FIG.
1. Map of the United States showing locations where soil samples were collected.
H. R. Co DIT
Eastman Kodak Company
Rochester, N. Y. 14650
The Spectral Reflectance
of American Soils
The characteristic can be predicted with sufficient accuracy from measurements
made at only five wavelengths.
] NTRODUCTIOX
has been expressed recen tly in the iden ti fication of terrestrial
features in aerial photography, and it is well
known that the government is sponsoring
several research projects pertaining to this
endeavor. Most of the land area of the earth's
surface consists of soils and their color represents a basic characteristic for use in their
iden tification. Therefore, spectrophotometric
curves of soil reflectance should be of particular interest in studies relating to the use
of aerial photography in the Earth Resources
M
UCH
INTEREST
Survey to be sponsored by XASA and other
agencies. To assist in this project, spectral
reflectances extending from 320 to 1,000
nanometers have been obtained from a large
proportion of the 285 samples of surface
soils (to a depth of about two inches) collected from 36 states. These samples represent
a wide variation in both color and reflectance.
The locations where they were collected are
shO\\'n in Figure 1. I n most instances the
samples collected were representative of the
soil most prominently evident in that particular area.
955
956
PHOTOGRAMMETRIC ENGINEERING
MEASUREME TS
Measurements extending from 320 to 800
nm were made with a Cary Recording Spectrophotometer Model 14 equipped with an
attach men t for the measuremen t of diffuse
reAectance. Measurements were extended
from 800 to 1,000 nm by use of a manually
operated Beckman DU Spectrophotometer,
also equipped with an attachment for making
diffuse reAectance measurements. \iVith both
shape of the curves can be classified into three
types. Type 1, represen ted by the spectral
reAectance of a chernozem-type soil collected
20 miles east of Lincoln, Tebraska, is illustrated in Figure 2. The distinguishing feature
of these curves is that over any range of wavelengths the slope, with minor exceptions,
either increases or is nearly constant. In the
majority of examples of this type the slope increases throughout the whole spectral range
ABSTRACT: Spectral reflectances extending from 320 to 1,000 nanometers have
been obtained for 160 soil samples collected from 36 states. Measurements were
made of both wet and dry samples, which vary widely in color and reflectance. An
examination of the 160 sets of curves indicates that they can be classified into
three general types with respect to their curve shape. A characteristic vector analysis was made of the spectral reflectance data; it showed that by linear combinations of four vectors and the mean curve, each set of data could be reconstituted to
a high degree of accuracy. Theoretically then, the reflectance data for all wavelengths should be predictable from measurements made at as few as four wavelengths. Empirical regression equations have been derived which relate spectral
reflectance data at 35 wavelengths spaced at 20-nanometer increments to measurements made at only five specially selected wavelengths. To the extent that
soils may be identified by their -reflectance characteristics, this abridged technique
seems to have sufficient accuracy for the 160 samples which have been measured.
instruments the sample was illuminated at
normal incidence, measurements being made
at an angle of 45°.
Each sample was measured in both a wet
and a dry condition. \iVetting was accomplished by adding a fine spray of water to the
sample until it was almost-but not quite-saturated, after which spectral reAectances
were determined. The sample was then oven
dried at 110°F and another set of measurements were made.
A1l soil samples were thoroughly dried before they were prepared for measurement. In
some instances the soil dried in the form of
clumps or aggregates which were reduced in
size by placing them ina mortar and pressi ng
the pestle agains{ them with sufficient pressure to separate the aggregates into individual and sma1l groups of particles. The
samples were placed in holders containing a
circular cavity It inches in diameter and 3/8
inch in depth. The surface texture was made
relatively smooth by first applying a moderate
amount of pressure to the sample by means of
a Aat piece of Plexiglas fo1lowed by sliding
the Plexiglas across the top of the holder.
from 320 to 1,000 nm, as is shown in Figure 2.
In the other examples the slope increases in
the ultraviolet, blue and green region but is
nearly constant in the red and infrared region.
Most samples having Type 1 general shape
are those wi th rather low reAectances.
Type 2 is represented by the curves shown
in Figure 3, which are from a pedalfer-type
sil t co1lected 15 miles southeast of Hot Springs,
Arkansas. I n this type the reAectance increases fairly rapidly, especia1ly for the dry
curve, from 320 to about 450 nm, where a
70
60
~
c 50
~
~
40
'"
30
Gi
~
DRY
20
WET
10
0
300
CLASSIFICATJOX BY CURVE SHAPE
An examination of the 160 sets of curves
thus far obtained indicates that the general
SOil
20 mites E. of lincoln, Neb.
400
500
600
700
800
900
1000
Wavelength (nm)
FIG.
2. Type 1 curves for a sample of chernozem-type soil.
957
THE SPECTRAL REFLECTANCE OF AMERICAN SOILS
T
SILT
60
15 miles S.E. of Hot Springs. Arlc.
"~
50
DR'
~
"
40
Qj
'"
30
WET
20
10
o
300
400
I
I
I
I
500
600
700
800
900
1000
Wavelength (nm)
FIG.
3. Type 2 curves for a sample of pedalfertype silt.
slight or evcn moderate dip in the slopc occurs, followed by an incresae in the slope at
about 480 nm. At about 580 nm another decrease in the slope occurs. From 600 to about
700 nm a slight-to-moderate dip in the slope
is generally present. At about 750 nm the
slope decreases again. Beyond 780 nm the
slope usually changes very little with increasing wavelength. A few examples have been
found where the decrease in the slope at 750
nm is completely absent.
Type 3, illustrated in Figure 4, is from a red
quartz and calcite sand collected from Monument Valley, Utah. In this type the slope of
the curve increases at a moderate rate from
the ultraviolet region to about 530 nm, then
rises sharply to about 580 nm, where a definite decrease in the slope occurs. From about
620 to about 740 nm a slight-to-moderate dip
in the slope is usually present. At 740 nm another definite decrease in the slope occurs,
often dropping to or near zero. In about half
of the 33 sets of curves thus far obtained
70 ,----,------r---,--,------.----r--,---
70
~
60
SAND
60
ec so
which represent this type of curve shape, the
slope was found to rise again at longer wavelengths, as ill ustrated in Figure 4. I n the
other sets of curves of this type there occurred
a decrease not only in slope but in reRectance
from 760 to about 880 nm. Beyond 880 nm the
slope increased with increasing wa\·e!ength.
Such an example is illustrated in Figure 5,
which is the spectral reRectance of a lateri tetype soil collected 5 miles north of Charlottesville, Virginia. The curves given in this figure
and in Figure 4 are quite similar in shape
from 320 to about 750 nm, particularly in the
case of the dry samples. But, as shown in
Figure 5, the reRcctance decreascd from 760
to 880 nm with an increase in reflectance occurring from 880 to 1,000 nm. The necessity
for extending the spcctral reRectance measurements to at lcast 1,000 nm can be seen
from the fact that while the tlVO samples are
quite similar through thc ultraviolet and visible regions of the spectrum, they are quite
different in the infrarcd.
A numcrical value of the difference in curve
shape between Type 2 and Type 3 may be obtained by comparing the ratio in reflectance
measured at 760 and 500 nm of the dry and
wet curves for each type. From measuremen ts
made of 100 Type 2 curves the lowest ratio
for the dry curves (reRectance at 760 nm
divided by reRectance at 500 nm) was found
to be 1.2 the highest was 2.8, and the average
was 2.0. I n the occurrence of the 33 Type 3
curves, the lowest ratio was found to be 2.9.
the highest, 7.3, and the a\'erage 4.6. Note
that the lowcst ratio for the Type 3 cun'es is
higher than the highest ratio for the Type 2
curves.
For the wet curves, the lowest ratio for
Type 2 curves was found to be 1.2, the high-
SOIL
5 miles N. of Charlottesville, Va.
Monument Valley, Utah
DR'
DR'
~
40
'"~
30
Qj
WET
20
10
0l====:=L_.l-~_-L_...L._L-~
300
400
SOD
600
700
800
900
1000
Wavelength (nm)
FIG. 4. Type 3 curves for a sample of red
quartz and calcite sand.
300
400
500
600
700
800
900
1000
Wavelength (nm)
FIG. 5. Type 3 curves for a sample of laterite-
type soil.
958
PHOTOGRAMMETRIC ENGINEERING
est was 3.7, and the average 2.4. The lowest
ratio [or the Type 3 curves was 3.8, the
highest 8.2, and the average, 5.3. Again, the
lowest ratio [or the Type 3 curves is higher
than the highest ratio for the Type 2 curves.
32
28
24
20
[6 -
.!
;;
'"
12 -
L
~
/
v,
v,
-4
-8.=----;=:---------::=:--_L-_L-_.L...._.L...._.L....---.J
300
400
SOD
600
700
800
900
1000
Wavelength (nm)
FIG.
e
t:
~
6. Plot of the mean curve and four characteristic vectors.
SOIL
50
5 mile, N. of Charlottesville. Va.
740 nm
.!
40
640 nm
'"
)0
I
~
At an early stage in this study it became
eviden t that the measuremen t of the spectral
reflectance of soils by means o[ aerial photography would be greatly simplified if the
reflectance data at all 35 wavelengths (data
were digitized at 20-nm increments) could
be predicted from measurements made at
only a few wavelengths. The application of
characteristic vector analysis to photographic
and optical response data as described by
Simonds,[ of the Kodak Research Laboratories, was used in this study. For this analysis
200 spectral reflectance curves from 100 samples were selected, and from these, the mean
spectral reflectance data and [our vectors
were extracted. These curves are shown in
Figure 6. A measure o[ the variability of the
reflectance data is given in the trace, i.e., the
sum of the diagonal elements of the covariance matrix computed from the original
data. Characteristic vector analysis of these
data has shown that, in linear combination
with t.he mean curve, one vector can account
for 93.67 percent. of the t.race. T\\'o vectors
can explain 98.72 percent; three vectors,
99.60 percent.; and four vectors, 99.92 percent
of the trace. The value of the empirical eigenvector analysis lies in its finding that the
~
60
;;
ABRIDGED TECH [QUE OF
SPECTROPHOTOMETRY
ec
70
0
300
860 nm
I
DRV
WET
540 nm
20
10
t
440 nm
I
400
500
I
600
700
800
900
1000
Wavelength (nm)
FIG. 7. Location of the five selected wavelengths
with respect to the curve shape of one of the soil
samples.
dimensionality of the data variability is small;
it seems evident that reflectance data at all
wavelengths can be accurately predict.ed [rom
measurements made at only a few selected
wavelengths.
From an examination o[ the eigenvectors
from the covariance matrix, several possible
wavelengths were selected. Linear stepwise
regression analysis was used to find the best
linear combination of a specified number o[
predictor variables. These are best in the sense
that the particular combination gives a minimum standard error of estimate of data reconsti tu tion. This analysis revealed that reflectance measurements made at only 5 wavelengths should be required to predict with
su fficien t accuracy the reflectance at the
other 30 wavelengths. Ideally the selected
wavelengths should be 400, 540, 640, 740,
and 920 nm.
The energy reflected by the sample at these
wa velengths can be measured ei ther photoelectrically or photographically. If one uses
the latter technique, the soil, along with a
target of known spectral reflectance, would be
photographed through five interference filters whose peak transmittances correspond
with the five wavelengths. Knowledge of the
spectral sensitivity characteristics o[ the film
allows one to determine the energy reflected
at the five wavelengths by the standard
methods o[ photographic photometry. From
these energy measurements, the percent reflectance of the sample at these wavelengths
can be obtained. Unfortunately, the aerographic film most suited [or this workKODAK Infrared AEROGRAPHIC Film 2424
(ESTAR Base)-has, for all practical purposes,
no sensitivity at 920 nm. Inspection of the
film's spectral sensitivity curve indicates
959
THE SPECTRAL REFLECTANCE OF AMERICAN SOILS
TAI3LE
II'avelength
(nm)
1. WAVELENGTI'I-DEPENDE T ADDITIVE TERM A D REGRESSION COEFFICIENTS
Regression Coefficients
I Vavelength-lJependent
A dditive Term
ao
at
a2
a3
a.
a.
320
340
360
380
0.6670
0.5448
0.3545
0.2008
1.2289
1.3231
1.3876
1.3910
-0.7553
-0.7552
-0.7145
-0.6090
0.5264
0.4587
0.3579
0.2414
-0.3616
-0.2967
-0.2241
-0.1423
0.0589
0.0371
0.0293
0.0183
400
420
440
460
480
0.0823
0.0232
0.0000
-0.0260
-0.1005
1.3363
1.2159
1.0000
0.8972
0.8446
-0.4358
-0.2490
0.0000
0.1355
0.2020
0.1132
0.0236
0.0000
-0.0004
0.0145
-0.0492
-0.0002
0.0000
-0.0067
-0.0562
-0.0005
-0.0009
0.0000
0.0026
0.0450
500
520
540
560
580
-0.0870
-0.0738
0.0000
0.0062
0.0522
0.6537
0.3584
0.0000
-0.2325
-0.2073
0.3990
0.6879
1.0000
0.9766
0.5690
0.0049
-0.0072
0.0000
0.3661
0.9328
-0.0623
-0.0433
0.0000
-0.1742
-0.3901
0.0532
0.0385
0.0000
0.0300
0.0666
600
620
640
660
680
0.1068
0.0329
0.0000
-0.0793
-0.1054
-0.0936
-0.0331
0.0000
-0.0073
0.0073
0.2150
0.0669
0.0000
-0.0017
-0.0235
1.1539
1.1094
1.0000
0.8295
0.6405
-0.3199
-0.1412
0.0000
0.1558
0.3405
0.0248
-0.0085
0.0000
0.0221
0.0355
700
720
7-l0
760
780
-0.0788
-0.0283
0.0000
0.0020
O.03-l-l
0.0115
O.OO3-l
0.0000
-0.022-l
-().0-l0-l
-0.0243
-0.0131
0.0000
0.0182
0.037-l
0.4242
0.19-l1
0.0000
-0.0809
-0.1102
0.5852
0.8314
1.0000
0.9911
0.8630
0.0098
-0.0155
0.0000
0.0936
0.2-l50
800
820
8-l0
860
880
0.0579
-0.0461
-0.0567
0.0000
0.0467
-0.0520
-0.054U
-0.0372
0.0000
0.0455
0.0465
0.0498
0.0326
0.0000
-0.0402
-0.0657
-0.0352
-0.0066
0.0000
-0.0173
0.6233
0.3953
0.1759
0.0000
-0.1138
0.-l379
0.6355
0.8296
1.0000
1 .1330
900
920
940
960
980
0.0890
0.1465
0.1709
0.2361
0.3094
0.0981
0.1638
0.2329
0.2878
0.3437
-0.0817
-0.1472
-0.2117
-0.2637
-0.3171
-0.0884
-0.1558
-0.2637
-0.3774
-0.4999
-0.1375
-0.1435
-0.0817
-0.0095
0.1017
1. 2247
1.3069
1.3587
1.4034
1.4195
1,000
0.4297
0.3911
-0.3634
-0.6271
0.2334
1.4179
that the measurement in the infrared region
should be made at a wavelength no longer
than 860 nm. I n addition, since atmospheric
haze is generally very high in the ultraviolet
and far blue region of the spectrum, the measurement at 400 nm should be shifted to 440
nm. A shift of these two wavelengths will, in
some instances, lead to less accurate predictions in the regions of 320 to 380 nm and from
about 920 to 1,000 nm. In summary then, the
final choice of five wavelengths to be used in
this abridged technique of spectrophotometry
was 440.540,640,740, and 860 nm. The Joca-
tion of these selected wavelengths. with respect to curve shape of one of the soil samples,
is shown in Figu re 7.
From the 200 sets of measurements made
at the five selected wavelengths, regression
coefficients and wavelength-dependent additive terms were obtained, by a standard leastsquares regression analysis, for use in predicting the reflectance data at each of the other 30
wavelengths. This empirical deviation indicates that, to determine the full spectral
reflectance curve for a soil sample, one need
only measure the reflectances at 440, 540,
960
PHOTOGRAMMETRIC ENGINEERING
640, 740, and 860 nm and insert them into
linear regression equations of the form:
R x = ao.x
+ al,xR••o + a,.x!?.o + a3.xR MO
740
The wavelength-dependent additive term,
ao,X, and the regression coefficien ts, al,X, a2,X,
aa.x, etc., for each of the 30 prediction equations, plus the five equations corresponding
to the wavelength at which the reflectance
measurements were made, are listed in Table
1. Applying these regressions to the original
data yields low errors in predicting the reflectance data in this abridged manner. The
2.
STANDARD ERROR OF ESTIMATE
1Vavelength (nm)
.
~
+ a•. xR + a•. xRB,o.
TABLE
l
~
..
1.07
.89
.68
.48
400
420
440
460
480
.32
.2-1
.00
.21
.3-1
50
40
'"
30
20
.30
.2-1
.00
.30
.41
600
620
640
660
680
.33
.21
.00
.12
.16
700
720
740
760
780
.17
.11
.00
.15
.19
800
820
840
860
880
.24
.20
.16
.00
.18
900
920
940
960
980
.31
.47
.66
.82
.99
1,000
1.17
WET
10
O~~~~----l...-.---l....-~
300
400
500
600
700
BOO
900
1000
Wavelength (nm)
FIG. 8. Type 1 curves for a sample of chernozemtype soil. Measured values are shown by lines;
predicted values by a series of circles.
70
SilT
60
15 miles S.E. of Hoi Springs, Ark.
ec
50
~
40
~
DRY
~
'"~
WET
o
300
500
520
540
560
580
SOil
20 miles E. of lincoln. Neb.
~
Standard Error of Estimate (% Reflectance)
320
340
360
380
60
L
400
I
SOO
I
!
600
700
800
900
1000
Wavelength (nm)
FIG. 9. Type 2 curves for a sample of pedal fertype silt. Measured values are shown by lines;
predicted values by a series of circles.
70
60
ec
~
~
SAND
Monument Valley. Utah
50
40
Q;
'"
30
20
10
-l
600
1
I
700
BOO
~--"--
900
1000
Wavelength (nm)
FIG. 10. Type 3 curves for a sample of red quartz
and calcite sand. Measured values are shown by
lines; predicted values by a series of circles.
standard error of estimate, expressed in terms
of per cent reflectance, is tabulated for each
wavelength in Table 2. It should be pointed
out, however, that this abridged technique is
applicable only for materials whose spectral
961
THE SPECTRAL REFLECTANCE OF AMERICAN SOILS
Let us now see how \\'cl1 this abridged technique of spectrophotomctry has been able to
predict thc curvcs gi"cn in Figure 2 to 5.
These same cunTS are shO\\'n again in Figure
8 to 11. The actual measurements are shown
by the solid linc; the predicted values are
representcd hy the circlcs. It wil1 be noted
that except for thc mcasuremcnts from about
940 to 1,000 in Figure 8,10, and 11, the prcdicted values arc extrcmely accurate.
70
60
ec
50
~
';
40
'"~
30
..
SOil
5 miles N. of Charlottesville, Va,
20
10
o~~~C
300
400
500
I
I
L
800
900
1000
SPECTRAL REFLECTANCE CURVES
Wavelength (nm)
FIG. It. Type 3 curves for a sample of lateritetype soil. Measured values are shown by lines;
predicted values by a series of circles.
reflectance distributions lie within thc domain
of variability spanned by thc characteristic
vectors derived in this experiment.
TABLE
Fig. No.
3.
In the scction which fol1o\\'s, figures show
spectral rcflectances for "arious soil samplcs.
The soil sam pies, in cach of the "arious curve
type groups, arc presen ted in order of increasing reflectance of thcir dry curve at 700
nm, which is about the midpoint of the spectral range covered in this study.
I t has been found that any lack of a match
EVALUATION OF THE MATCH BETWEEN PREDICTED AND MEASURED
SPECTRAL REFLECTANCE DATA
320-380-nm Region
Dry Cur-de
Wet Curve
400-900-nm Region
Dry and Wet Curves
920-J,OOO-nm Region
Wet Curve
Dry Curve
Excellent
Excellent
Very good
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Fair
Excellent
Very good
Good
Fair
Excellent
Excellent
Good
Fair
Good
Excellent
Very good
Good
Very good
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Very good
Good
Good
Good
Excellent
Good
Fair
Excellent
Very good
Excellent
19
20
21
22
lexcellent
I'oor
Fair
Poor
Poor
Very good
Poor
Fair
Poor
Poor
Excellent
Excellent
Excellent
Excellent
Excellent
Very good
Excellent
Good
Very good
Fair
Excellent
Excellent
Excellent
Excellent
Fair
23
24
25
26
27
Very good
Excellent
Excellent
Poor
Very good
Excellent
Excellent
Excellent
Poor
Excellent
Excellent
Excellent
Excellent
Very good
Excellent
Excellent
Excellent
Excellent
Very good
Very good
Fair
Excellent
Excellent
Fair
Excellent
28
29
30
31
32
Poor
Poor
Excellent
Excellent
Good
Poor
Excellent
Excellent
Excellent
Good
Excellent
Excellent
Excellent
Excellent
Excellent
Very poor
Very good
Good
Poor
Excellent
Very poor
Very poor
Good
Fair
Very good
33
34
35
36
37
Excellent
Excellent
Excellent
Very good
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Excellent
Fair
Fair
Fair
Excellent
Excellent
Good
Fair
Excellent
Excellent
Fair
8
9
10
11
12
13
I~
15
16
17
18
962
PHOTOGRAMMETRIC ENGINEERING
SOil
.. miles N. of Neodesha, Ken.
60
ec
~
.
.
2
.
u
c
SO
- -r
60
SOil
1 mile E. of lindsborg. Kon.
SO
u
40
~
Gi
'"~
-,
70
70
'"~
30 -
40
DRY
30
20
20
DRY
WET
10
10
WET
0
300
400
500
600
700
800
900
400
1000
500
600
700
800
900
loaD
Wavelength (nm)
Wavelength (nm)
FIG. 13. Type 1 curves for a sample of chernozem-type soil.
FIG. 12. Type 1 curves for a sample of chernozem-type soil. Measured values are shown by
lines; predicted values by a series of circles.
70
70~ -'------'1--~--,-'- -,---,-----,
60 .
ec
~
~
ec
SOil
.. mIles SE of Avent,
Olclo.
~
.
50-
SO,
40
Gi
'"~
40
Gi
'"
SAND
Public Beach, Newport, R.I.
60
30
30
20
WET
10
400
400
SOD
600
700
800
900
14. Type 1 curves for a sample of chernozem-type soil
70~
.
40
~
~ ::r
Snoke River at Copper MI. in Hell's Canyon
SO -
Gi
'"
~
- r - -,
30
0
: ... :
... : ••
:
•
800
900
1000
' :
J
aL_.
I/
:1 0
' : WET _
I
SANO
I
SAND
ec
700
FIG. 15. Type 2 curves for a sand sample contai ni ng quartz, rock fragments, and shell fragments. The unusual feature of these curves is that
they are nat over such a long wavelength span: 600
to 1,000 nm.
70
60
600
Wavelength (nm)
Wavelength (nm)
FIG.
500
1000
Pacific Ocean Beach, Sonia Monica, Calif.
/---~~
~
...
w"
30Lo--L----.JL--60.L0--7-l0-0--8,--OOL---:9~00::---:-,0~0::0:--03 ooLI--..,..,---,-:-,----,--L-!,-----::ilc:--~I'----,----,-----,.....,-Wavelength (nm)
400
SOO
600
700
800
900
1000
Wavelength (nm)
FIG.
16. Type 2 curves for a sample of quartz and
rock-fragment sand.
FIG.
17. Type 2 curves for a sample of qllartz and
rock-fragment sand.
963
THE SPECTRAL REFLECTANCE OF AMERICAN SOILS
10
I
SAND
.
~
60 -
Cousins Island in
CO$CO
Boy, Moine
~
SO
2u
..
~
'"
.
40
/p
30 -
20
/ /
,,- ,-
.--
t
,,<> ............ ""..,. .. "
DRY -
/
~~-~4WET
10~
::I~':O
MIami Beoch, Flo .
SO
DRY
~ :~lc~:~:~--'~
. . .
oloooor
~~00;-~4:-!0:;;0-~50t,;0:----;6-!:-00::----:7:-!OC::O----'-80Lo'-----;9J.0"-0--,-.J00LO----.J
300
400
1
I
I
!
500
600
700
800
FIG. 18. Type 2 curves for a sample of quartz and
rock-fragment sand.
~
..
~
'"
~
carbonate sand.
70
SOIL
6.0
1 mile S. of Sagamore Hills, Ohio
ec
ec
50
~_~OO.Y
40
50
2
~
40
'"
30
G;
30~
~
20
20
10
10
ol~
300
I
I
-,-
/._
I
10 miles S. of lyman, Neb.
...J:l....lLg
.....
DRY
WET
o L_..-L_-L_---I_---".L-_....L_--"-,---,-,JL-_
400
500
600
700
800
900
1000
300
400
sample.
cSO~
q. ......
40 _
••• • A
700
800
900
1000
FIG. 21. Type 2 curves for a pedocal-type soil
sample. Note t he small difference III reflectance
between the dry and wet curves.
rl
CLAY
J miles E. of Paris, Mo.
e
600
--"T,---r--,---
,--,,--~,-- -
60 -
500
Wavelength (nm)
FIG. 20. Type 2 curvps for a pedocal-type soil
~
I
SOil
Wavelength (nm)
70
1000
FIG. 19. Type 2 curves for a sample of quartz and
, - - - ' - ,- - . - - - - I
60
900
WET
Wavelength (nm)
Wavelength (nm)
70
--,
r
•
•
:~~
•• ,
e ..
~
SAND
20 miles N. of Coos Bay, Oregon
DRY
._~~. WET
: :~.y/::::::- .../
oL
300
L
----~---"---L--.,.L-----'-----'
400
500
600
700
800
900
1000
Wavelength (nm)
FIG. 22. Type 2 curves for a clay sample. Note
that the difference in reflectance between the dry
and wet curves in this figure is fairly small.
o
~
300
400
.L-_--'-_ _...L_ _L...~~_ _-'-_---'
500
600
700
800
900
1000
Wavelength (nm)
1'10.
23. Type 2 curves for a sample of quartz
sand.
%4
l'HOTOGRAMMETRIC ENGINEERING
r
C
.::r
~
70,---.,---,---.----.---,---,----r---,
SAND
40
~30y'
- -.
~
20
60
~DRY
3 mil., E. of Mountain View, Mo.
DRY
ec SO
~
50
2
~
SOil
Woikiki Beoch. Honolulu, Hawaii
e _ _ 'lo-"~
WET
~..
40
Qi
'"
WET
30
20
10
OL-_-L_ _L-_-L_ _L--_-L_ _L-_--L_---l
300
400
500
600
700
800
900
1000
500
Wavelength (nm)
2~.
FIG.
600
700
800
900
1000
Wavelength (nm)
Type 2 curves for a carbonate sand
sample.
,FIG.
25. Type 2 curves for a pedalfer-type soil
sample.
70
.
~
SOil
60
12 miles N. of Dolton, Go.
DRY
50
70
..
40
60
'"
30
~
~
ec
3 miles W. of Phillipsburg, Mo.
DRY
,
...
0
50
~
.
20
~
'"~
10~
a
r----,--~SO-I-l-'~---,-~-:~-~:-~--,---~
000
300
0
400
500
600
700
800
900
1000
Wavelength (nm)
FIG. 26. Type 2 curves for a sample of a pedalfertype soil. The curves shown in this figure are
difficult to classify. Portions of the curves more
nearly resemble the Type 3 shape but they were
placed in the Type 2 classification because of the
rapid rise in renectance in the region from 320 to
SOO nm, which is particularly characteristic of the
Type 2 curves. This is one of only two such Type 2
curves thus far obtained in which the reflectance
decreases by any appreciable amount in the region
beyond 760 nm.
400
500
600
700
800
900
1000
Wavelength (nm)
FIG.
27. Type 2 curves for a pedocal-type soil
sample of very high reflectance.
80,.---,-------,----,-----,---.,-----,-------,-DRY
70
~ooo
~ WET
SAND
Notional Monument, N. Mex.
10
OL-_---'-_ _-'---_---'-_ _-'---_--'-_ _-'--_--'-_---'
300
40Q
500
600
700
Wav~lenlJth
800
(nm)
900
1000
FIG. 28. Type 2 curves for a sample of gypsum
sand. Note that beyond 920 nm there is a sharp
decrease in the reflectance for both the wet and
dry curves, the abridged technique was unable to
predict correctly values beyond 920 nm.
965
THE SPECTRAL REFLECTANCE OF AMERICAN SOILS
8o,1-.------.---,--.---,----r-=-""'0.~
70
--,-
.,
--,
DRY
60
e ::~
c
~ 50
c
WET
-;
0<
40
0<
40
0
-;
~
~
~
50
2u
CU
SOil
5 mil.s S.E. of Rowe, N.M.
~
SAND
20
Fl. Walton Beoch. flo,
30/
30
~DRY
WET
10
20
O~~~...-L...-...L..-.-.i---l.-~
10
300
400
500
~0;;0;--:4::0::0--5~0::0-:-:6~00=--=7,.t0-,0--80LO--9.L00--IO.lO-0-....J
Wavelength (nm)
FIG. 29. Type 2 ('urves for a sample of quartz
,;and, collected at Ft. Walton Beach, Florid,!. This
sample was found to have the highest spectral
reAectance of any thus far measured, reachillg a
value of 79 percent at 1,000 11m.
70
50
~
40
u
~
-;
0<
~
60
SOIL
"c
~
_~ ••• ~DRY
~
o~
30
20
~"
0<
~
~ WET
400
500
600
700
800
40
30
0~~~:L---..L_---l.-_l-.----L.._..l-~
900
300
1000
50D
600
700
800
900
1000
FIG. 32. Type 3 curves for a sample of a lateritetype soil.
70
SOil
60
60
Woodlawn, Colo.
"c
50
~
40 -
~
400
Wavelength (nm)
70
~
c
/ D o·
RY
,~Q." •• ~
30
~WEI
10
FIG. 31. Type 3 curves for a sample of a pedal fertype soil.
0<
1000
50
Wavelength (nm)
-;"
900
SOIL
2 miles N.E. of lu:ington, N. Cor.
20
~o.
10
300
BOO
70
Gorden of the God" Colo.
"c
700
FIG. 30. Type 3 curves for a sample of a very low
reAectance, pedalfer-type soil. Note the very
sl11all difference in reAectance in the wet and dry
curves.
I
60
600
Wavelength (nm)
-
~
-;"
50
0
~
30
00
20 -
20
10
10
SOD
600
700
800
900
~ D R00Y
40
0<
~WET
o·
400
SOIL
12 miles W. of Elk City, Oklo,
1000
Wavelength (nm)
FIG. 33. Type 3 curves)or a sample of a pedal fertype soil.
WET
o~~~----,,-l-:---,l-,----~.l:-:300
400
500
600
700
800
900
1000
Wavelength (nm)
FIG. 3-l. Type 3 curves for a sample of a pedocaltype soil sample.
966
PI-IOTOGRAMMETRIC ENGINEERING
70 , - - - , - - - - . - - - - - - - . - - - , - - - , - - - - . - - - - - , - - - - - ,
70
1
.
~
~..
SOil
60
60
6 miles N. of Del Rio, Texas
~
c:
SO
o
40
Q 0
0
DRY
Qj
'"
30
WEI
400
500
600
700
800
900
1000
'"
30
70
'~--'---""'---'---r
60
SOil
-4 miles N. of Griffin, Ga.
O'-:----'----....L_......J_ _L..._...L_....L_......J_----J
300
400
500
600
700
800
900
1000
Wavelength (nm)
,-
FIG.37. Type 3 curves for a sample of quartz sand
with a hematite stain.
with which the abridged technique was able
to predict spectral reflectance val ues.
CONCLUSIONS
50
~ORY
40
Qj
~
'"
~
10
FIG. '>5. Type.> curves for a sample of a pedal fer··
type soil.
.
40
10
Wavelength (nm)
~
JQj
20
o l..!!:!E~~=----.l..._---L_J-_..L--_.L-~
~
c:
DRY
50
2
20
300
SAND
Bole Tower, Fla .
30
WEI
o
...J
300
400
-I................L.
500
600
L
700
800
_
..l
900
1000
Wavelength (nm)
FIG. 36. Type 3 curves for a sample of a pedalfertype soil.
between predicted and measured values of
spectral reflectance usually occurs in the regions from 320 to 380 nm and 920 to 1,000
nm. I n Table 3 is given an eval uation of how
well the predicted values match the measured
values in these two regions as well as in the
region from 400 to 900 nm for the 30 samples
discussed in this paper.
Figures 12 to 14 reproduce Type 1 curves,
Figure 15 to 29 show Type 2 curves, and the
remaining figures (30 to 37) show Type 3
curves for representative soil samples examined.
Of the 30 sets of curves presented in this
paper, 18 were from the 100 sets used in the
characteristic vector analysis. They are shown
in Figure 8-12, 15-18, 21-24, 26, 27, 32, 35,
and 37. The curves not included in the
analysis are shown in Figure 13, 14, 19,20,25,
28-31,33,34, and 36. It is important to note
that no significant difference exists between
the two groups with respect to the accuracy
From a study of the degree to which the
predicted spectral reflectance data matched
the measured values, it seems evident that the
spectral reflectance of a wide variety of soils
(a basic characteristic in their iden tification)
can be predicted with sufficient accuracy from
measurements made at only five waYelengths.
ACKNOWLEDGMENT
The author gratefully acknowledges the
assistance of three of his Kodak Research
Lab. colleagues: Leroy E. DeMarsh, who supplied the compu tel' program for the characteristic vector analysis, David J. Paine for his
assistance in the stepwise regression analysis
and in the determination of the regression coefficien ts, and John L. Simonds for his preparation of the computer program used in predicting the spectral reflectance values by
means of the abridged technique of spectrophotometry and for his many helpful suggestions in the preparation of this paper. The
author also wishes to thank Dr. Robert G.
Su tton, Professor of Geology at the U niversity of Rochester, for iden tifying the soil samples, and Donald J. Belcher, Professor of
Civil Engineering and Director for Aerial
Photographic Studies at Cornell University,
for general counsel on the characteristics and
geographic location of various types of soils.
REFERE 'CE
1. Simonds, ]. L., "Application of Characteristic
Vector Analysis to Photographic and Optical
Response Data," J. Upt. Soc. Am. 53: 968-974,
August 1963.