In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Voi. XXXVIII, Part 7B 
SMALL-SCALE ROUGHNESS EFFECTS ON THERMAL IR SPECTRA 
L. K. Balick 3 ’ *, I. Danilina b , A. Gillespie b , J. Jolin c , A. Mushkin d 
a Space and Remote Sensing Sciences, Los Alamos National Laboratory, Los Alamos, NM, USA - lbalick@lanl.gov 
b Earth and Space Sciences, University of Washington, Seattle, WA, USA - (arg3, Danilina)@uw.edu 
c Physical Chemistry and Applied Spectroscopy, Los Alamos National Laboratory, Los Alamos, NM, USA -jolin@lanl.gov 
d Geological Survey of Israel, Jerusalem IS, Mushkin@gsi.gov.il 
Commission VII Symposium 
KEY WORDS: Hyper spectral, Multispectral, Thermal, Infrared, Surface, Radiometry, Measurement 
ABSTRACT: 
Thermal infrared (IR) spectra of materials are affected by subpixel surface roughness that increases interactions between surface 
facets, thereby shifting the spectra toward a blackbody. Roughness also creates directional effects due to differential solar heating 
and view geometry. Three types of ground-based experiments were conducted to quantify roughness effects at scales of about 10 cm 
or less. First, a radiosity model was implemented and validated for natural and artificial surfaces using an imaging spectrometer. 
Area and resolution create practical computational limits so most simulations were performed at a resolution near 1 cm over a 1 m 
area. Surfaces were specified using laser profilometer data but can be simulated. Second, a well-calibrated radiometer was used to 
measure radiance for emissivity retrievals of different sized gravels and at different solar and view geometries. Finally, a reflectance 
spectrometer measured spectra for soft rocks sanded to different roughnesses. 
Measurements of soft rocks with single mineral features (alabaster, soapstone, and chlorite) sanded to different roughnesses show a 
decrease of spectral peak height with roughness when the roughness scale is significantly larger than the wavelength. Precise 
measurements of two types of gravel, in three size classes of gravels, with a non-imaging spectrometer show an apparent saturation 
of roughness effects and a probable increase of directional effects with roughness. The modelling results show that a simple radiosity 
model can broadly simulate the effects of roughness. The shape of the roughness elements has a significant impact. Imaging 
spectrometers permit observation of small-scale spatial variations, which are not observed at pixel scales of a meter or more. Spectra 
go to blackbody spectra in small cracks and crevices. 
1. INTRODUCTION 
1.1 Background and Objectives 
The radiance observed by remote sensing platforms varies 
significantly with subpixel surface roughness. Subpixel surface 
roughness occurs at spatial scales at which facets of the surface 
interact with each other resulting in multiple reflection or 
absorption/emission of photons but are not resolved in an 
image. The concept is quite simple but the spatial limits are not 
abrupt and hard to quantify. They vary with wavelength at the 
lower limit and pixel size at the upper limit. At the small end of 
the scale, facets or characteristic dimensions are large enough— 
on the order of 100 times the wavelength—that non-linear 
scattering processes are unimportant. At the upper limit, the 
dimensions need to be small enough so that a pixel contains a 
representative sample of the geometric variation of the surface: 
perhaps a tenth of the pixel area. The key concept is that 
subpixel roughness effects occur between surface facets that are 
not resolved in a remotely sensed image pixel. The net effects 
include darkening pixels in the reflectance domain, brightening 
pixels in the emissive domain (increasing both temperature and 
emissivity), reducing the depth of spectral features, and 
producing directional variations of observed radiance. Surfaces 
composed of different materials (stones in a soil matrix, for 
instance) will undergo non-linear spectral mixing when facets 
of different materials interact with each other. At thermal 
infrared wavelengths, the effect is that emissivity spectra are 
moved closer to a blackbody spectrum in which extremely 
rough surfaces or cavities have an emissivity approaching 1.0 at 
all wavelengths. (The effects are broadly similar with surface 
roughness and volume scattering, such as occurs with 
vegetation or loosely packed particles, but volume scatting is 
more complex.) This paper presents studies of subpixel 
roughness effects across a variety of spatial scales in the long 
wave infrared (LWIR, 8- \4 ¡um) spectral range. 
Because pixels can be quite large, they can contain subpixel 
surface roughness effects across a range of spatial scales and 
phenomena. Typically, one spatial scale dominates observations 
or, at least, one spatial scale is of primary interest. Therefore, 
studies of roughness effects often concentrate on a single scale 
or phenomenon. 
1.2 Objectives 
This paper presents summaries of four studies of subpixel 
roughness effects that range from the upper limits to the lower 
limits of spatial scales relevant to remote sensing. First, at the 
finest scale, spectral measurements of rock surfaces roughened 
with different grits of sandpaper are examined and related to 
micro-profilometer measurements of the surfaces. 
Corresponding author