oak are the dominant forest communities, though

numerous small agricultural holdings are common. Hardwood

swamp forests are found around Lake George and along the

Oklawaha and St. Johns Rivers.

The Atlantic Coastal Region is an area of cooler

temperatures extending from the Florida-Georgia border to

the southern end of the study area and inland 20-25 km. The

region includes soils from the North Florida Flatwoods and

South and Central Florida Flatwoods. In Duval and Nassau

counties these include Tidal Marsh and Swamp, Coastal Beach

and Dunes, Chipley-Kureb-Lakeland, Plummer-Rutledge and








Leon-Pelham-Mascotte associations. All but the Coastal

Beach and Dunes and Chipley-Kureb-Lakeland associations are

poorly drained soils. South of Duval county are the

Riviera-Winder, Coastal Beach and Dunes, Pomona-Wachula-

Placid, Istokpoga-Samsula and Tidal Marsh and Tidal Swamp

associations. All except the Coastal Beach and Dunes

association are poorly drained. No distinct relationship

between soil associations and surface temperature patterns

is evident in the Atlantic Coast Region. Vegetation

communities include coastal strand, salt marsh, pine

flatwoods, longleaf pine-turkey oak and hardwood swamp

forests.

The last region, the Central Region, includes the portion

of the study area between the Suwannee Agricultural Region

in the west and the St. Johns River in the east, excluding

the Interlachen Karst and Lake George Regions. The western

and southern boundaries of the Central Region correspond

closely to those of the well drained Central Florida Ridge

soils. The northern section includes the Leon-Pelham-

Mascotte and Plummer-Rutledge associations, both comprised

poorly drained soils. The southern section in Alachua,

Putnam and Marion counties includes the Adamsville-

Lochloosa-Sparr, Pomona-Wauchula-Placid, Istokpoga-Samsula

and Bushnell-Boca associations. All are comprised of poorly

drained sandy soils. The primary vegetation community in

the region is pine flatwoods, with longleaf pine-turkey oak








forests found on well drained uplands in Nassau, Clay,

Putnam and Alachua counties. Pine forests are subject to

intensive lumbering throughout the region. Hardwood and

cypress forests are present along river valleys, in wetter

areas around lakes and rivers and in the Okefenokee Swamp

near the Florida-Georgia border. Extensive freshwater

marshes are found in Alachua county.

Night temperature patterns for the dates used in this

study do not correspond to the above thermal regions as well

as do the day data. Some of this lack of correspondence is

because of the overall poor quality of the available night

imagery. In spite of this lack of correspondence the

thermal regions will be used to describe nighttime

temperature patterns. This is done for two reasons. First,

the daytime patterns appear to follow land cover and land

use patterns in the study area relatively closely, providing

a physical basis for the use of these regions for discussion

purposes. Second, the use of the thermal regions allows a

uniform format for discussing temperature patterns.



Daytime Surface Temperature Patterns

November 5, 1978


Meteorological conditions. The November 5, 1978 HCMM

data are cloud-free and of good quality (Fig. 7 and Table

4). The 0700 EST surface weather map shows that the area

was under the influence of a high pressure cell centered








over Tennessee and the barometric pressure was rising.

Windspeeds at the time of imaging were low and from an

easterly direction. No rainfall was recorded in the study

area on the day of the satellite overpass. Only three

stations within the study area measured any rainfall within

five days prior to imaging, Fernandina Beach, St. Augustine

and Daytona Beach, all below 0.18 cm. October 1978 was

drier than normal for most of the study area, with 1.3 to

9.1 cm of rainfall. Daytona had anomalously high rainfall,

with 21.0 cm for the month.


Gulf Coast Region. Temperatures in the Gulf Coast Region

range from 25.2 to 26.0 C for coastal marshes and wetlands

to greater than 35.9 C over interior agricultural areas.

Most of the region has temperatures between 26.3 C and 29.4

C. Temperatures in the region correspond closely to

vegetative land covers. Coastal marshes are among the

coolest land covers, though these wetlands are poorly

delimited for most of the Gulf Coast Region, tending to

blend in with many of the wetter forests. This is

particularly evident along the Suwannee River where

temperatures of 25.2 to 26.0 C extend several kilometers

upstream of the salt marshes.

A close correspondence between surface temperature and

forest type as found on the 1972 1:250,000 USGS Land Use and

Land Cover Map may be observed in the image. Cooler areas

(25.2 to 27.1 C) correspond closely to the forested wetland

























Figure 7: November 5, 1978 HCMM IR image.























































O0
0








TABLE 4

November 5, 1978 meteorological conditions.


Station


Jacksonville
Tallahassee
Orlando
Tampa


Temperature Wind Wind Relative Pressure
Speed Direction Humidity


23.8
25.5
26.6
26.1


0.0
5.5
12.9
11.1


1018.6
1018.2
1017.9
1017.8


Daily air temperatures:
(shelter height)




Soil temperatures:
(10.2 cm depth)


Maximum
Minimum
Average


of 28.9 at Live Oak
of 23.8 at Jacksonville
of 26.1


Maximum 25.6 at Monticello
Maximum 21.6 at Gainesville


All temperatures are given in degrees C, windspeeds in
km/hr, wind direction in tens of degrees counterclockwise
from north, relative humidity in percent and pressure
in millibars. Temperature, wind speed, wind direction
and relative humidity for Jacksonville, Tallahassee,
Orlando and Tampa were taken at 1300 EST. Pressure was
obtained from the 0700 EST surface weather map.








category while warmer areas (27.6 to 29.4 C) match the

evergreen forest land category. Evergreen forests are

dominated by pines interspersed with cypress domes and are

subject to intensive logging. Logging of the pine forests

leads to a wide variety of stand ages and heights within the

region, ranging from recent clearcuts to mature pine

forests. Forested wetlands also contain pines, but

frequently have a higher number of cypress and hardwood

species. These areas are subject to less intensive logging.

Within the Gulf Coast Region are areas of notably higher

temperatures. Warmer temperatures (in excess of 34.1 C)

around Old Town and Fanning Springs in Levy and Dixie

counties correspond to agricultural land covers. Several

smaller areas can be identified with forests that have been

recently clearcut or are in various stages of regrowth.

Examples of such areas can be found in northern Lafayette

county, southern Lafayette county approximately 20 km west

of the junction of Santa Fe and Suwannee Rivers, just

northeast of the mouth of the Suwannee River, 15-20 km

northeast of Cedar Key and north of Waccasassa Bay.

Temperatures are from 2.3 to 8.7 C higher than those of

nearby pine flatwoods and higher by another 1.0 to 2.0 C

over nearby forested wetlands.


Suwannee Agricultural Region. Temperatures are higher in

the Suwannee Agricultural Region than in dominantly forested

regions; the mean temperature in the region is 33.5 C as





83

compared to 27.6 C in the Gulf Coastal Region. Temperature

patterns are complex in the Suwannee Agricultural Region

because of the mix of bare soil, grass, crops and forest in

the area. Minimum temperatures of 27.6 C are found along

the Suwannee River and over forested areas; maximums exceed

39.0 C over agricultural lands.

Vegetation density appears to be the primary factor

controlling surface temperatures in the Suwannee

Agricultural Region. Hardwood forests along the Santa Fe

and Suwannee Rivers have temperatures between 27.6 and 27.9

C, with planted pine forests generally 2.0 C or more warmer.

Temperature differences within pine forests, and between

pine forests and nearby agricultural lands are also evident.

A planted pine forest located on well to moderately drained

soils has temperature differences of 1.0 to 1.5 C between

the most densely forested and less heavily forested

sections. The pine forests are 6.0 to 8.0 C cooler than

adjacent agricultural fields. Similar temperature

differences are found in comparisons between planted pines

around the Deerhaven Power Plant north of Gainesville and

nearby agricultural lands. The sparse longleaf pine-turkey

oak forests found in Gilchrist and Levy counties are 1.5 to

2.0 C warmer than dense pine forests and experience

temperatures similar to those of young planted pine forests.

Agricultural areas with minimal vegetation cover

experience the highest temperatures in the region. By early





84

November vegetation cover over agricultural lands is minimal

because most crops have been harvested. These sparsely

vegetated agricultural areas have temperatures greater than

32.0 C, with extensive areas greater than 35.3 C. Maximum

temperatures in the region exceed 39.0 C in western Marion

county.

Within or bordering the Suwannee Agricultural Region are

three urban or industrial features influencing surface

temperature patterns; they are the open-pit phosphate mine

in Hamilton county and the cities of Lake City and Ocala.

Examination of the isotherm map shows that the phosphate

mine causes the 29.0 C degree contour to extend eastward to

include the mine, with temperatures approximately 1.0 to 2.0

C higher than surrounding forested areas. More pronounced

are the urban heat islands of Lake City and Ocala. Lake

City has a maximum temperature difference of 10.0 C over

nearby pine forests. Ocala has a less pronounced heat

island, with temperatures 2.0 to 4.0 C above surrounding

agricultural lands.


Interlachen Karst Region. Minimum temperatures in the

Interlachen Karst Region are 26.3 to 27.1 C for pixels

influenced by lakes, while maximums exceed 37.0 C over

agricultural lands. The majority of the region exhibits

temperatures greater than 29.7 C and large areas are warmer

than 33.0 C.





85

High temperatures in the region are related to the sparse

vegetation and well drained soils found there. The dominant

vegetation community, longleaf pine-turkey oak forest,

generally has an open canopy and minimal understory. This

means that a greater percentage of the higher-temperature

soils will be visible to the satellite sensor, resulting in

elevated surface temperature measurements. The potential

for high temperatures in these areas is further increased by

clearing of the already sparse forests for housing

developments. An example of this is found in a region of

longleaf pine-turkey oak forests east of Lake Geneva where

temperatures exceed 34.1 C. Similar temperatures are

located around the Interlachen area and south of Kingsley

Lake over Camp Blanding. Comparison to pine flatwoods

adjacent to the Interlachen Karst Region show longleaf pine-

turkey oak forests to be 5.0 C or more warmer.

Highest temperatures (37.0 C) in the region are located

over agricultural areas where maximum soil exposure is

found. Temperatures over an open-pit heavy mineral mine

southeast of Kingsley Lake reach 36.3 C, only slightly lower

those found over agricultural areas.


Hastings Agricultural Region. As in the Suwannee

Agricultural Region, lowest temperatures (25.2 to 26.0 C)

are associated with forested inliers. In the Hastings

Agricultural Region these are primarily hardwood swamp

forests along the St. Johns River and its tributaries.





86

Temperatures over forested areas are between 25.2 and 27.1

C, and pastures are mostly 1.0 to 2.0 C warmer.

Temperatures greater than 35.3 C are common in cultivated

areas, with a significant amount of the region exhibiting

temperatures greater than 36.3 C. In spite of the poorly

drained soils in the region surface temperatures are similar

to those found over more well drained agricultural areas

elsewhere.


Lake George Region. Lowest temperatures, 25.2 to 26.0 C,

in the Lake George Region are found over swamp forests and

maximum temperatures exceed 34.8 C for an agricultural area

approximately 15 km north of Lake George. The majority of

the land surface in the region lies between 27.6 and 33.8 C.

Highest surface temperatures are found over three land

covers, sparsely vegetated longleaf pine-turkey oak forests,

logged areas and agricultural sites. Elevated surface

temperatures, up to 31.6 C, occur over an area of longleaf

pine-turkey oak forest south of Lake Kerr. An adjacent

clearcut area has temperatures between 33.0 and 33.8 C.

Another area of longleaf-pine turkey oak south of Rodman

Reservoir has temperatures from 30.9 to 31.6 C.

Temperatures in the northeast corner of the Ocala

National Forest range from 28.3 to 33.8 C with most between

29.7 and 30.9 C. This area is covered by a relatively dense

sand pine forest that has been extensively logged, therefore

high surface temperatures would be expected. Because








temperatures are for the most part lower than found over

clearcuts elsewhere in region it is plausible that

temperatures in the area are being masked by the relatively

coarse resolution of the HCMM sensor. The logged areas are

generally small, below 0.3 km2 and irregularly shaped.

Overlaying of a template representing the original

600-by-600-m HCMM temperature measurements onto NHAP air

photos shows that pixels seldom fall entirely on clearcuts,

thus the temperatures measured will commonly include both

clearcut and forest. The resulting mixed pixels will lead

to increased temperatures for forests and decreased

measurements for clearcuts.

High surface temperatures are found over agricultural

lands and pastures between Lake George and Crescent Lake.

Surface temperatures here are generally between 30.9 and

35.9 C. Temperatures between 26.3 and 28.6 C are found over

forested areas and around small lakes.


Atlantic Coastal Region. In the Atlantic Coast Region

minimum temperatures are 23.7 to 24.9 C along the Atlantic

coast and for salt marsh and lagoon areas. Maximum

temperatures of 30.9 to 31.6 C are associated with urban and

interior agricultural areas. The majority of the region

lies between 25.2 and 27.1 C.

As in the Gulf Coast Region, temperature differences can

best be explained by vegetation cover and surface moisture,

with wetlands being cooler than dry forests. Agricultural





88

and logged areas tend to exhibit higher temperatures (by 1.0

to 2.0 C) than do forested areas. Because of their

influence on the vegetation distribution in the Atlantic

Coast Region Pleistocene beach ridges are visible in some

portions of the region. The lack of definition of these

features is probably caused by the coarse resolution of the

HCMM thermal data.

Urban areas exhibit a notable influence on surface

temperatures in the Atlantic Coast Region. Eastern

Jacksonville is approximately 2.0 to 4.5 C warmer than rural

areas and a slight heat island can be seen on Jacksonville

Beach. Daytona Beach has a maximum urban-rural.temperature

difference of approximately 7.0 C.


Central Region. Minimum temperatures (excepting water

bodies) in the Central Region are 25.2 to 26.0 C and are

found over freshwater marshes and hardwood swamp forests.

Maximum temperatures exceed 37.0 C over agricultural areas

and at the heavy mineral mine northeast of Starke. The

majority of the region exhibits temperatures between 26.3

and 31.6 C.

As elsewhere, temperature differences can be related to

variations in vegetation covers. Coolest temperatures are

associated with wetlands. Wetland forests north of Lake

Santa Fe and north of Newnans Lake are between 26.0 and 26.8

C. Similar temperatures are found north of Ocean Pond and

in the Okefenokee Swamp. Freshwater marshes in Paynes








Prairie, Lake Levy and around Orange Lake have roughly

equivalent temperatures, between 26.0 and 26.8 C.

Temperatures over pine flatwoods range from 26.3 C in the

wettest areas to approximately 29.0 C in drier areas. Mixed

pine-cypress forests north of Gainesville have temperatures

as low as 26.3 while those around Lake Sampson and in Austin

Cary Forest north of Newnans Lake are between 27.6 and 28.9

C. Drier pine forests northeast of Kingsley Lake are

slightly warmer, between 28.3 to 28.9. Temperatures over

dominantly hardwood forests in San Felasco Hammock are

between 27.6 to 27.9 C, roughly equivalent to the coolest

pine forests.

As in the Gulf Coast Region, comparisons with air photos

and Landsat imagery show a close correspondence between

areas approximately 2.0 to 7.0 C warmer than surrounding

forests, recently logged or regrowth sites. A clearcut in

Austin Cary Forest is up to 5.0 C warmer than surrounding

forests. Another clearcut north of Lochloosa Lake is

similarly warmer than surrounding forests.

Urban and industrial areas exert important local controls

on surface temperatures in parts of the Central Region. The

northern heavy mineral mine east of Starke has temperatures

greater than 37.0 C, making it one of the warmest areas in

the region. The urban area of Jacksonville has a well-

developed heat island, with downtown temperatures

approximately 6.0 C above surrounding rural areas. Also








evident are the heat islands of Orange Park and Palatka.

Gainesville has maximum temperatures of 36.6 C,

approximately 9.0 C greater than nearby forests and more

than 10.0 C greater than nearby wetlands.

It is interesting to note that airports are consistently

some of the warmest cultural land covers, often warmer even

than downtown urban areas. Jacksonville and Cecil Field

Naval Air Stations Naval Air Stations both have temperatures

between 34.1 and 34.8 C, and Jacksonville International

Airport has temperatures up to 31.6 C. Even smaller

airports at Gainesville and the abandoned Green Coves

Springs military base display temperatures approximately 4.0

C warmer than adjacent rural regions.



December 17, 1978


Meteorological conditions. The December 17, 1978 HCMM

data are the lowest quality day imagery used in this study

(Fig. 8 and Table 5). The 0700 EST surface weather map

shows a cold front running from Jacksonville to Cedar Key.

This front appears to have stalled and is visible over the

northern part of the study area. Clouds are evident

northeast of Jacksonville and light cloud cover influences

temperature measurements over much of the study area north

and east of the Suwannee River. A north-south temperature

gradient is noticeable in the imagery, with regions north of

a line running from Jacksonville to Steinhatchee notably




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs