Tennessee
Valley
Authority
Office of
Natural Resources
Chattanooga, TN 37401
United States
Environmental Protection
Agency
Office of Energy, Minerals, and
Industry
Washington DC 20460
TVA/ONR- 79/01
EPA-600/7-79-138'
June 1979
Research and Development
Remote Sensing of
Sulfur Dioxide
Effects on
Vegetation
Photometric
Analysis of Aerial
Photographs
Interageney
Energy/Environment
R&D Program
Report
-------
EPA-600/7-79-138
TVA/ONR-79/01
REMOTE SENSING OF SULFUR DIOXIDE EFFECTS ON VEGETATION-
PHOTOMETRIC ANALYSIS OF AERIAL PHOTOGRAPHS
by
C. Daniel Sapp
Office of Natural Resources
Tennessee Valley Authority
Muscle Shoals, Alabama 35660
Interagency Agreement EPA-IAG-D8-E721-DJ
Project No. E-AP 80 BDJ
Program Element No. INE 625C
Project Officer
James Stemmle
U. S. Environmental Protection Agency
401 M Street, SW
Washington, DC 20460
Prepared for
OFFICE OF ENERGY, MINERALS, AND INDUSTRY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, DC 20460
-------
DISCLAIMER
This report was prepared by the Tennessee Valley Authority and has
been reviewed by the Office of Energy, Minerals, and Industry, U.S.
Environmental Protection Agency, and approved for publication. Approval
does not signify that the contents necessarily reflect the views and
policies of the Tennessee Valley Authority or the U.S. Environmental
Protection Agency, nor does mention of trade names or commercial products
constitute endorsement or recommendation for use.
11
-------
ABSTRACT
Spectral reflectances were measured by tri-band densitometry of
aerial color-infrared photographs of soybean [Glycine max (L.) Merr.]
fields that had been affected by sulfur dioxide (S02) emissions from
large, coal-fired power plants in northwestern Alabama and western
Tennessee. The photographs were photometrically calibrated.
Results indicate that, at very light levels of foliar injury, the
infrared-to-red reflectance ratio decreased with increasing injury.
This behavior was in accordance with theory. However, at moderate and
severe levels of injury, the ratio increased with injury. The infrared
component increased, and the red component decreased as injury level
rose. Two other ratios of reflectance (infrared-to-green and red-to-
green) did not correlate significantly (a = 0.05) with injury. The best
indicator of crop yield was green band reflectance, but the red and
infrared bands were nearly as good. Ratios produced no significant
correlations with yield. The yield variable actually increased with
the level of injury, apparently because of field-to-field variations in
canopy density.
This report was submitted by the Tennessee Valley Authority, Office
of Natural Resources, in partial fulfillment of Energy Accomplishment
Plan 80 BDJ, under terms of Interagency Agreement D8-E721-DJ with the
Environmental Protection Agency. Work on this phase of the project was
completed as of September 1978.
111
-------
CONTENTS
Abstract
Figures
Tables '.'.'.'.'.'.'.'.'.'.'.'.'.'. vl
1. Introduction •>
General -^
2. Conclusions and Recommendations 2
Conclusions 2
Recommendations 2
3. Methods and Instruments 4
Overflights '.'.'.'.' 4
Photometric analysis 4
4. Results and Discussion g
Colbert test '.'.'.'.'.'' 8
General g
Measurement and comparison of reflectance 8
Johnsonville test ' 12
General ^2
Measurement and comparison of reflectance 12
References 21
Appendixes
A. Explanation of procedure for estimating foliar
effects of S02 on crop species 22
B. Photometric calibration 23
IV
-------
FIGURES
Number Fage
1 Colbert Steam Plant area in northwestern Alabama .... 5
2 Johnsonville Steam Plant area in western
Tennessee 6
3 Generalized relationship of IR/R ratio to spectral
curves of healthy and stressed vegetation 7
4 Flight lines and S02-affected areas near Colbert .... 9
5 Aerial color-infrared photograph showing Colbert
area 10
6 Location of densitometer sampling points within
soybean fields near Colbert Steam Plant 11
7 Regression of reflectance ratios and foliar injury
levels for Colbert soybean fields 14
8 Flight lines and S02-affected areas in
Johnsonville area 15
9 Aerial color-infrared photograph showing selected
soybean fields near Johnsonville Steam Plant 16
10 Statistical regressions of reflectance, injury
levels, and yield for Johnsonville soybean
fields 20
B-l Comparison of D-log E curves — red filter 24
B-2 Comparison of D-log E curves—green filter 25
B-3 Comparison of D-log E curves—blue filter 26
v
-------
TABLES
Number
1 Spectral Reflectance Ratios for Selected Soybean
Fields Near Colbert 13
2 Comparison of Ground Truth with IR/R Ratio
Statistics for Selected Soybean Fields Near
Colbert 13
3 Reflectance, Foliar Injury Levels, and Yield for
Johnsonville Soybean Fields and Bare Soil 18
4 Correlation Coefficients for Reflectance, S02
Injury, and Yield of Soybeans 19
B-l a's and p's for Colbert Photographs of
August 29, 1977 27
B-2 Of's and P's for Johnsonville Photographs 28
B-3 Spectral Reflectance Gradient Constants 29
B-4 Measurements of Reflectance-Field B 30
B-5 Reflectances Corrected for Illumination Gradient .... 30
B-6 Comparison of Photometric Interpretation Accuracy
With and Without Preliminary Illumination
Gradient Correction 31
VI
-------
SECTION 1
INTRODUCTION
GENERAL
The purpose of the project is to study the acute effects of sulfur
dioxide (S02) emissions from large, coal-burning power plants on vege-
tation. The technique of remote sensing—a term used by earth scientists
to describe the study of phenomena from a distance, without having the
sensor in direct contact with the phenomena being sensed—is the tool
chosen to accomplish this purpose.
Problems encountered during traditional field surveying include the
time-consuming nature of field surveillance, inaccessibility of many
areas, and subjectivity of descriptions of foliage color. Remote measure-
ment of reflectance with calibrated instruments, followed by objective
analysis of data in the laboratory, may bypass some of these problems.
The color of a leaf may be described merely as a hue, or it may be
defined objectively by measuring reflected radiation. A spectral plot
of radiant energy reflected by either a plant canopy or an individual
leaf can define (1) the color of the leaf or canopy in the visible spec-
trum and (2) its radiant energy characteristics in the adjacent near-
infrared spectrum. When a plant is stressed by S02 or some other agent,
the shape of its radiant energy curve changes.1
The classical procedure for acquiring radiant energy data in the
field involves making measurements at discrete points with a portable
spectroradiometer.2 Comparable data may also be obtained directly by
making point measurements of the optical density of aerial photographs.
With appropriate calibrations, calculations, and corrections, either
technique can yield spectral reflectance data that may indicate S02~
induced stress. This report describes "photometric analysis," the
aerial photographic technique for acquiring reflectances.
This report is the first of two reports on remote sensing of reflec-
tance characteristics of S02~affected vegetation. The second report
describes spectroradiometry, in which a radiometer is used for laboratory-
based measurements of stressed soybeans, winter wheat, and cotton plants.
Research on a third technique, concerning close-range overhead photography
of test plots of S02~stressed soybeans, was also undertaken. However,
results of these close-range photographic experiments were inconclusive,
and further work is being done before publishing them.
The primary hypotheses to be tested in this study are that (1) spec-
tral reflectances are related to the levels of foliar injury to soybeans
under the experimental conditions imposed on the study, and (2) reflectance
is also related to yield of the crop.
-------
-2-
SECTION 2
CONCLUSIONS AND RECOMMENDATIONS
CONCLUSIONS
Overflights of two coal-fired power plants were made to photograph
foliar effects of S02 emissions on soybeans. Spectral reflectance data
were obtained from the color-infrared photographs, after which the
statistical means and standard deviations for reflectance were compared
with the levels of S02 injury for each field of soybeans.
Several statistically significant (ot = 0.05) relationships between
reflectance and injury and between reflectance and yield were discovered.
1. At very light levels of S02 injury to soybeans, where less than
about 5 percent of the foliage was chlorotic,* the ratio of infrared
to red (IR/R) reflectance decreased with the level of injury, as
predicted. However, at higher levels of injury, where 20 percent
or more of the foliage was chlorotic or necrotic,** the IR/R ratio
increased with the level of injury. This behavior is the opposite
of what was expected. The IR component increased and the R component
decreased as percentage injury rose above 20 percent, thus raising
the ratio. Two other reflectance ratios, infrared to green (IR/G)
and red to green (R/G), did not correlate significantly with injury
levels.
2. Single-band reflectances were not significantly related to injury.
This includes the R, G, and IR bands.
3. The best indicator of yield was G reflectance, which was positively
correlated (significant, ot = 0.05). The R and IR bands were nearly
as good. Ratioing of the spectral bands produced no significant
relationships.
RECOMMENDATIONS
The experiments described in this report involved comparing the
field-to-field variations in reflectance with levels of injury and yield
for each field. The work should be repeated in more detail, concentrating
on a few fields of mature soybeans within which detailed surveys of the
patterns of reflectances and foliar injury are known. This procedure
should isolate many of the masking variables described above.
Because the effects of the masking variables cannot be completely
eliminated, future investigations should concentrate on fields of mature
soybeans that have a nearly continuous canopy and little infestation by
weeds.
*Chlorosis is defined as a visible yellowing of leaf tissue. The
markings may be acute or chronic, permanent or temporary.
**Necrosis is acute injury characterized by marginal or intercostal
areas of dead leaf tissue.
-------
-3-
The relationship of reflectance to yield needs further study. In
particular, the increase in the IR/R reflectance ratio with increasing
injury warrants further investigation, because the findings are the
reverse of current theories published on the subject.
-------
-4-
SECTION 3
METHODS AND INSTRUMENTS
OVERFLIGHTS
Overflights of two of TVA's 12 coal-fired power plants were per-
formed during the 1977 growing season when the foliar effects of S02
were still visible to ground observers. We were focusing on the effects
on soybeans [Glycine max (L.) Merr.], an important cash crop in the
Tennessee Valley, growing near Colbert Steam Plant, in northwestern
Alabama; and Johnsonville Steam Plant, in western Tennessee (Figures 1
and 2).
Aerial color-infrared photographs were obtained from altitudes
ranging from 500 to 1800 m above ground level. Large-format aerial
mapping cameras equipped with 152-mm focal-length lenses were used. The
resulting range of image scales varied from about 1:3000 for the low-
altitude runs to 1:12,000 for the high-altitude runs. TVA used its Wild
RC-8 camera to obtain coverage of Johnsonville, and EPA used one Wild
RC-10 to fly Johnsonville and another to fly Colbert. EPA also flew a
Daedelus multispectral scanner (MSS) over Colbert for TVA. The results
of the MSS data analysis will be published later as a separate report.
PHOTOMETRIC ANALYSIS
The objective of photometric analysis was to detect S02 effects in
soybean fields by measuring differences in spectral reflectance. A
second objective was to relate the reflectance measurements to yield of
soybeans. To do this, systematic errors in the photographs were measured
and eliminated through an image calibration process. The errors resulted
from film processing, atmospheric effects, and variation in illumination.
The reflectance patterns and trends were then compared with ground-truth
data on S02 effects to ascertain whether any relationships existed.
Photometric analysis involved use of the Calspan Scene Color Stand-
ard (SCS) technique. Specific details of the SCS technique have been
published.5 The basics of photometric analysis have been described
expertly by Lillesand in an introductory text.4 Lillesand calls the
technique photographic radiometry. The SCS procedure permits deter-
mination of the atmospheric and illumination variables directly from the
photograph. One advantage is that no a priori knowledge of the reflec-
tances of ground objects is necessary, because all the information for
calculating reflectances is available from the photograph itself.
Reflectances of specific objects, such as individual soybean plants on
large-scale images or integrated spot canopy measurements on small-scale
images, may be calculated from spot density measurements. Two densito-
meter apertures, 1.0 mm and 150 |Jm, were used. The 1-mm densitometer
aperture was adequate for canopy sampling on the l:12,000-scale photo-
graphs, where the spot covered a ground area 12 m in diameter. The
150-|jm aperture was used to make density measurements in shadow areas.
Also measured in the calibration process were images of asphalt surfaces
(roofs and roads) and bare soil. Reflectances from these surfaces are
-------
i
Ln
i
Figure 1. Colbert Steam Plant area in northwestern Alabama.
-------
Figure 2. Johnsonville Steam Plant area in western Tennessee.
-------
-7-
relatively constant temporally and can be used to subtract the effects
of atmosphere and illumination. Reflectances of other objects can be
measured to an accuracy of about 5 percent of their true values.3 This
performance is comparable with that expected from a field radiometer.
The image calibration procedure requires rigorous control of vari-
ables. A density step wedge was processed with the film so that film
processing effects could be measured. This procedure used sensitometric
curves (D-log E) that plot density against relative log exposure values
(Appendix B). Tri-band (analytical) densities were then measured directly
from the film and converted to changes in relative exposure.
Reflectances can be obtained from each of the three spectral bands
comprising color film, whether the emulsion is true color or color-infrared.
The spectral coverage of a particular band is determined by the sensitivity
of that component of the film emulsion. For color-infrared film, the
emulsions are sensitive to either green, red, or near-infrared wavelengths.
Both densitometers were equipped with selectable filters (nos. 92, 93,
94, and 106), enabling the operator to measure density in any one or all
three emulsion layers.
Reflectance values from one band may be divided by reflectance
values from other bands. The ratio thus obtained may be a sensitive
indicator of stress. Three simple ratios were calculated in this study:
infrared to red (IR/R); infrared to green (IR/G); and red to green
(R/G). Figure 3 illustrates the relationship of the IR/R ratio to
stress as shown by spectral curves of reflectance. The ratio approaches
1.0 as the curve flattens with stress from S0£ or some other agent.
The spectral changes that occur in plants in response to stress
have been described by Murtha.1 We investigated the three single-band
reflectances and the three reflectance ratios described above to determine
whether they correlated with levels of foliar injury to soybeans.
w
u
WAVELENGTH
HEALTHY
w
u
H
O
W
W
WAVELENGTH
STRESSED
Figure 3. Generalized relationship of IR/R ratio to spectral curves of
healthy and stressed vegetation. "R" indicates a reflectance
measurement in the red region, and "IR" indicates a reflectance
measurement in the near-infrared region.
-------
-8-
SECTION 4
RESULTS AND DISCUSSION
An operational test of the photometric analysis technique was
performed to determine whether it could be used to derive spectral
reflectances at sample points within selected S02-affected soybean
fields. The fields near Colbert exhibited foliar effects that were
generally below 10 percent* and in the very light to light category. In
contrast, Johnsonville had effects ranging from very light to severe.
There were also coexisting foliar effects induced by herbicides or
drought rather than 803. Fortunately, ground-truth data were available
for the fields.
COLBERT TEST
General
Exposures to S02 caused visible foliar injury to soybeans in the
project area on August 3 and August 26, 1977 (Figure 4). The Environ-
mental Protection Agency (EPA) was asked to perform an overflight after
the first exposure; but because of concurrent requests, the aircraft did
not arrive until August 29. Drought-induced senescence and, in areas of
adequate moisture, growth of the soybean canopy and weeds tended to
dilute the S02 effects. Effects from the August 26 injury were still
fresh. An attempt was made to distinguish and separate these types of
stress through photometric analysis of the aerial photographs.
Analysis of the Colbert photographs focused on five soybean fields
that fell within a single frame (Figure 5). Four of the fields were
affected by S02 and one was unaffected. The photographs also show a set
of six test panels (arrow) for calibrating reflectance in the photometric
analysis procedure. The soybeans were mature (generally 7 to 10 nodes)
plants that had stopped showing new growth, but were not yet senescent.
Some areas were infested with cockleburs [Xanthium strumarium (L.)].
The soybean canopies were generally continuous, with only a few areas of
soil showing from overhead. The effects of S02 exposure consisted of
light levels of chlorosis, but no necrosis.
Measurement and Comparison of Reflectance
The sample points were selected systematically within each field
(Figure 6), and the densitometer was used to measure the optical density
of the photograph at each point. Exceptions to the uniform spacing of
the sample points were made when required by the shape of a field. The
''^Calculations of injury used the LAP method (Appendix A). Foliar injury
was classified as follows: light injury, up to 10 percent; moderate
injury, 11 to 25 percent; severe injury, over 25 percent. Very light
injury was the lower half of the light category.
-------
COLBERT STEAM PLANT
I
VD
Figure 4. Flight lines and S02-affected areas near Colbert. (Boxed
numbers locate fixed 862 monitoring stations.)
-------
-10-
Figure 5. Aerial color-infrared photograph showing Colbert area. (Letters
identify selected soybean fields discussed in report. Scale
1:12,000).
-------
-11-
0
<
0.5km
Figure 6. Location of densitometer sampling points within
soybean fields near Colbert Steam Plant.
-------
-12-
mean and standard deviations of the reflectance values from each spectral
band with foliar injury levels obtained from ground-truth indicated poor
correlations. The next step was to ratio the near-infrared (IR), the
red (R), and the green (G) reflectances (Table 1). The IR/R ratio pro-
vided the best separation of affected and unaffected soybeans. The
S02-affected soybeans had lower IR reflectance and higher R reflectance.
This finding is in accordance with published theories on the behavior of
stressed plants.1
Comparison of the foliar injury data with the IR/R reflectance
ratios for the five fields (Table 2) showed that the field with the
highest mean reflectance (A) was unaffected by S02; field B, with 2
percent injury, had the next highest mean; and field C, with the highest
injury level, had the lowest mean reflectance. Therefore, it appears
that the smaller the IR/R ratio, the greater the S02 effects. Figure 7
further illustrates the relationship between injury and each reflectance
ratio. Field D was infested with weeds and had a discontinuous canopy;
this heterogeneity is reflected in the high standard deviation (1.11)
for reflectance for the field. A low standard deviation for the reflec-
tance values from a field indicates a homogeneous canopy and few weeds.
JOHNSONVILLE TEST
General
Two incidents of S02 injury to vegetation occurred in the photo-
graphed Johnsonville area during July 1977. The effects were classified
generally as light to moderate. The earliest incident occurred on July
3 in an area northwest of the plant (Figure 8). The effects persisted
and were photographed by EPA on July 21. TVA acquired ground truth and
obtained a duplicate copy of the film from the EPA Vint Hill Farms
Station. Another S02 incident occurred on July 23 in the same general
area. Injury to soybeans was still visible in the field on August 2,
the date of the TVA overflight.
The photometric analysis focused first on 15 soybean fields in the
Johnsonville area where the S02 plume contacted the crop. The number
was later reduced to 9 because of differing stages of growth. One
aerial color-infrared photograph (Figure 9) shows some of the fields.
As with the Colbert data analysis, the field-to-field variations in
reflectance ratios were studied. Next, the relationship between the
ratios and the S02 injury levels was investigated. Finally, the relation-
ships among reflectance, S02 injury levels, and yield were explored.
Measurement and Comparison of Reflectance
Reflectance values from soybean fields located northwest of the
Johnsonville Steam Plant were compared to determine whether they were
related to foliar injury levels. The S02 effects ranged from very light
to severe, providing a full scale for study. The affected area contained
no-till fields, tilled (plowed) fields, and barren (unplanted but recently
plowed) fields. Some fields contained mature soybeans, whereas others
-------
TABLE 1.
-13-
SPECTRAL REFLECTANCE RATIOS FOR SELECTED
SOYBEAN FIELDS NEAR COLBERT
Field
a
no.
A
B
C
D
E
IR/Rb
Mean
5
5
3
4
4
.77
.34
.01
.55
.91
0
0
0
1
0
S.D.
.338
.676
.486
.109
.569
Mean
5.75
4.99
3.67
4.47
4.96
IR/Gb
0
0
0
0
0
S.D.
.128
.407
.399
.763
.444
R/Gb
Mean
0
0
1
1
1
.99
.94
.12
.01
.02
0
0
0
0
0
S.D.
.052
.062
.199
.112
.052
, Keyed to Figure 6.
Abbreviations: IR--infrared reflectance; R--red; and
G—green.
TABLE 2. COMPARISON OF GROUND TRUTH WITH IR/R RATIO
STATISTICS FOR SELECTED SOYBEAN FIELDS NEAR COLBERT
Observed level
Field
a
no.
A
B
C
D
E
of injury
0
2
4
2C
2
Mean
5.7
5.3
3.0
4.6
4.9
IR/R
S.D.
0.34
0.68
0.49
1.11
0.57
,Keyed to Figure 6.
Field estimate based on degree and extent of injury to
foliage in affected parts of field (L x A method—see
Appendix A).
Estimated because of conflicting data.
-------
6.0-
cr
4.0-
6.0-
o:
4.0-
0.0 2.0 4.0
INJURY(%)
r = -0.96!
0.0 2.0 4.0
INJURY(%)
CD
-\
o:
1.0-
r=0.699
0.0
2*0
INJURY(%)
4.0
Figure 7. Regression of reflectance ratios and foliar
injury levels for Colbert soybean fields.
-------
-15-
Figure 8. Flight lines and S02-affected areas in Johnsonville area.
-------
-16-
Figure 9. Aerial color-infrared photograph showing selected soybean fields
near Johnsonville Steam Plant (scale 1:12,000). Acquired by EPA
on July 21, 1977, from 1800 m above ground level.
-------
-17-
had young plants. Some fields had been planted twice. Weeds, mainly
cockleburs, were prevalent in some of the no-till fields. Some no-till
fields contained wheat stubble between the bean plants.
The image densities were measured, and the values were converted to
percentage reflectance. Means, standard deviations, and ratios were
computed (Table 3 and Appendix B). Because previous analyses indicated
the relative superiority of the IR/R ratio, this ratio was used instead
of the R/G and IR/G ratios in the search for a relationship between
reflectance and injury levels.
Correlation coefficients were generated for reflectance, S(>2 injury,
and yield (Table 4). The IR/R increased with the level of injury, the
opposite of what was expected (Figure lOa). The IR component of the
ratio increased and the R component decreased with injury (Figures lOb
and c). The G reflectance also decreased with increasing injury (Figure
lOd). None of the three single-band reflectance correlations were
significant (a = 0.05), but the IR/R ratio was significant. The cor-
relation coefficients are shown on the illustration.
The best indicator of yield was the G reflectance (Figure lOe).
The R and IR reflectances (Figures lOf and lOg) were also good, and all
three correlations of yield and reflectance were significant (a = 0.05).
The IR/R ratio results (Figure lOh) were not significant. Yield increased
with injury (Figure lOi). This last correlation, which may appear sur-
prising at first, is explained by the association of yield with canopy
density. The density variable, as measured from overhead photographs,
is associated with stage of growth, availability of soil moisture, soil
fertility, and many other factors that affect plant conditions before,
during, and after an incident of exposure to S02. Certainly, the more
dense canopy is associated with higher yield. Injury level was apparently
not a sufficiently powerful factor to override the density (and therefore
yield) factor. More study is needed to define the relationship between
foliar injury levels and yield before reflectance can be used as a
surrogate for yield.
The relationship between the effects of power plant emissions and
productivity of crops is not well known. In general, crop yields are
not affected by S02 exposure unless visible foliar effects occur.
Evidently, over 5 percent of the leaf area must be affected to measurably
reduce yield.7 Common practice for estimating yield reduction involved
(1) field sampling to determine the percentage of leaf area destroyed
and (2) applying an empirically or theoretically derived factor to
calculate loss. The difficulty is that many uncontrollable cultural,
edaphic, and climatic factors also affect yield. Certainly, the stage
of growth at the time of exposure to S02 is one of the more important
factors. Significant reductions in yield caused by S02 exposures during
the pod-filling stage of growth in soybeans has been documented and
related to the amount of foliar chlorosis.8 However, other exposures to
soybeans during the prebloom stage did not reduce yield significantly.9
-------
TABLE 3. REFLECTANCE, FOLIAR INJURY LEVELS, AND YIELD FOR
JOHNSONVILLE SOYBEAN FIELDS AND BARE SOIL
Field
designation
F
G
H
I
J
K
L
Bare soil
IR
Mean
21.1
23.2
24.8
24.0
17.9
22.6
25.7
20.6
S.D.
3.7
4.5
2.8
4.3
1.7
1.8
1.0
1.9
R
Mean
6.4
6.5
7.3
6.8
6.6
6.7
4.6
17.1
G
S.D.
1.0
0.9
0.8
0.7
0.6
0.6
0.8
2.0
Mean
7.2
7.5
7.9
7.8
7.0
7.8
4.7
13.7
S.D.
0.8
1.3
0.6
0.6
0.7
0.5
0.6
1.0
Observed level
IR/R of injury3 Yield
Mean
3.3
3.6
3.4
3.5
2.7
3.4
5.6
1.2
S.D. (%)
0.66 34.8
0.73 44.6
0.55 20.2
0.65 30.8
0.35 20.0
0.46 22.0
0.95 14.5
0.10
bu/acre m
22.5 0
23.4 0
28.0 0
28.5 0
22.4 0
24.9 0
31.0 0
0
a/ha
.300
.312
.373
.380
.299
.332
.413
.000
kg/ha
3676b
3823b
4574
4656
3659
i
4068 £
i
5064C
0
o
, L x A method
"•V-: ^T A — i/o -
(see Appendix
A.)
c •
Tilled field; all others are no-till.
-------
-19-
TABLE 4. CORRELATION COEFFICIENTS FOR REFLECTANCE,
S02 INJURY, AND YIELD OF SOYBEANS3
Injury
Yield
Gb
(-0.125)
0.860
Rb
(-0.574)
0.806
IRb
(0.209)
0.784
IR/Rb
(0.565)
0.480
Injury
-
0.265
of = 0.05, except as noted by coefficients in parentheses, which
are not significant at this level.
Abbreviations: G = green band reflectance; R = red; IR = infrared;
IR/R = ratio of infrared to red reflectance.
"Injury is the percentage for fields calculated by the L x A x P method
(Appendix A). Yield data were collected in field at harvest.
-------
-20-
r = 0.565
3.5-
3.0-
20.0 30.0 40.0
INJURY
r* 0.209
25.0-
20.0-
20.0 30.0 40.0
INJURY
r=-0.574
.7.0-
6.5-
20.0
30.0
INJURY
40.0
a 7.5-
7.0'
r--O.I25
20.0
30.0
INJURY
40.0
5000-
J4000-
r= 0.860
72 7.6
G
5000-
4000-
r = 0.806
6.4
68
R
5000-
1
04000-
r= 0.784
16.0 20.0 24 0
IR
5000-
•3
i
34000-
r = 0.430
2.8
32
IR/R
3.6
500O-
3 4000-
r= 0.265
20.0
30.0
INJURY
40.0
Figure 10. Statistical regressions of reflectance, injury levels,
and yield for Johnsonville soybean fields.
-------
-21-
REFERENCES
1. Murtha, P. A. Remote Sensing and Vegetation Damage: A Theory for
Detection and Assessment. Photogramm. Eng., 44(9):1148-50, 1978.
2. Reeves, R. G., Abraham Anson, and David Landen, eds. Manual of
Remote Sensing (1), 1975. p. 845.
3. Piech, K. R., and J. E. Walker. Interpretation of Soils. Photogramm.
Eng., 38(l):87-94, 1974.
4. Lillesand, T. M. An Introduction to Photographic Radiometry and
Spectral Pattern Recognition. State University of New York,
Syracuse, New York, 1976. p. 1.
5. Walker, J. E. Personal communication, 1977.
6. Eastman Kodak Company. Infrared and Ultraviolet Photography.
Part 2, Applied Infrared Photography, Tech. Pub. M-27/28-H.
Rochester, New York, 1972.
7. Barrett, T. W., and H. M. Benedict. Sulfur Dioxide. In: Recogni-
tion of Air Pollution Injury to Vegetation. Air Pollution Control
Association, Pittsburgh, Pennsylvania, 1970. pp. C1-C17.
8. Jones, H. C., Frances P. Weatherford, W. S. Liggett, Jr., and J. R.
Cunningham. Effect of Foliar Injury Caused by Exposure to Sulfur
Dioxide on Yield of Soybeans - Results of a Large Scale Field
Investigation. Division of Environmental Planning, Tennessee Valley
Authority, Muscle Shoals, Alabama. Manuscript in preparation.
9. Jones, H. C., J. R. Cunningham, S. B. McLaughlin, N. T. Lee, and
Shirley S. Ray. Investigation of Alleged Air Pollution Effects on
Yield of Soybeans in the Vicinity of the Shawnee Steam Plant.
E-EB-73-3, Division of Environmental Planning, Tennessee Valley
Authority, Muscle Shoals, Alabama, 1973. 36 pp.
-------
-22-
APPENDIX A
EXPLANATION OF PROCEDURE FOR ESTIMATING FOLIAR EFFECTS
OF S02 ON CROP SPECIES
L = Percentage of leaves affected on an average affected
plant
A = Percentage of leaf area affected on an average affected
leaf
P = Percentage of plants affected in field
L x A = Percentage of total leaf area affected on an average
affected plant
LxAxP=T= Average of total leaf area affected in a given field
-------
-23-
APPENDIX B
PHOTOMETRIC CALIBRATION
The atmospheric and illumination contributions to the total exposure
of an image on film are, by definition,1
(1) a multiplicative attenuation of the energy for every object in
the scene (a) and
(2) an additive contribution of energy from the optical path between
the objects and the sensor (p).
To measure these two parameters (a and P), one must first derive
the relationship between image density (D), which is measured from the
images, and the exposing energy (E) that caused the image density. This
is accomplished by using a step wedge (series of known exposures) that
has been imaged on the film before it is processed. TVA furnished one
set of EPA aerial photographs with a step wedge (Colbert) and two sets
of TVA photographs without wedges (Johnsonville). The densities of the
steps in the Colbert wedge were measured and plotted against their known
relative log exposure values. The resultant curves are shown in Figures
B-l, B-2, and B-3 for the infrared energy (red filter), red energy
(green filter), and green energy (blue filter). The Eastman Kodak
Handbook curve for CIR film was also plotted. The comparison of these
curves showed a significant difference in the low-exposure end, with
severe density compression present in the EPA film. Examination of both
the wedge and scenes under SOX magnification revealed a microscopic
pattern generally related to processing to correct for underexposure
during data collection. This processing problem could manifest itself
in several ways in photometric analyses if the analyst were not aware of
its presence.
The first manifestation could be in the photometric calibration
process to derive the additive contribution to image exposure (p). This
task is accomplished by performing a regression analysis of the exposures
of dark and light objects in the scene illuminated by skylight (shadow)
and also by sunlight plus skylight. Because of density compression due
to processing, the exposure range between objects illuminated by skylight
and those illuminated by skylight plus sunlight is reduced. Thus, the
plotting scale used in the computerized regression program to derive P
must be expanded to obtain meaningful values for p. This actually
occurred in the first attempt to derive p for the Colbert data, with
negative p's being output by the regression program. Expanding the
plotting scale, however, did allow the derivation of meaningful p's.
The expanded scale and plot allowed the operator to recognize erroneous
data points and eliminate them from the analysis, an important interactive
step in calibration.
1. Piech, R., and J. R. Schott. Atmospheric Corrections for Satellite
Water Quality Studies. Proceedings of the SPIE, 51:84-89, 1974.
-------
o
o
3.0l
2.0
-24-
o HANDBOOK
• EPA
RED FILTER
_L
o • .
I
o o
0.0 1.0 2.0
RELATIVE LOG EXPOSURE
Figure B-l. Comparison of D-log E curves — red filter.
-------
-25-
3.0
2.0
H
1.0
O HANDBOOK
• EPA FILM
GRKKN FILTER
_L
0.0
1.0 2.0
RELATIVE IDG EXPOSURE
Figure B-2. Comparison of D-log E curves--green filter.
-------
-26-
o
3.0
H
CO
2.0
1.0
0.0
o
o HANDBOOK
• EPA FILM
BLUE FILTER
•
O
1.0
RELATIVE LOG EXPOSURE
2.0
Figure B-3. Comparison of D-log E curves—blue filter.
-------
-27-
Another manifestation that could occur is that the density of
healthy soybean fields located in the extreme corners of the film format,
especially in the red information band (green filter-chlorophyll absorp-
tion band), could reach the maximum density limit imposed by the pro-
cessing problem; even with correction for lens fall off, an incorrect
value of red reflectance could result. This did not occur in the case
of Colbert data, but the analyst should be aware that processing problems
can result in maximum density compression.
A third manifestation to be aware of, because of this type of
processing problem, is the effect of the microscopic pattern on densito-
metry. If a large-aperture (1 mm) densitometer is used, it integrates
the variations in density within the microscopic pattern anomaly with
little or no effect on the densitometry. However, if a densitometer
with a small sample aperture (50 (Jm) is used, the placement of the
aperture is critical. The processing pattern consists of circular
areas, about 100 |Jm in diameter, over the entire film format. In the
center of each circular area of normal dye concentration is an abnormal
"snowflake" pattern that is generally of higher density than the normal
dye concentration surrounding the "snowflake." Therefore, the analyst
must be sure he places the small aperture in the normal dye area and not
on a "snowflake" when making an image density measurement.
Fortunately, the Colbert photographs were acquired in late August,
when most soybean crops had reached full cover condition. Therefore, it
was possible to use the 1-mm aperture and have the effect of the proces-
sing problem averaged into the noise of the overall measurement process.
The a's and p's derived for the Colbert photographs are shown in
Table B-l.
TABLE B-l. a'S AND P'S FOR COLBERT PHOTOGRAPHS OF AUGUST 29, 1977
Infrared band Red band Green band
a
^
206.92 157.14 174.70
6.72 3.45 10.37
For the Johnsonville photographs it was necessary to assume charac-
teristic curves to obtain a's and (5's. The Johnsonville films were
examined under 30X magnification to determine whether the microscopic
pattern present in the Colbert wedge and scenes was still present. The
pattern was not present in either the July 13, 1977, films furnished by
EPA .to TVA or in the August 2, 1977, TVA films. The maximum densities
of these films were checked (i.e., zero exposure in borders), with the
expectation that they would be significantly higher. However, because
the maximum densities were not found to be significantly higher, we
-------
-28-
decided to use the original wedge data for the Colbert scene to derive
a's and P's for Johnsonville. The a's and f3's for the two sets of
Johnsonville films are given in Table B-2.
TABLE B-2. a1S AND 6'S FOR JOHNSONVILLE PHOTOGRAPHS
Infrared band Red band Green band
July 13, 1977
a 322.588 130.486 138.611
P 13.39 3.21 4.27
August 2, 1977
Of 329.139 214.169 219.411
P 7.15 3.28 3.45
The assumptions necessary to obtain a photometric calibration of
these films were pointed out to the TVA representative. The necessity
for the exposure of a 21-step wedge on the original aerial films and a
duplicate of this wedge for subsequent duplicates of scenes to be analyzed
in any future field experiment was also emphasized. This requirement
was the first experimental design criterion resulting from this effort.
Because of the problems of sensitometric and densitometric control
for the aerial photographs, absolute spectral reflectances and reflec-
tance ratios could not be obtained from photometric interpretation;
however, important relative reflectance information within any one scene
could still be obtained, depending on the level of reflectance. The
film covering the Colbert site (frames 5353 through 5355, August 29,
1977) was selected for analysis because it did have a wedge. Thus,
these reflectances would be closest to absolute values.
The TVA representative measured the spectral reflectance and
reflectance ratio properties of soybean plants in five fields at the
Colbert site (fields A through E). The number of measurements per field
was limited by the field of view of the densitometer used. This was a
1-mm-diam aperture, representing a 12-m-diam area on the ground. A
0.6-m-diam area could have been used, but this was not considered necessary
for these fields because they were mature fields with very full cover.
The 12-m-diam area still allowed about 50 samples to be measured per
field in less than 15 min.
This is considered a very important point relative to the comparison
of any potential aerial photometric method of soybean stress measurement
to a ground survey method. It would be very difficult to obtain a
sample rate of 200 samples/h per field when using a ground survey
method, whether they are spectral reflectances from a ground-based TSR
-------
-29-
or L x A x P (see Appendix A) estimates. This is a very practical
advantage of an aerial photographic approach. Furthermore, the photo-
graph would provide a valuable permanent record of the crop condition.
The TVA representative compared the mean IR/R reflectance ratios for
these fields and found the most injured (4 percent by the L x A x P method)
plants had a ratio of 3.010.49, whereas the least injured (0 percent) plants
had a ratio of 5.7±0.34. In previous photometric interpretation analyses
of vegetation stress (chlorosis-necrosis) by Calspan, the mean IR/R reflec-
tance ratio has always been high when stress was low and low when stress
was high. Thus, even with known processing problems, the reflectance and
ratio results obtained were consistent with previous studies.
However, one inconsistency was noted. The standard deviation in the
red (chlorophyll absorption) band is usually extremely large in comparison
with that in the infrared band. These data showed the infrared band to
have the larger standard deviation. The presence of a sunspot image in
the Colbert scene could have caused these inconsistencies.
Field B, having 2 percent injury by the L x A x P method, appeared
on three frames (nos. 5353, 5354, and 5355) at different format positions
relative to the sunspot and center of the format. Spectral reflectance
gradient functions from the center of the sunspot image through the center
of each format to the edge of the film away from the sunspot were derived
from regressions of soybean reflectances against distance from the center
format. Deciduous tree reflectance was also regressed relative to distance
with almost identical results. This result would be expected since both
canopies are highly textured and contain multiple reflectances.
A second-order regression resulted in a functional relation with high
correlation coefficients (r2):
RA = B2X2 + BjX + B0,
where X = the distance of an image from the center of the film format (0)
to (+) the sunspot and away from (-) the sunspot, in inches along the
sunspot, center format line.
The constants BO, Bj, and B2 are shown in Table B-3, with the
correlation coefficients for each spectral band.
TABLE B-3. SPECTRAL REFLECTANCE GRADIENT CONSTANTS
RIR RR RG
B0 0.0248 0.0575 0.0539
Bj 0.0139961 0.008437 0.0076
B2 0.001159 0.0007353 0.0008
r2 0.728 0.879 0.865
-------
-30-
Eight to nine spectral reflectance measurements were made in
field B on each of the three frames on which the field was imaged.
The results of the measurements are shown in Table B-4.
TABLE B-4. MEASUREMENTS OF REFLECTANCE--FIELD B
Frame
no.
5353
5354
5355
All data
RIR
29.26
28.89
24.88
27.5
a(%)3
±7.5
±3.0
±5.0
±9
RR
05.65
05.76
04.51
05.14
a(%)a
±3.4
±6.0
±4.0
±5.3
RG
06.14
05.83
04.38
05.38
a(%)a
±4.1
±7.0
±5.0
±15.8
RIR/RR
5.2
5.4
5.5
5.4
a(%)a
±8
±4
±2
±6
Standard deviation, as percentage of mean reflectance.
Next, a preliminary gradient correction method under development
at Calspan was applied to the spectral measurements. The corrected
reflectance data for the field are shown in Table B-5.
TABLE B-5. REFLECTANCES CORRECTED FOR ILLUMINATION GRADIENT
Frame
no.
5353
5354
5355
All data
RR
31.96
32.60
31.67
32.06
a(%)a
±7.6
±3.3
±4.7
±5.2
07
07
08
07
RR
.01
.10
.04
.37
a(%)a
±3.3
±6.0
±4.1
±7.9
07
07
06
07
RG
.39
.44
.94
.24
a«)a
±4
±7
±5
±6
.1
.4
.2
.4
RIR/RR
4.5
4.6
3.0
4.3
a(%)a
±8
±4
±2
±9
o
Standard deviation, as percentage of mean reflectance.
-------
-31-
The preliminary correction for illumination gradient normalizes all
reflectances to a common point in the scene--in this case, the sunspot
location. Theoretically, the mean values of reflectance for field B
should be identical on all three frames. Therefore, we can compare the
difference of the mean reflectance values for any single frame with the
mean reflectance for all frames, as a measure of the accuracy of the
photometric interpretation method with and without the correction for
illumination gradient. The results of such a comparison are shown in
Table B-6. There can be no doubt from these results that a correction
provides a major improvement.
TABLE B-6. COMPARISON OF PHOTOMETRIC INTERPRETATION ACCURACY
WITH AND WITHOUT PRELIMINARY ILLUMINATION GRADIENT CORRECTION
Frame
no.
5353
5354
5355
R
With
0.3
2
1
jR(%) RR(%)
Without
6
5
10
With
5
4
9
Without
10
12
12
RG(%)
With
2
3
4
Without
12
8
18
Accuracy in % = RX (all data) - RX (field) x 100
RA (all data)
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/ 7-79-1 38
2,
4. TITLE ANDSUBTITLE
REMOTE SENSING OF SULFUR DIOXIDE EFFECTS ON
VEGETATION—PHOTOMETRIC ANALYSIS OF AERIAL
PHOTOGRAPHS
7. AUTHOR(S)
C. Daniel Sapp
9. PERFORMING ORGANIZATION NAME M
Office of Natural Resource.
Tennessee Valley Authority
Chattanooga., TN 37401
JD ADDRESS
3
12. SPONSORING' AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Research & Development
Office of Energy, Minerals & Industry
Washington, B.C. 20460
3. RECIPIENT'S ACCESSION- NO.
5. REPORT DATE
June
1979
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO.
TVA/ONR-79/01
10. PROGRAM ELEMENT NO.
INE 625C
11. CONTRACT/GRANT NO.
80 BDJ
13. TYPE OF REPORT AND PERIOD COVERED
Mil estone
14. SPONSORING AGENCY CODE
EPA/600-17
15. SUPPLEMENTARY NOTES
This project is part of the EPA-planned and coordinated Federal Interagency
Energy/Environment R&D Program; Project Officer: James Stetnmle
16. ABSTRACT
Spectral reflectances were measured by tri-band densitometry of aerial color-infrared
photographs of soybean [Glycine max (L.) Merr.3 fields that had been affected by
sulfur dioxide (SOj emissions from large , coal-fired power plants in northwestern
Alabama and western Tennessee. The photographs were photometrically calibrated.
Results indicate that, at very light levels of foliar injury, the infrared-to-red
reflectance ratio decreased with increasing injury. This behavior was in accordance
with theory. However , at moderate and severe levels of injury, the ratio increased
with injury. The infrared component increased , and the red component decreased as
injury level rose. Two other ratios of reflectance (infrared-to-green and red-to-
green) did not correlate significantly (a = 0.05) with injury. The best indicator
of crop yield was green band reflectance, but the red and infrared bands were nearly
as good. Ratios produced no significant correlations with yield. The yield variable
actually increased with the level of injury, apparently because of field-to-f ield
variations in canopy density.
17. (Circle One or More)
a. DESCRIPTORS
Environments
Geography
Other:
13. DISTRIBUTION STATEMENT
Release to Public
KEY WORDS AND DOCUMENT ANALYSIS
b. IDENTIFIERS/OPEN ENDED TERMS
Transport Processes
Charac . Meas . & Monit .
19. SECURITY CLASS (This Report)
Unclassified
20. SECURITY CLASS (This page)
Unclassified
c. COSATI Field/Group
6F 8A 8F
8H 10A 10B
7B 7C 13B
21. NO. OF PAGES
31
22. PRICE
EPA Form 2220-1 (9-73)
------- |