'RS, Vol. XXXVIII, Part 7B 
In: Wagner W„ Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Voi. XXXVIII, Part 7B 
57 
>ECTRA 
lbalick@lanl.gov 
ilina)@uw.edu 
JSA -jolin@lanl.gov 
factions between surface 
differential solar heating 
s at scales of about 10 cm 
an imaging spectrometer, 
ion near 1 cm over a 1 m 
d radiometer was used to 
•ies. Finally, a reflectance 
erent roughnesses show a 
the wavelength. Precise 
7w an apparent saturation 
ow that a simple radiosity 
nificant impact. Imaging 
' a meter or more. Spectra 
at emissivity spectra are 
aim in which extremely 
ssivity approaching 1.0 at 
idly similar with surface 
such as occurs with 
5, but volume scatting is 
its studies of subpixel 
ipatial scales in the long- 
:ral range. 
ley can contain subpixel 
lge of spatial scales and 
e dominates observations 
mary interest. Therefore, 
:entrate on a single scale 
bur studies of subpixel 
upper limits to the lower 
lote sensing. First, at the 
rock surfaces roughened 
examined and related to 
e surfaces. 
Figure 1. Flat rock surfaces sanded to U.S. grit sizes shown. 
Alabaster is at the upper left, soapstone at the upper right, 
and the two chlorite slabs are at the bottom. 
Secondly, retrievals of emissivity of gravels of different sizes 
are made using data from a well-calibrated, highly stabilized 
FTIR spectrometer in the field, Balick & al., (2009). Third, a 
radiosity model of the impacts of roughness at centimetre scales 
was developed and verified. Simulations and data from special 
cases are presented. Finally, multi-directional satellite images 
are used to retrieve roughness information about the surface. 
2. SURFACE ROUGHNESS STUDIES 
2.1 Roughened Surfaces 
Three types of fairly soft rocks were first smoothed and 
flattened, then hand-sanded with different grits, ranging from 
very smooth to very coarse and using diamond micromesh; 
about 2 pm to 350 /im average grit size. The rock types are 
commonly used in sculpture and are alabaster (gypsum, a 
sulfate), soapstone (talc, a phyllosilicate), and chlorite (another 
phyllosilicate closely related to soapstone). The rock surfaces 
and grit sizes are shown in Figure 1. Note that the different 
roughnesses vary in appearance in the visible wavelengths with 
the rougher surfaces appearing as a flat grey and becoming 
darker or more colourful with smoothness. Diffuse reflectance 
thermal IR spectra were measured with an A2 Technology 
Exoscan FTIR spectrometer (A2 Technology, 2010). Only the 
measurements in the LWIR (8-14 /urn) were used. Like most 
diffuse reflectance spectrometers, a small spot on the surface 
was illuminated with a broad-band source, and the energy 
reflected was measured at some range of off-nadir angles. 
Spectral Peak Height Variation with Grit Size 
Figure 2. Plots of the spectral peak heights with mean grit 
size of the sandpaper used to roughen the rock. This is not 
the same as the actual roughness of the surface, but it is 
proportional to it. 
The measurement is only truly diffuse reflectance if the surface 
is a diffuse reflector, and then Kirchhoffs Law holds: 
reflectance, p, is related to emissivity, e, by p = 1 - e at any 
wavelength. The measure of roughness effects used here is the 
height of the reflectance peak. 
Figure 2 shows the height variation of the spectral features with 
grit size for all three rock types. At values of sandpaper grits 
greater than about 25 pm, diffuse reflectance decreases with 
roughness, albeit slowly, for all three rock types. Below 25 pm, 
the curves are very steep. Soapstone and chlorite, which are 
closely related (both are steatites) increase to a peak around 25 
pm, while the alabaster diffuse reflectance decreases in this 
range. At roughness sizes near the size of the wavelength, non 
linear scattering processes occur and reflectance is no longer 
dominated by interactions between facets. This might be seen in 
these measurements between 2 pm and 25 pm, with a possible 
transition out to 50 pm. Of course, sandpaper grit sizes do not 
actually represent the actual surface roughness. The surfaces 
were scanned with a Nanovea PS-50 (Nanovea, 2010) optical 
microprofilometer at a nominal resolution of 10 pm in x, y, and 
z. The sizes of the finest grits (1,000 US grit scale and higher) 
are below the resolution of the profilometer. Also, the measured 
root mean square RMS values of the rock surfaces are well 
below the average grit size of the sandpaper so the values of grit 
size cannot be interpreted as the same as the surface roughness, 
just proportional to it. The actual roughness for the very fine 
grits must be viewed with caution. Nevertheless, spectral peak 
heights do decrease with roughness down to spatial scales that 
approach the wavelength. 
2.2 Gravel: Spectral Emissivity Retrieval 
Highly accurate and precise measurements of spectral 
emissivity in the field are notably difficult to make for a variety 
of reasons, including characterization of “downwelling” 
radiance from the surroundings and sensor calibration and 
stability. In this work, careful measurements were made with an 
extremely well-calibrated and stabilized FTIR for different 
roughness, nadir angle, material type, and time of day. The 
retrieval of emissivity generally follows that described by 
Salvaggio & Miller (2001) and is conceptually simple. The