REMOTE SENS. ENVIRON.

47:345-361 (1994)

Emissivity of Terrestrial Materials in the

3-5/ m Atmospheric Window
John W. Salisbury* and Dana M. D'Aria*

A c c u r a t e satellite remote sensing of surface temperature surface properties to account for surface upwelling radia-
has become more and more important in the context of tion. Because of the multiple uses to which our data can
calculating the radiation balance input to climate models be put, digital records of all spectra used in the figures
for assessment of global change. Surface temperatures are available from 2.01 12m to 14 ltm upon request.
can be determined from radiance measurements in both
the 3-5 l~m and 8-14 ltm atmospheric windows, given
some estimate of surface emissivity. The 3-5 tJm region INTRODUCTION
offers a potential advantage in the accuracy of tempera-
ture determination, because temperature estimates are In a previous article (Salisbury and D'Aria, 1992a), we
more than two times less sensitive to emissivity errors discussed the 8-14/~m emissivity of earth surface mate-
in this region. However, very little is known about the rials in the context of the problem of separating tempera-
emissivities of land surface materials in the 3-5 l~m ture and emissivity, which was the theme of the special
region, If emissivities were more variable in the 3-5 I~m issue of Remote Sensing of Environment in which it was
region than in the 8-14 I~m region, this might cancel the published. Indeed, accurate satellite remote sensing of
advantage of lower sensitivity to emissivity error in the surface temperature has become more and more im-
shorter wavelength region. This article documents the portant in the context of calculating radiation balance
directional hemispherical spectral reflectance, from which input to climate models for assessment of global change.
emissivity can be calculated, of a wide variety of earth While remote sensing of sea surface temperature has
surface materials in the 3-5 l~m region, and compares been rather successful (McClain et al., 1985) due to the
their average reflectances in A VHRR Band 3 (3.55-3.93 relative uniformity of the target, the variability of land
12m) to that in Bands 4 (10.3-11.3 l~m) and 5 (11.5-12.5 surface materials and general ignorance of their emissivi-
I~m). It appears from these data that, based on natural ties has made land remote sensing more difficult (Price,
variations in emissivity of different surface materials, 1983). Simplifying assumptions are often used over a
AVHRR Band 5 is best used for temperature remote variable surface, such as that all polar ice is snow-
sensing of rocks, soils, and senescent vegetation, while covered (Key and Haefliger, 1992), or that land surfaces
Band 3 is best for green vegetation and water. Band 4 is simply emit as black bodies (Price, 1983). Clearly, more
preferred for thermal remote sensing of ice, especially detailed knowledge of surface emissivities is necessary
rough ice. Emissivity and reflectivity in the 3-5 ltm region for more accurate temperature measurements, and our
affect other significant remote sensing measurements of previous article began the task by documenting emissiv-
global change besides those of surface temperature alone. ity of surface materials in the 8-14/~m region.
The accuracy of atmospheric sounding of CO near 4.7 While the 8-14/tm atmospheric window is widely
I~m, for example, depends in part on knowledge of such used in remote sensing of temperature, because typical
earth surface temperatures produce peak radiance in
that wavelength range, the 3-5/2m window may also
be useful. Volcanic eruptions and fires exhibit peak
*Department of Earth and Planetary Sciences, Johns Hopkins radiance in the shorter wavelength region, but this
University, Baltimore region has a significant potential advantage for remote
Address correspondence to John Salisbury, Dept. of Earth and sensing of cooler temperatures as well. This advantage
Planetary Sciences, Johns Hopkins University, 34th and Charles
Streets, Baltimore, MD 21218. is revealed by a Planck's law calculation showing that
Received 6 November 1992; revised 3 April 1993. the temperature error in the shorter wavelength region
0034-4257 / 94 / $7.00
©Elsevier Science Inc., 1994
655 Avenue of the Americas, New York, NY 10010 345
346 Salisbury and D'Aria

due to an error in estimating emissivity is less by a factor of different types of surface materials can be obtained
of more than 2 (that is, at average Earth temperatures an from the senior author by sending him three high den-
emissivity error of 0.01 results in a temperature error sity 3.5 in. disks, and specifying MS-DOS or Macintosh
of approximately 0.25°C at 4/zm, compared to a temper- format for the digital data.
ature error or 0.6°C at 10/zm). Since surface tempera-
ture accuracy is so important in radiation balance calcu-
SPECTRAL BEHAVIOR OF ROCKS
lations, this reduced sensitivity to emissivity error
advantage of the 3-5/zm region might easily outweigh Igneous Rocks
the energy disadvantage. However, the range of temper- Spectra of suites of igneous rocks have been published
ature error in each wavelength region depends on the by several other authors (e.g., Lyon, 1964), but, with
range of emissivities found there. The emissivity range the exception of spectra of two igneous rocks by Hovis
for the 3-5 /~m region has been very poorly known and Callahan (1966), all others have skipped the 3-5
(Hovis and Callahan, 1966), and it is the purpose of this /zm region, except for Salisbury et al. (1988). Figure 1
article to correct this deficiency by documenting the shows representative spectra taken from Salisbury et al.
3-5 gm emissivities of a wide variety of earth surface (1988) of the full range of igneous rock types from felsic
materials. We tabulate the average emissivity of these to ultramafic. With the exception of quartz, the spectra
materials in Advanced Very High Resolution Radiome- of major igneous rock-forming minerals are featureless
ter (AVHRR) Band 3 (3.55-3.93/zm) and, for convenient in the 3-5 /zm region. They usually display a broad
comparison, in Bands 4 (10.3-11.3 /zm) and Band 5 reflectance maximum in the middle of the region be-
(11.5-12.5/zm) as well. cause of absorption bands at shorter wavelengths due
This tabulation of reflectances in AVHRR band re- to water in fluid inclusions, and at longer wavelengths
gions illustrates one application of these spectral data. due to overtone / combination tone bands of the funda-
However, emissivity and reflectivity in the 3-5/zm re- mental molecular vibration bands in the 8-20/zm region
gion affect many other significant remote sensing mea- [the origins of molecular vibration bands in minerals
surements of global change besides those of surface are discussed in detail in Farmer (1974), Farmer and
temperature alone. The accuracy of atmospheric sound- Palmieri (1975), and Estep-Barnes (1977) and summa-
ing of CO near 4.7/zm, for example, depends in part rized in Salisbury et al. (1992)]. By contrast, the basalt
on knowledge of such surface properties to account for spectrum (dash-dot curve in Fig. 1) shows a lack of
surface upwelling radiation. Thus, several instruments fluid inclusions in the minerals comprising the basalt
on board the Earth Observing System (EOS) will need by the rise in reflectance toward 3 /zm. The granite
emittance and reflectance data in the 3-5 gm and spectrum shows the presence of quartz by the very
8-14/zm atmospheric windows (e.g., Fujisade and Ono, weak quartz absorption bands near 4.5/zm and 4.7 gm,
1991; Salomonson et al., 1989; Drummond, 1991). Be-
cause of the multiple uses to which our data can be
put, digital records of all spectra used in the figures are
available from 2.01 /~m to 14 /zm upon request (see Figure 1. Directional hemispherical reflectance spectra of
below). rough surfaces of igneous rocks, including a basalt (dash-
dot curve), granite (dotted curve), syenite (solid curve), and
dunite (dashed curve).
EXPERIMENTAL TECHNIQUE
30 , , ' ' 1 ' ' ' ' 1 ' ' ' ' 1 '''°
The details of sample acquisition and measurement
technique are given in Salisbury and D'Aria (1992a).
Briefly, near-normal (10 °) directional hemispherical re- Igneous Rocks
flectance spectra of well-characterized samples were G)
measured on a Nicolet System 510 FTIR spectrometer o 20
with 4 wavenumber resolution throughout the spectral
range from 2.08/~m to 14 gin. These reflectance spectra
can be used to predict directional emissivity using Kirch- rr
hoffs law, as demonstrated with emittance spectra of
quartz sand (Salisbury and D'Aria, 1992a). For igneous
rocks, most spectra are of freshly broken rock surfaces.
For sedimentary and metamorphic rocks, very coarsely
particulate (250-1500/zm size range) material was used I I I I I I I i I I I I I I I I I I I

as a good analog of a fresh, rough surface.

Digital data are recorded for each spectrum. The 3 3.5 4 4.5
data used in the figures illustrating the spectral behavior Wavelength (l~m)
 Emissivity of Terrestrial Materials at 3-5 ~am 347

Table 1. Average Reflectances from 3 . 5 5 - 3 . 9 3 / a m (AVHRR Band 3), 1 0 . 3 - 1 1 . 3 / a m (AVHRR Band 4), and 11.5-
12.5/~m (AVHRR Band 5)

% Reflectance
Samples Filename Band 3 Band 4 Band 5
Rocks
Igneous Aplite.hl 16.74 8.37 4.41
Granite.hi 8.72 8.54 4.80
Granite.h2 12.54 8.88 4.62
Granite.h3 9.45 7.07 3.53
Granite.h5 9.16 7.04 3.56
Obsidian.h 1 4.83 11.31 5.48
Rhyolite.hi 16.42 9.58 4.82
Andesite.hl 4.24 10.41 5.22
Andesite.h2 11.87 6.68 3.46
Andesite.h4 6.95 9.59 4.18
Diorite.hi 12.81 9.09 4.49
Granodior.hl 10.63 10.60 5.47
Granodior.h2 4.71 9.54 4.71
Monzonite.hl 12.45 8.95 4.80
Qmonzonite.hl 11.67 7.00 3.30
Syenite.hi 7.60 10.06 5.08
Syenite.h2 14.26 8.80 4.18
Tonalite.hl 12.19 11.26 5.58
Anorthosite.hl 8.54 7.40 3.97
Basalt.hi 3.57 9.57 5.00
Basalt.h2b 7.05 10.44 6.81
Basalt.h5 11.11 13.36 7.15
Basalt.h7 11.40 9.66 4.74
Basalt.h9 9.24 11.08 5.42
Basalt.hi0 13.39 11.91 6.41
Diabase.hl 9.84 11.02 6.09
Diabase.h2 8.55 11.59 6.16
Gabbro.hl 7.41 12.87 5.64
Ijolite.hl 14.92 8.84 5.26
Lamprophyre.hl 5.39 10.84 6.06
Norite.hl 14.29 10.10 4.48
Norite.h2 4.99 12.36 6.33
Dunite.hl 7.24 14.60 5.23
Picrite.hl 11.31 28.81 12.24
Picrite.h2 16.13 14.79 6.65
Sedimentary Greywacke .h I 15.94 5.01 2.77
Limestone.hi 10.58 5.57 3.40
Limestone.h2 10.58 6.37 3.40
Limestone.h3 4.72 9.43 3.49
Sandstone.hi 16.81 3.60 2.20
Sandstone.h2 11.52 3.23 2.07
Sandstone.h3 12.17 4.02 2.54
Sandstone.h4 13.65 4.02 2.54
Sandstone.h5 36.18 4.21 2.69
Shale.hi 13.61 2.61 1.86
Shale.h2 6.03 3.55 2.50
Shale.h3 11.05 3.53 2.86
Shale.h4 5.28 2.99 2.31
Shale.h5 8.79 3.15 2.15
Shale.h6 5.21 3.34 2.68
Siltstone.hl 23.82 3.69 2.96
Siltstone.h2 28.93 3.25 2.22
Metamorphic Gneiss.hla 31.99 4.03 2.56
Gneiss.h3 26.12 5.07 2.77
Gneiss.h4 18.08 6.46 3.29
Hornfels.hla 19.50 6.08 3.41
Marble.h2 5.47 4.67 2.37
Marble.h3 9.77 5.09 4.62
Marble.h4 8.85 6.51 4.74
348 Salisbury and D'Aria

Table 1. (continued)
% Reflectance
Samples Filename Band 3 Band 4 Band 5
Marble.h5 7.70 4.77 2.92
Phyllite.hl 7.04 7.63 4.93
Quartzite.hi 22.04 2.81 1.97
Quartzite.h4a 28.30 4.74 2.98
Quartzite.h6 32.51 4.63 2.90
Schist.h3a 10.24 5.80 3.80
Schist.h6a 45.06 10.34 4.36
Schist.h7 22.04 8.59 4.15
Slate.hla 11.38 5.30 3.05
Slate.h2a 18.91 5.19 3.02
Slate.h3 4.11 4.18 2.73
Rock Coatings
Desert varnish Rhyolite.l~ 19.67 5.21 3.06
Rhyolite.vb 11.26 8.65 4.53
Basalt.tz 5.88 8.65 4.53
Basalt.vb 7.42 4.67 3.34
Ijolite.f~ 7.42 4.67 3.34
Ijolite.v b 12.66 7.05 4.83
Lichens Crustose.10 8.60 4.02 2.66
Crustose.65 10.39 4.11 3.35
Soils
Entisols 0135 17.12 3.55 2.18
0149 8.75 2.40 1.89
2230 14.31 2.70 1.95
Vertisols 0475 13.64 2.69 1.99
Inceptisols 0138 10.43 2.41 1.91
0215 10.36 2.84 1.95
0224 10.25 3.48 2.73
0226 16.71 3.13 2.34
0227 11.65 3.68 2.42
0209 25.34 3.61 3.05
2671 14.58 2.29 2.11
Aridisols 0139 27.82 3.20 2.76
0147 28.41 4.77 2.68
0148 33.14 3.28 3.48
0150 9.95 2.12 1.63
0151 28.40 3.01 2.76
0152 23.29 2.29 2.36
2659 22.89 2.65 2.85
1530 21.74 3.21 2.88
1536 22.95 2.85 2.61
3721 21.50 2.36 2.37
2695 25.48 2.82 3.05
Mollisols 0134 20.75 3.12 2.24
0140 24.53 2.93 2.35
0211 16.24 2.68 2.05
0212 11.62 2.57 2.09
0213 20.40 2.50 2.27
0216 19.20 2.22 1.86
0225 12.46 2.77 2.07
Spodosols 0127 10.99 3.04 2.86
Alfisols 0128 29.60 3.44 2.77
0129 20.13 3.57 2.34
0132 22.62 3.10 2.31
0133 26.02 3.02 1.98
0137 20.60 3.34 2.29
0214 19.46 2.62 2.28
0217 11.90 2.71 2.48
0219 22.89 3.55 2.42
0221 22.06 2.99 1.91
0222 27.23 3.46 2.66
 Emissivity of Terrestrial Materials at 3-5 gm 349

Table 1. (continued)
% Reflectance
Samples Filename Band 3 Band 4 Band 5
Ultisols 0136 22.24 3.86 2.42
0145 27.65 4.70 2.81
0146 32.52 4.89 3.16
0208 24.13 4.22 3.16
0210 27.39 4.03 3.12
0223 29.76 4.63 2.99
0220 21.08 4.00 2.67
Oxisols 4717 11.08 1.73 1.35
Vegetation
Lichens Foliose.1 6.36 3.13 2.02
Foliose.2 5.77 2.90 2.08
Fruticose. 1 4.93 2.89 2.18
Green foliage Beech (Fagus grandifolia) 4.91 4.72 3.99
Cherry (Prunus serotina) 3.17 3.55 2.89
Hickory (Carya glabra) 4.85 1.82 1.61
Laurel (Kalmia latifolia) 4.13 4.68 4.64
Redmaple (Acer rubrum) 3.64 4.70 3.82
Redoak (Quercus rubra) 3.38 5.29 5.35
Sugarmaple (Acer saccgaryn) 3.20 4.41 3.45
Yellowpoplar (Liriodendron tuliperfera) 3.51 3.23 2.17
Conifer (Pinus sp.) 2.32 2.04 2.03
Cedar (Cedrus deodara) 2.20 2.17 2.23
Whitepine (Pinus strobus) 2.38 2.21 1.83
Bigbluestem (Andropogon gerardii) 3.23 3.58 2.22
Indiangrass (Serghastrum nutans) 3.19 3.77 2.35
Switchgrass (Panicum virgatum) 4.64 3.35 2.75
Moss (unknown) 4.32 1.99 1.65
Senescent foliage Senbeech (Fagus grandifolia) 24.83 17.67 14.84
Senredoak.hl (Quercus rubra) 15.53 9.26 8.25
Senpine (Pinus strobus) 3.86 2.25 1.94
Senryegrass (Secale cereale) 15.12 8.94 8.61
Tree bark Beechbark (Fagus grandifolia) 11.75 4.16 4.59
Oakbark.1 (Quercus alba) 12.23 8.10 8.93
Oakbark.2 (Quercus rubra) 8.78 4.42 4.71
Ypoplarbark (Liriodendron tuliperfera) 12.57 6.29 5.84
Pinebark.1 (Pinus taeda) 9.04 5.13 5.23
Grapebark (Vitus sp.) 6.99 4.13 3.67
Decomposing soil litter Wood 8.85 3.48 3.23
Deciduous 7.63 3.99 3.67
Coniferous 6.42 2.40 1.84
Water and ice
Water Seawater 2.53 0.961 1.40
Distwater 2.51 0.93 1.45
Ice Seaice.smooth 3.85 2.31 2.72
Seaice. 100grit 7.45 1.44 2.94
Distice.smooth 3.04 2.36 6.38
Distince.100grit 5.35 1.35 2.83
Suspended sediment Qtzwater.7 2.53 1.16 1.58
Qtzwater.23 2.54 1.07 1.55
Qtzwater.64 2.54 0.973 1.54
Water coatings Foam 3.36 0.955 1.31
Oi115465 3.95 3.77 3.77
Oi135473 3.77 3.63 3.63
Oi134792 4.15 4.03 4.01
Oi139076 4.28 4.43 4.47
Oil42667 4.17 3.67 3.60
Qtzfloat 2.61 1.68 1.72
Soilfloat 2.67 1.71 1.81
a Fresh surface.
Varnish.
350 Salisbury and D'Aria

which are better displayed in spectra of quartz-rich increases the reflectance variations in Band 3, as the
sedimentary and metamorphic rocks discussed below. effects of compositional differences add to the effects
Table 1 shows the reflectances of different Earth of differences in surface roughness and the abundance
surface materials convolved with AVHRR Bands 3, 4, of scattering centers.
and 5, assuming square filters in these bands. Igneous The spectra of sandstones (Fig. 2a) are typically
rock spectra display their greatest variation in reflec- dominated by the reflectance of quartz, as illustrated
tance in AVHRR Bands 3 and 4, but for different reasons by the dotted curve. This shows a peak reflectance near
in the different bands. The variability of reflectance in 3.92/~m, with a falloff toward shorter wavelength due
Band 3, and the lack of dependence of this variability to absorption by water in fluid inclusions near 2.9/tm,
on rock type, are due to the relative transparency of all and toward longer wavelength due to the increasing
silicate rock-forming minerals in this region (Salisbury et strength of S i - O overtone/combination tone bands
al., 1992). Photons enter such semitransparent minerals near 4.29/~m, 4.48/~m, 4.69/.tm, and 4.95/~m.
easily, and the number backscattered depends on the Many quartz grains in sandstone are cemented by
roughness of the surface and the abundance of scattering carbonate, which displays strong absorption bands that
centers (grain boundaries, cracks, fluid and solid inclu- affect AVHRR Band 3. The ferruginous sandstone (solid
sions, etc.) below the surface. Because the surface curve in Fig. 2a) has such a cement, and the two
roughness and abundance of scattering centers is highly doublets near 3.36 / 3.49/~m and 3.85 / 3.99/~m are the
variable from rock to rock, the reflectance is also. resulting carbonate bands.
Band 4, on the other hand, lies in a wavelength In Figure 2b the graywacke spectrum (solid line)
region that commonly includes the fundamental molecu- displays weak quartz overtone/combination bands, but
lar vibration ("reststrahlen") bands. The wavelength of otherwise shows the typical broad rise in reflectance
these features varies systematically with mineralogy, centered near 4 /~m, as does the feldspatic siltstone
which has been known for quite a long time (Lyon, spectrum. The argillaceous siltstone (dashed curve) also
1964), and which was illustrated in Salisbury and D'Aria contains carbonate, as demonstrated by the presence of
(1992a). The intensity of these reststrahlen bands does the stronger carbonate doublet near 3.85/3.99 /tm.
vary with surface roughness, but most of the samples These materials have a generally higher reflectance than
measured have comparably rough, freshly broken sur- sandstones because of their smaller grain size (4-63
faces. Thus, the reflectance variations in Band 4 (Table /~m, as compared to 63 /Jm-2 mm), which increases
1) are primarily a function of composition. backscatter.
Band 5 (Table 1) shows relatively little variation in Figure 2c illustrates spectra of typical shales. The
reflectance for igneous rocks compared to Bands 3 and calcareous shale spectrum (dotted curve) displays very
4, because it lies to longer wavelength than the most strong carbonate bands, including features near 4.3/tm
intense reststrahlen bands (with the exception of one and 4.7/~m that overlap with quartz absorption bands.
of the picrites). These features are so weak that they are not usually
seen unless carbonate is very abundant or the bands
Sedimentary Rocks are enhanced by the scattering induced by very fine
Very few spectra of sedimentary rocks covering the particle size. The carbonaceous shale spectrum (dashed
3-5/~m region are available in the open literature. In curve) has a low reflectance due to the presence of
fact, we could find only one such spectrum (Hovis and opaque organic matter, but weak carbonate bands can
Callahan, 1966), as both Hunt and Salisbury (1975) still be discerned. The illitic shale contains a little car-
and Lang et al. (1990) skipped this wavelength region. bonate, but the spectrum (solid curve) has the largely
Consequently, we show a relatively large number of featureless broad rise associated with all silicates lacking
representative spectra in Figure 2, and comment briefly spectral features in this region.
on their individual spectral features. Generally speaking, Not surprisingly, the limestone spectra in Figure
it is apparent from the relatively low reflectance and 2d display very strong carbonate bands. As in the case
lack of variability in Band 4 of Table 1 that sedimentary of the carbonaceous shale, however, the carbonaceous
rocks do not display strong reststrahlen bands in Band limestone spectrum (dotted curve) has a low reflectance
4, in contrast with igneous rocks. This is because silicate due to the high opacity of the disseminated organic
minerals having reststrahlen bands in this region, such matter. Major spectral features for the dolomitic lime-
as feldspars and mafic minerals, are relatively scarce in stone (dashed curve) are shifted to shorter wavelength
sedimentary rocks. At the same time, the carbonate than those exhibited by the calcareous limestone (solid
minerals common in sedimentary rocks typically display curve), due to the substitution of Mg for Ca, and addi-
their bending vibration bands between Bands 4 and 5. tional bands occur near 4.29 /~m and 4.43 /~m. Note
On the other hand, carbonate minerals usually display that the weak spectral features in the carbonaceous
strong absorption bands in the 3-5 /~m region. This limestone, partially masked by the opaque carbonaceous
 Emissivity of Terrestrial Materials at 3-5 lzm 3 5 1

30 ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' 40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' '

Graywacke and
i i
siltstones
Sandstones 30
@ f \ ~ / \
0
¢- 20 0
c- t "_~Z_ a
t~

',,;..,...
L) I
/ ,
_~ 20
ft. rr
10
10

I I I I I t ' ' ' I , , ~ , I I I I I . . . . I , , , , I , , , , I , , , ,

0
3 3.5 4 4.5 5 3 3.5 4 4.5 5
Wavelength (Ixm) Wavelength (I.tm)
(a) (5)

30 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' 30 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' '

I1)
Limestones
o 20
t- Shales o 20
t"

t~

r I se ~# ,i

rr i
i
i
t
# t
l _

~ 10 ~ 10
il, ,,I
.... ".', ' P-" '-:'"'-,'_7." ....

0 0 , , i i I . , . I l i i a I I , i i i

3 3.5 4 4.5 5 3 3.5 4 4.5 5

Wavelength (!am) Wavelength (I.tm)
(e) (d)
Figure 2. Directional hemispherical reflectance spectra of coarsely particulate (250-1500/zm size range) samples of different
sedimentary rocks: a) sandstones, including ferruginous (solid curve), glauconitic (dashed curve), and arkosic (dotted curve)
samples; b) graywacke (solid curve) and argillaceous (dashed curve) and feldspathic (dotted curve) siltstones; c) shales, includ-
ing illitic (solid curve), carbonaceous (dashed curve), and calcareous (dotted curve); d) limestones, including fossiliferous
(solid curve), dolomitic (dashed curve), and carbonaceous (dotted curve) samples.

material, clearly indicate by their wavelength positions briefly discuss the spectral features of individual sam-
that the carbonate present is dolomitic. ples. Generally speaking, metamorphic rock spectra dis-
play slightly more variability in AVHRR Bands 4 and
5 than do sedimentary rocks because of the greater
Metamorphic Rocks variability in their compositions. However, it is Band 3
As in the case of sedimentary rocks, we were unable to that shows the greatest variability, exceeding that of
find more than a single spectrum of a metamorphic rock both sedimentary and igneous rocks, because of the
covering the 3-5/tm region (Hovis and Callahan, 1966), more strongly expressed absorption bands of the recrys-
because Hunt and Salisbury (1976) skipped this wave- tallized minerals in metamorphic rocks.
length region. Thus, we also present a relatively large Despite the variation in composition of the gneisses
number of metamorphic rock spectra in Figure 3 and for which spectra are shown in Figure 3a, the spectra
352 Salisbury and D'Aria

40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' 40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' '

• . -# , - , ' , . - • Gneisses Marbles

o
30 30
o ° -I
E
O • / \ "~ -I 0
_~ 2 0 ," / ~ \ ', • 20 ,'-I i~ -i~ I"-.
0 )

:, ~, J v'-zN', ,
,>/- ,,q. --,,
10 /- \,',., 1 0 ~t~]~ ~l ~

0 , , , , I , , , , I , , , ' l l l I u
O ~ o'--'-

3.5 4 4.5 5 3 3.5 4 4.5

Wavelength (tam) Wavelength (I.tm)
(a) (b)

40 ii ii illU i,~ I l n l l l U l l l U Ilnllll III IIIlU u II 50 r II

i' ' I ~' 'I I "~. I' ~''~#..
' ' I . . . . I ' ' ' '
'~ ~ Quartzites 40 I ~ t ~ Schists
~-~
30 o r~ I
I I

I~l
I

~\
o¢- o
E
t~ N 30
o
_~ 2 0
o
n- 7 .., i t ,.,
20 • kl %
,, ,- ,, , , ,, ; , ;,,

10 o S •

10

iiii iiiiii iiiiiiiiiiiiiiiiiiiiiiiiii i I I , , , I I I I

0 , , , , , , , , , , ,

3 3.5 4 4.5 5 3 3.5 4 4.5 5

Wavelength (I.tm) Wavelength (tam)
(c) (d)

40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' '

@ 30 Slates
ot-
t~
20
o
Figure 3. Directional hemispherical reflectance spectra of
coarsely particulate (250-1500 pm size range) samples of
different metamorphic rocks: a) gneisses, including syenitic
10 (solid curve), felsitic (dashed curve), and chloritic (dotted
curve); b) marbles, including dolomitic (solid curve), calcitic
(dashed curve), and serpentinitic (dotted curve); c) quartz-
l l l l l l l l l l l l l l | l l l l ites, including purple (solid curve), green (dashed curve),
and red (dotted curve); d) schists, including muscovite
3 3.5 4 4.5 schist (solid curve), tremolite schist (dashed curve), and
chlorite schist (dotted curve); e) slates, including gray slate
Wavelength (tam) (solid curve), green slate (dashed curve), and chiastolic slate
(e) (dotted curve).
 Emissivity of Terrestrial Materials at 3-5 ltm 353

themselves show little variation from the overall broad may contain kaolinitic clay in Western Australia (R.
silicate rise in reflectance in the middle of the 3-5 Bm Lyon, personal communication). This varnish material
range. Both the felsitic (dashed curve) and chloritic commonly coats the surfaces of rocks in a desert environ-
(dotted curve) gneiss spectra also display very weak ment, and the clay S i - O reststrahlen band may sub-
quartz overtone/combination bands, while the syenite stantially mask the rock spectrum in the 8-14/~m region
gneiss spectrum (solid curve) has a weak 3.97/4.12/tm as the varnish thickens. Rock specimens selected to
doublet that we cannot explain. illustrate this effect in the previous article (Salisbury
Figure 3b shows spectra of typical marbles. As in and D'Aria, 1992a) had continuous coatings of varnish
the case of the limestone spectra described above, major that were sufficiently thick to completely obscure the
spectral features for the dolomific marble (solid curve) underlying mineral grains from view, and their spec-
are characteristically shifted slightly to shorter wave- tra were dominated by the clay reststrahlen band near
length than those of the calcitic marble (dashed curve) 9.7/2m.
by the substitution of Mg for Ca, and two additional Figure 4 shows the effect of these coatings in the
bands occur near 4.29 btm and 4.43/2m. The presence 3-5/2m region. Figure 4a depicts the spectrum of a
of an opaque mineral (magnetite?) in the serpentin- fresh surface of a rhyolite (solid curve) compared to that
itic marble (dotted curve) greatly reduces reflectance of a completely coated surface (dashed curve). The
throughout the wavelength region, as well as partially spectrum of the varnish coating is featureless in this
masking the carbonate bands. region. Its relative opacity reduces the reflectance, espe-
Figure 3c shows spectra typical of quartzites. The cially at the shorter wavelength end of the 3-5 /2m
color of quartzites is usually caused by trace amounts region, and effectively masks most of the spectral fea-
of different ferric oxide minerals or widely disseminated tures in the fresh rhyolite spectrum. These spectral
fine mica. Thus, our purple and red quartzites are col- features are due, oddly enough, primarily to a small
ored by ferric oxide, while the green sample contains amount of calcite, instead of to the quartz which is also
fine chlorite. These colorants have no spectral signatures present.
in the 3-5/2m region, and, consequently, most quartzite Figures 4b and 4c illustrate the effect of varnish
spectra are totally dominated by quartz overtone/com- coatings on mafic and ultramafic rocks, the spectra of
bination tone bands superimposed on the broad rise in which are fairly featureless to begin with. In both cases,
reflectance, as in the case of our purple (solid curve) the relative opacity of varnish and substrate is reversed
and green (dashed curve) quartzites. The red quartzite compared to the rhyolite. Thus, the varnish coating
(dotted curve), however, has calcite filling the interstices (dashed curve) slightly raises the refectance, especially
between quartz grains and displays strong carbonate at the long wavelength end of the window. Otherwise,
bands as well. the spectral effect is small.
Spectra of schists are shown in Figure 3d. Most Vegetation may also mask the spectral signatures
schists derive their schistosity from micas, and thus the of underlying rocks, changing their reflectivities and
spectra of our chlorite (dotted curve) and muscovite emissivities. When green trees or shrubs are present in
schists (solid curve) are more typical than that of the sufficient density, they are easily distinguished in remote
tremolite schist (dashed curve). The latter has a very sensing data by their characteristic reflectance signature
high reflectance (note scale change) and displays charac- in the visible / near-infrared. The mixing model problem
teristic tremolite overtone/combination tone bands becomes more difficult when the vegetation is very
near 3.18/2m and 3.35/2m, plus weak quartz features sparse and / or senescent, and senescent grass is a partic-
at longer wavelengths. ularly common scene contaminant. However, lichen
Spectra of slates in Figure 3e show a wide range of coatings on rocks may have the most subtle effect on
reflectance due primarily to the very low reflectance of rock emissivity, because they are not obvious, unless
the chiastolic slate (dotted curve) associated with the outcrops are observed closely in the field. In the 8-14
presence of very abundant opaque organic matter. Spec- /~m region, lichen acted as almost a neutral mask on the
tra of the gray (solid curve) and green (dashed curve) rock spectrum, reducing the height of the reststrahlen
slates show evidence of minor calcite in the weak car- reflectance peaks roughly in proportion to area of cover-
bonate bands near 3.5/tm and 4 gm, and the gray slate age. Figure 5 shows that in the 3-5/~m region lichen
also displays weak quartz bands near 4.5/.tin and 4.7/2m. actually increases the reflectance of a sandstone as areal
coverage increases, except in the lichen hydrocarbon
absorption bands near 3.4/~m. Thus, the average emis-
The Effect of Rock Coatings on Emissivity sivities of rocks in the 3-5/2m region may be either
Desert varnish in the United States is typically composed increased or decreased by vegetation or varnish coat-
of manganese and ferric oxides intimately mixed with ings, which must be borne in mind when modeling
montmorillonitic clay (Potter and Rossman, 1979), but surface emissivity.
354 Salisbury and D'Aria

4 0 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' 0 I l l l i l l l l l l l l l l l l l l

30 30
(1)
O Rhyolite O Basalt
t- t'-

t~ t~
O

_~ 2 0 _~ 20
n- n"

10 10

, , , , I , , , , I , , , , I , , , ,
0 I I ' , I , , , , I , , , , I , , , ,
0
3.5 4 4.5 5 3 3.5 4 4.5 5
Wavelength (t.tm) Wavelength (~tm)
(a) (b)

40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' '

30
O Ijolite
C

_~ 20
w..

o~ 10

0
3 3.5 4 4.5 5
Figure 4. Directional hemispherical reflectance spectra of
Wavelength (Ixm) fresh (solid curve) and varnished (dashed curves) rock sur-
(c) faces: a) rhyolite; b) basalt; c) Ijolite.

40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' Discussion of Rock Emissivities
The departure of reflectance from that of a black body
in AVHRR Band 3 is much greater than in Bands 4 and
30 Lichen-covered sandstone 5 for sedimentary and metamorphic rocks. Igneous rock
O
c-
spectra show a large departure from black body behavior
t~ in Band 4 in addition to Band 3. For our particular
"6 collection of rocks, which we believe is fairly representa-
%,,..® 20
tive, both the magnitude of the average reflectance and
rr /
the variability of reflectance in Band 3 increases from
/ igneous (avg--10.05, l a = 3 . 6 0 ) to sedimentary (avg--
10 / /
13.82, l a = 8.38) to metamorphic rocks (avg--18.28,

, , , , I , , , , I , , , , I , , , ,
Figure 5. Directional hemispherical reflectance spectra of li-
chen-covered sandstone with 10% coverage (solid curve)
3 3.5 4 4.5 and 65% coverage (dashed curve). Lichen was of the crus-
Wavelength (~m) tose variety.
 Emissivity of Terrestrial Materials at 3-5 l~m 3 5 5

40 ' ' ' ' 1 ' ' ' ' 1 ' ' ' ' 1 , , , , 40 .... I .... I .... I ....

30 30
{J o
t- c-

O "6
20 20
cr N
rr
I .-'" \
10 10

i ~ .' ~. .' ~. ' I I . . . . I . . . . I , , , ~ ,'I

0 I ' ' ' I , , , , I , , , , I , , I I

3 3.5 4 4.5 5 3 3.5 4 4.5 5

Wavelength (~m) Wavelength (~m)
(a) (b)

40 . . . . I ' ' ' ' I ' ' ' ' I ' ' ' '

I1) 30
o

o
20
IE

10
"/ ....-
Figure 6. Directional hemispherical reflectance spectra of
0 , , , , I , , , , I , , , , I , I I I soils: a) spectra of Entisol 0135 (solid curve), Vertisol 0475
(dashed curve), and Inceptisol 0226 (dotted curve); b) spec-
3.5 4 4.5 5 tra of Aridisol 1530 (solid curve), Mollisol 0211 (dashed
Wavelength (lLtrn) curve), and Spodosol 0127 (dotted curve); c) spectra of Alfi-
sol 0222 (solid curve), Utisol 0145 (dashed curve), and Oxi-
(c) sol 4717 (dotted curve).

l a = 11.0). Rock coatings may alter the reflectance of D'Aria, 1992a), and we do not repeat that explanation
rocks, but Band 3 variability and absolute reflectance here. Sampling locations and brief physical descriptions
typically remain greater than in Bands 4 and 5. of our soil samples are given in Salisbury and D'Aria
It is clear from the reflectance values of rocks in (1992b). Suffice it to say here that, except for the aridi-
Table 1 that, in remote sensing of their temperature, sols, soil samples were obtained from all over the United
the advantage of the lower sensitivity of Band 3 to States. The aridisols come primarily from the Mid-
error in estimating emissivity is more than offset by the dle East.
magnitude and variability of emissivity in the Band 3 Soil composition is often dominated by quartz, be-
region. Band 5 provides a much more stable emissivity cause quartz is both a common mineral and resistant to
that is also much closer to that of a black body. weathering. Because of its abundance and the fact that
the quartz reststrahlen bands are stronger than those of
any other soil-forming minerals (Salisbury et al., 1992),
SPECTRAL BEHAVIOR OF SOILS
quartz reststrahlen bands dominate soil spectral behav-
The complex soil classification system of the U.S. De- ior in the 8-14/~m region. The much weaker overtone /
partment of Agriculture Soil Conservation Service was combination tone absorption bands of quartz are seen
briefly explained in our article on the 8-14/~m spec- in the 3-5/~m region, especially the strongest of these
tral behavior of Earth surface materials (Salisbury and bands near 4.5/~m and 4.7/~m. These features can be
356 Salisbury and D'Aria

seen in Figure 6, most prominently displayed in spectra gm region than they are in the 8-14 ~tm. As in the
of the Entisol (Fig. 6a, solid curve) and Ultisol (Fig. 6c, longer wavelength region, however, lichens of all kinds
dashed curve). However, weak quartz features can be display a remarkable uniformity of spectrum in the 3-5
seen in spectra of the Vertisol, Inceptisol, and Alfisol. gm region [compare spectrum of 65% crustose lichen
Often carbonate bands are more prominent in the cover in Fig. 5 (dashed curve) with that of the foliose
3 - 5 / t m region than quartz bands, however, especially variety in Fig. 7a]. As illustrated in Figure 7a, their
the strongest carbonate doublet near 4.0/2m. Such bands spectra have a low reflectance in the region of the
are best displayed in the spectra of the Aridisol (solid fundamental H - C stretching vibration bands near 3.4
curve) and Mollisol (dashed curve) in Figure 6b. Soils pm and 3.5 gm, and rise to a much higher reflectance
of arid regions not only typically contain abundant car- at longer wavelengths. This high reflectance occurs be-
bonate, but also typically are exposed at the surface, cause they do not have the cellular water content of
and thus a factor in remote sensing observations. Be- green leaves (see below). Different lichens in our collec-
cause the carbonate bands affect emissivity in AVHRR tion vary by about 1% in reflectance at the short wave-
Band 3, abundant carbonate in soil may strongly affect length end of the 3-5/~m window and by about 3 % at
surface temperature estimates using this band. the long wavelength end.
Many soils are rich in organic matter, which affects
their spectra. However, the organic matter in soils is Green Foliage
not nearly as opaque as that found in shales and slates. Green leaves of all kinds display a uniformly low re-
The insoluble organic matter (kerogen) that survives flectance throughout the 3-5 /~m region, except for
weathering to be buried in shales is more carbon-rich, weak reststrahlen reflectance peaks near 3.43/tm and
and hence more opaque, than soluble organic matter. 3.51/tm associated with the H--C vibration bands, as
Further loss of the more volatile hydrocarbons, due to shown in Figure 7b. These vibration bands are expressed
high temperature metamorphism of shale to slate, re- as reflectance peaks, rather than troughs, because of
sults in an even more graphitic and opaque residue the dominance of surface scattering. Volume scattering,
(Rowan et al., 1991). Thus, despite the very high which would cause absorption bands, is virtually elimi-
(28.5%) extractable organic carbon content of the soil, nated due to very strong absorption by cellular water
the Spodosol spectrum in Figure 6b (dotted curve) has in the leaves, which absorbs any photons that penetrate
a relatively high reflectance over much of the 3-5/2m the surface. Like most surface scattering reststrahlen
range. The Mollisol (dashed curve) has a more typical bands, these features are offset to slightly longer wave-
and much lower extractable organic carbon content of length than the absorption band minima resulting from
2.49%, but its reflectance is not much different from volume scattering (Salisbury et al., 1992).
that of the Spodosol. This is because the overall re-
flectance is controlled more by carbonate and quartz Senescent Foliage
absorption bands than by the organic matter, except in The spectral reflectance of senescent leaves and pine
the fundamental H-- C stretching vibration bands at 3.4 needles varies sharply from that of green materials,
/2m and 3.5 gm. primarily because of the loss of cell water, but also
because the waxy cuticle flakes off of the broad leaves
Discussion of Soil Emissivities after senescence. As in the 8-14/tm region, the spectra
The relatively fine grain size of soils, which typically of the broad leaves (solid and dashed curves) and dry
contain a clay-size fraction ( < 2/~m), maximizes scatter- rye grass (dotted curve) shown in Figure 7c are close
ing, making soil reflectances in AVHRR Band 3 greater to that of pure cellulose, while that of pine needles
than that for rocks (avg--20.28 for our soil collection). (dash-dot curve) is less reflective and more featureless.
Backscatter intensity is dominated by the finer particles Without water present to absorb photons penetrating
in soils, which tend to coat the larger grains. As a result, the surface of the senescent broad leaves and grass,
there is relatively little variation in Band 3 reflectance volume scattering results in absorption band troughs in
( l a = 6.68). Both Bands 4 and 5 display a much lower the hydrocarbon bands near 3.4/2m and 3.5/2m. How-
reflectance and variability (Band 4, avg -- 3.15, ltr = 0.72; ever, reflectance from the pine needles does display
Band 5, avg= 2.43, ltr= 0.45). Thus, as in the case of some surface scattering, presumably because the resin-
rocks, Band 3 offers no advantage for remote sensing of ous material of the needle is highly absorptive, with or
temperature. without water present.

Bark
SPECTRAL BEHAVIOR OF VEGETATION
The similarity of spectra of different barks in Figure 7d
Lichens reflects the fact that all bark is tignified cellulose. The
As pointed out under the discussion of lichens as a rock differences in overall reflectance are apparently related
coating material, lichens are more reflective in the 3-5 to differences in texture, which modifes scattering.
 Emissivity of Terrestrial Materials at 3-5 ltm 357

40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' 40 , , , , i , , ' , i , ' ' ' I ' ' ' '

30 30
o
t'-
Lichen o
t'-
Green foliage
t~
,,I--,
o o
_~ 20 _~ 20
rr n-

~ 10 10

, , . , I . . , , I , , , , I , , , , i i i i I I I i i i i ! ! i i i i

3.5 4 4.5 5 3 3.5 4 4.5

Wavelength (~m) Wavelength (~tm)
(a) (b)

40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' '
40

S e n e s c e n t foliage

(D
30 I'
J 30 Bark
O o
t.- t-

g 11 f ~ ~ .---'
o
i® 20 _~ 20
rr
iI ~i ~
10 'I I ~ ' 10
#

:. . . . - - - I i I I I . . , , I . , , . I , . , ,

3 3.5 4 4.5 5 3 3.5 4 4.5 5

Wavelength (I.tm) Wavelength (!am)
(c) (d)

40 , , , , l , , , , i , , , , i , , , ,

3O
,'- Soil litter
o
Figure 7. Directional hemispherical reflectance spectra of
_e 20 representative samples of vegetation cover: a) spectrum of
n" the foliose lichen Parmotrem (sp?), a tree lichen from Mary-
land; b) spectra of typical green foliage, including red oak
leaves (solid curve), Indian grass blades (dashed curve), and
white pine needles (dotted curve); c) spectra of typical se-
nescent foliage, including beech (dashed curve), red oak
leaves (solid curve), rye grass (dotted curve), and white pine
0 needles (dash-dot curve); d) spectra of typical bark samples,
including white oak (solid curve), yellow poplar (dashed
3 3.5 4 4.5 5 curve), and Loblolly pine (dotted curve); e) spectra of typi-
Wavelength (I.tm) cal decaying soil litter, including deciduous leaf fragments
and twigs (solid curve), mostly decaying wood (dashed
(e) curve), and decaying white pine needles (dotted curve).
358 Salisbury and D'Aria

Soil Litter emissivity of essentially 1.0 in all bands at all seasons.

Spectra of decaying soil litter are shown in Figure 7e. Thus, there is a seasonal effect that must be considered
Freshly fallen leaves, bare wood, and dry grass display along with cover type in assigning emissivities to for-
the spectrum of cellulose but, as this material begins to ested or grassland surfaces.
decay, its spectrum becomes flatter and less reflective.
In the previous article (Salisbury and D'Aria, 1992a), we SPECTRAL BEHAVIOR OF WATER
suggested that the ultimate reflectance of fully decayed
organic materials is probably well represented by the Sea Water
spectrum of the organic-rich Spodosol surface horizon Water is often assumed to have an emissivity of 1.0, but
shown in Figure 6b (dotted curve). While this may be it is well known that this is not true (Pinkley and
true in the 8-14 Mm region, the overall reflectance of
the soil is higher than that of our litter samples, probably
because of the presence of some fine quartz in the soil. Figure 8. a) Directional hemispherical reflectance spectra
of sea water (solid curve) and sea foam (dashed curve), b)
Directional hemispherical reflectance spectra of distilled wa-
Discussion of Vegetation Emissivity ter with suspended quartz sediment: Solid, dashed, and dot-
Lichens. Lichens may be locally abundant on the sur- ted curves represent spectra recorded at 7, 23, and 64 min
faces of rock types favorable for their growth, especially after mixing, as the quartz grains settle out; these curves
overlap so closely that they cannot be distinguished.
in high alpine and Arctic climates. However, complete
coverage is rare and lichens appear slightly transparent 40 , , , , I , , , , i , , , , i , , , ,
(1-2%) to rock reflectance as well. Thus, lichens can
be a locally significant factor in rock reflectance, but do
not typically dominate over large areas.
(D 30
Green vegetation. Green vegetation, on the other 0
t,-
hand, can completely cover a large surface. Based on t~
Seawater and foam
their reflectance, the emittance from individual leaves 0

and conifer needles is close to that of a black body in _m 2o

I1)
the 3-5/~m region. As discussed in our previous article rr
(Salisbury and D'Aria, 1992a), black body behavior will
be enhanced by canopy scattering, Thus, closed cano- 10
pies can reasonably be assigned an emissivity of 1.0.
Clearly, this value must decrease as vegetation becomes
sparse and soil or soil litter and bark contribute to , , , , I , , , , l l , , , l l
0 i I

emittance. However, emissivity should remain close to

3.5 4 4.5 5
1.0 as long as the leaf area index is great enough
to suggest multiple scattering in the canopy. Wavelength (gm)
AVHRR Bands 3, 4, and 5 display about the same (a)
reflectances for green vegetation. Thus, from emissivity
considerations alone, the lower sensitivity of Band 3 to 40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' '

emissivity errors is an advantage for remote sensing of

the temperature of green vegetation. The greater energy
available in Bands 4 and 5, and the lesser effect of 30
atmospheric water vapor on energy received at the o

sensor in these bands, are additional variables not con-

r- Suspended quartz sediment
sidered here. Obviously, these considerations might dic-
~ 20
tate the use of the longer wavelength bands, despite
the greater effect of emissivity error in those bands. rr
Senescent vegetation. Emittance from individual se-
nescent broad leaves in the 3 - 5 / t m region is very far 10
from that of a black body, as is that of bark. Thus,
deciduous forests in the Fall, with individual trees sur-
, , , , I , , , , I i i i i I i
rounded by fallen leaves, may display an emissivity 0 i !

sufficiently far from black body behavior in AVHRR 3.5 4 4.5 5

Band 3 to counteract any emissivity error advantage
Wavelength (gm)
compared to Bands 4 and 5. This is also true for dried
grass, while conifers would be expected to retain an (b)
 Emissivity of Terrestrial Materials at 3-5 Ilm 359

Williams, 1976), as we illustrated in our previous article in refractive index occur, despite similar absorption
(Salisbury and D'Aria, 1992a) for the 8-14/~m region. coefficients, because of differences in lattice vibrations.
Figure 8a (solid curve) shows the directional hemispher- It is reasonable to assume that the presence of dissolved
ical reflectance of sea water in the 3-5 pm region. salt would, indeed, distort the crystal lattice, while the
Although Pinkley and Williams (1976) point out the fundamental absorption coefficient of water ice could
small spectral differences between distilled water and remain unchanged.
sea water, these differences are too small to affect most Whatever the cause of spectral differences in the
temperature estimates. In fact, they are of the same 8-14 /~m region, the two kinds of ice appear quite
magnitude as small differences in emissivity expected similar in their spectral behavior in the 3-5/tm region.
from temperature changes alone (Pinkley et al., 1977). The relative heights of the reflectance peak near 3.1
To this point, we have not considered the effects /~m in Figure 9a may be due to small differences in
of the bi-directional reflectance function (BDRF) on surface roughness of these particular samples, as other
reflectance and emissivity, because the materials consid-
ered so far have had rough surfaces or complex canopies
from which we expect generally uniform directional
spectral behavior, except perhaps at grazing angles (Nor- Figure 9. Directional hemispherical spectra of sea water ice
man et al., 1990). Water may differ in this regard, (solid curves) and distilled water ice (dashed curves): a) orig-
inal smooth surfaces; b) after roughening with 100-grit sand-
because of its exceptional range of possible surface paper,
roughness. However, Fresnel equation calculations of
directional reflectance based on the optical constants of 40 .... I .... I .... I ....
Pinkley et al. (1977) show that there is little effect until
an angle of about 45 ° is reached (see also Fig. 3-111 in
Wolfe and Zissis, 1989). At that point, reflectance begins Sea ice and distilled water ice
G)
30
to rise, but the shape of the spectral curve remains tj smooth
E:
constant. The statistical distribution of wave slopes, t~
tj
which is largely a function of wind speed (Cox and 20
Munk, 1954), can be combined with calculated direc-
tional reflectance to yield an average reflectance or rr
emissivity under different sea surface conditions. JI

Despite its high absorption coefficient, water does 10

not entirely obscure the reststrahlen bands of suspended
sediment in the 8-14 /~m region, at least when that
sediment is quartz with very strong reflectance peaks 0
(Salisbury and D'Aria, 1992a). However, Figure 8b 3.5 4 4.5
shows that suspended sediment, lacking strong bands
Wavelength (~m)
in the 3-5/~m region, has no effect on water reflectance
(a)
there.
40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' '
Ice
Salisbury and D'Aria (1992a) made preliminary spectral
measurements of sea water ice and distilled water ice Sea ice and distilled water ice
(1) 30
which showed that the spectra of smooth samples o rough
c
differed substantially in the 8-14 pm region, while
.i,-,
roughening the surfaces with 100 grit sandpaper re- o

sulted in very strong spectral similarity. The spectral 20

(1)
similarity of rough surfaces was attributed to the similar- rr
ity of absorption coefficient of the two ices, which is
the most important parameter when volume scattering 10
dominates reflectance. Surface scattering becomes more
important for smooth surfaces and involves the refrac-
tive index as well as the absorption coefficient, so it was I I i i I , , , , I , , , , I , , , I

concluded that sea water ice and distilled water ice

3.5 4 4.5
differed in refractive index to account for their spectral
differences when smooth. What was not offered was an Wavelength (I.tm)
explanation of how this can occur. Such differences (b)
360 Salisbury and D'Aria

40 ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' by about 1%, we conclude that all ice has essentially
the same spectral signature when rough, which was our
conclusion in the 8-14/lm region as well.
30
¢J
¢.- Coatings
t~
.a..J Crude oil slicks Different coatings on water naturally may change its
0
20 spectral behavior. Common coatings are foam, oil slicks,
(9 and windblown dust.
rr
Volume scattering in bubbles makes sea foam bright
o~ in the visible, but our previous article (Salisbury and
10
D'Aria, 1992a) showed that the absorption coefficient
of water is so high in the 8-14/lm region that negligible
volume scattering takes place in foam bubbles. Conse-
l l i , l i . , , l ~ i l I [ l l l l
0 quently, foam is not brighter than sea water in the
3.5 4 4.5 5 8-14/tm region. This is also true in the 3-5/lm region,
Wavelength (~m) as indicated by the fact that sea water and foam have
very close to the same reflectance in Figure 8a.
(a) Crude oil slicks were made in the laboratory by
pipetting known volumes of oil onto a sea water sample
40 in a Petri dish. These oils are of very different compo-
sitions and viscosities, as described in Salisbury and
D'Aria (1992a), but display virtually identical infrared
30 spectra, as shown in Figure 10a. A detailed study of
O
i- these slicks is reported in Salisbury et al. (1993). Briefly,
N Floating sediment it was found that spectra of oil slicks are fiat with a
0
m 20 reflectance of about 4% throughout the spectral range
from 3-14/~m, with the exception of the small reststrah-
len reflectance peaks associated with the C - H bands.
10 This spectral behavior is essentially not affected by slick
thickness or age, and can be used to detect oil slicks
by its contrast with the variable spectral reflectance of
I I I ' I . ~ , . I . i I I I I ' ' '
water in the 8-14/Jm region.
Very fine windblown dust may float on a calm
3.5 4 4.5 5 water surface, supported by surface tension. Such dust
Wavelength (gm) particles typically agglomerate into small patches a few
millimeters across, and these patches may coalesce into
(b)
larger patches or streamers under the influence of slow
Figure 10. Directional hemispherical reflectance spectra of current action. Such coatings are ephemeral and easily
floating materials on water: a) continuous oil slicks 2-3 mm destroyed, but have a large effect on the spectrum of
thick on sea water made with five different crude oil sam- water in the 8-14 pm region, because that is where the
ples; b) patchy monolayers of very fine windblown quartz silicate reststrahlen bands lie. In the 3-5/~m region,
powder (solid curve) and a montmorillonitic soil (dashed
curve) floating on tap water; these curves overlap so closely however, the water spectrum is unaffected by such
that they cannot be distinguished. coatings, as can be seen in Figure 10b.

Discussion of Emissivities of Water and Ice

Water. Water is closer to a black body in AVHRR Band
samples showed more similar spectral features here. It 4 than in Bands 3 and 5, but water should provide a
is, in any event, the spectral behavior of rough surfaces target with a known and relatively uniform emissivity
that are of most importance, because we expect most in all bands, compared to land surfaces. Surface coatings
natural ice surfaces to be rough. The biggest spectral and suspended sediment change the emissivity of water
change from smooth to rough surfaces is the introduc- very little, except for oil slicks. However, extreme sur-
tion of a broad reflectance maximim centered about 3.5 face roughness can raise reflectance substantially in all
/~m. While there appears to be a minor shift in the peak bands, making the lower sensitivity of Band 3 to emissiv-
of this maximum for sea ice compared to distilled water ity errors more attractive if surface roughness is not
ice and average reflectances in AVHRR Band 3 differ known.
 Emissivity of Terrestrial Materials at 3-5 I~m 361

Ice. As for ice, Band 4 offers an emissivity closer 2.5 and 8.0 to 12.0/lm) of Phanerozoic strata, Wind River
to a black body and less variable than do both Bands 3 Basin and southern Bighorn Basin areas, Wyoming, J. Sed.
and 5. Pet. 60:504-524.
Lyon, R. J. P. (1964), Evaluation of Infrared Spectrophotometry
for Compositional Analysis of Lunar and Planetary Soils.
We are grateful to Ken Watson for a very helpful critique, and Part II: Rough and Powdered Surfaces, NASA-CR-100,
to a number of people who provided samples for measurement,
including Cathy Ager for lichen samples, Larry Brown for soil Washington, DC, 262 pp.
samples, Michael Chenevey for varnished rhyolite and basalt McClain, E. P., Pichel, W. G., and Walton, C. C. (1985),
samples, Doug Nash for sea water, John Norman for prairie Comparative performance of AVHRR-based multichannel
grass samples, and Floyd Sabins for oil samples. Larry Rowan sea surface temperatures, J. Geophys. Res. 90:11,587-
collected the sample and measured the spectrum of the fresh 11,601.
and varnished ijolite on spectrometer equipment similar to ours' Norman, J. M., Chen, J., and Goel, N. (1990), Thermal emissiv-
at the U.S. Geological Survey (USGS) in Reston, Virginia. Vicki ity and infrared temperature dependence of plant canopy
Comer of the USGS identified the lichen samples. This research architecture and view angle, in Proceedings of the IEEE
was supported by the Solid Earth Science Branch of the U.S. International Geoscience and Remote Sensing Symposium,
National Aeronautics and Space Administration in support of 20-24 May, College Park, MD, Vol. 3, IEEE, New York,
the Earth Observing System (EOS).
pp. 1747-1750.
Pinkley, L. W., and Williams, D. (1976), Optical properties
of sea water in the infrared, J. Opt. Soc. Am. 66:554-558.
REFERENCES Pinkley, L. W., Sethna, P. P., and Williams, D. (1977), Optical
constants of water in the infrared: Influence of temperature,
Cox, C. and Munk, W. (1954), Statistics of the sea surface Opt. Soc. Am. 67:494-499.
derived from sun glitter, Sears Found. J. Mar. Res. 13:198- Potter, R. M., and Rossman, G. R. (1979), The manganese
227. and iron-oxide mineralogy of desert varnish, Chem. Geol.
Drummond, J. R. (1991), Measurements of pollution in the 25:79-94.
troposphere (MOPITT), in The Use of EOS to Study Atmo- Price, J. C. (1983), Estimating surface temperatures from
spheric Physics (J. Gille and G. Visconti, Eds.), North- satellite thermal infrared d a t a - a simple formulation for
Holland, Amsterdam. the atmospheric effect, Remote Sens. Environ. 13:353-361.
Estep-Barnes, P. A. (1977), Infrared spectroscopy, in Physical Rowan, L. C., Salisbury, J. W., Kingston, M. J., Vergo, N. S.,
Methods in Determinative Mineralogy, 2nd ed. (J. Zussman, and Bostick, N. H. (1991), Evaluation of visible, near-
Ed.), Academic, New York, Chap. 2, p. 529-603. infrared and thermal-infrared reflectance spectra for study-
Farmer, V. C., Ed. (1974), The Infrared Spectra of Minerals, ing thermal alteration of Pierre shale, Wolcott, Colorado.
Monograph No. 4, Mineralogical Society, London, 539 pp. J. Geophys. Res. 96:18,047-18,057.
Farmer, V. C., and Palmieri, F. (1975), The characterization Salisbury, J. W., and D'Aria, D. M. (1992a), Emissivity of
of soil minerals by infrared spectroscopy, in Soil Compo- terrestrial materials in the 8-14/~m atmospheric window,
nents, Vol. 2. Inorganic Components (J. E. Gieseking, Ed.), Remote Sens. Environ. 42:83-106.
Springer-Verlag, New York, pp. 573-670. Salisbury, J. W., and D'Aria, D. M. (1992b), Infrared (8-14
Fujisada, H., and Ono, M. (1991), Overview of ASTER design /~m) remote sensing of soil particle size, Remote Sens.
concept. Soc. Photo-Opt. Instrum. Eng. 1490:320-342. Environ. 42:157-165.
Hovis, W. A., and Callahan, W. R. (1966), Infrared reflectance Salisbury, J. W., Walter, L. S., and D'Aria, D. (1988), Mid-
spectra of igneous rocks, tufts and red sandstone from 0.5 Infrared (2.5 to 13.1~m) Spectra of Igneous Rocks, USGS
to 22 I~m, J. Opt. Soc. Am. 56:639-643. Open File Report 88-686, Washington, DC, 126 pp.
Hunt, G. R., and Salisbury, J. W. (1975), Mid-Infrared Spectral Salisbury, J. W., Walter, L. S., Vergo, N., and D'Aria, D. M.
Behavior of Sedimentary Rocks, Air Force Cambridge Re- (1992), Mid-Infrared (2.1 to 25 I~m) Spectra of Minerals,
search Laboratories Technical Report, TR-75-0356, Bed- Johns Hopkins University Press, Baltimore, 296 pp.
ford, MA, 49 pp. Salisbury, J. w., D'Aria, D. M., and Sabins, F. F. (1993),
Hunt, G. R., and Salisbury, J. W. (1976), Mid-Infrared Spectral Thermal Infrared remote sensing of crude oil slicks, Remote
Behavior of Metamorphic Rocks, Air Force Cambridge Re- Sens. Environ., forthcoming.
search Laboratories Technical Report, TR-76-0003, Bed- Salomonson, V. V., Barnes, W. L., Maymon, P. W., Montgom-
ford, MA, 67 pp. ery, H. E., and Ostrow, H. (1989), MODIS: advanced
Key, J., and Haefliger, M. (1992), Arctic ice surface tempera- facility instrument for studies of the Earth as a system,
ture retrieval from AVHRR thermal channels, J. Geophys. IEEE Trans. Geosci. Remote Sens. 27:145-153.
Res. 97:5885-5893. Wolfe, W. L., and Zissis, G. J. (1989), The Infrared Handbook.
Lang, H. R., Bartholomew, M. J., Grove, C. I., and Paylor, Environmental Research Institute of Michigan, Ann Arbor,
E. D. (1990), Spectral reflectance characterization (0.4 to 1710 pp.