453 
overflown these same sites with the NASA DC-8 aircraft fitted with AIRSAR, a three frequency, polarimetric SAR 
developed by JPL operating at P, L and C bands [Evans et al., 1986; Vogt, and Kobrick, 1991]. Multiple 
overflights of each site created a data set from which G° can be calculated at multiple incidence angles, frequencies 
and polarizations. In previous reports we have shown that in desert areas L-band (24-cm wavelength), cross- 
polarized (HV) radar backscatter taken at an incidence angle of 40 degrees demonstrates the best-behaved 
relationship to z Q [Greeley el al., 1991b]. Desert sites were chosen on lava flows, alluvial fans dominated by sand 
and gravel of differing ages and roughnesses, and on a silt-clay playa. There, roughness varied from a few cm to a 
few tens of cm, although rills and small washes 1 to 2 m wide were common. All sites were unvegetated with the 
exception of occasional creosote bushes in the washes and isolated patches of grasses. Signs of active sand transport 
and in some cases of abrasion, were evident 
We decided to compare this relationship to data from vegetated sites where wind erosion is a potential problem. In 
the desert sites surface roughness has two components: (1) size and arrangement of surface particles and (2) 
submeter-scale roughness of the surface. Vegetation adds not only the complication of a third factor (the new 
surface consisting of the stems, branches, and leaves), but also adds a possible new interaction mechanism because 
the surface is now "thick”; that is, it is at least partly penetrable by both microwaves and the wind. In the case of 
microwave energy, this adds the possibility of multiple-bounce geometries which tend to depolarize the returned 
energy. In the case of the wind the effect is far less clear. 
A site near the village of Stauning, Denmark, in the river Skjem delta area near Ringkobing Fjord, has been 
equipped with meteorological instruments for some time and was selected for study. The area is agricultural, 
including a variety of tree and crop types. Of the eight areas selected for this study, seven were sewn in various 
crops ( e. g„ barely, rape, beet, peas and grasses), and one is a stand of Spruce trees. Vegetation height varies from 
tens of centimeters to several meters, and the corresponding roughness lengths are z Q = 0.02 to 0.75 [Rasmussen et 
al., 1993]. Unlike the desert sites of the previous study, here all sites were essentially 100% vegetated. However, 
evidences of active aeolian transport were present, and 0.5 to 1 cm low rills were observed with spacing of 5 -10 cm 
or larger. 
DATA COLLECTION 
The DC-8 overflight was organized and executed in July 1991. The AIRSAR obtains data at three wavelengths 
simultaneously with 12.5 m resolution, but multiple flights over targets are required to assemble a multiple-angle 
data set. Radiometric calibration accuracy of the resulting images is estimated at +/- 3 dB [VanZyl, 1990]. Denmark 
targets, including the Stauning site, were imaged at multiple incidence angles near 1400 UT. A set of trihedral 
comer reflectors placed on the North Sea coast was also imaged to ensure radiometric calibration of the data. 
Images such as that shown in Figure 1 were produced from the AIRSAR data, and radiometric calibration was 
checked using the known response of the comer reflectors. Eight areas for which z Q is known or could be inferred 
from field work were identified in the images. 
Soil moisture samples were taken from some of the sites, and local observations were recorded. All sites had 
volumetric moisture content near 30%. Contemporary meterological data were obtained at the Stauning site, 
although the z Q data used in this study is from the larger set of data available for the site, as described below. 
ANALYSIS 
Patches representing the eight test areas were extracted from the data and were averaged in power to create estimates 
of L-band HV backscatter cross-section. These were compared with the corresponding roughness measures, 
stimates were made for all other wavelengths and angles and for HH and VV data but were not used for this study, 
olarization signatures were constructed for each site and checked for signs of contamination or other problems, and 
(with the exception of the P-band data, which was not used here) all data were judged to be reliable. 
Wind profile data for the Stauning site are available from June 1991. Those showing near-neutral thermal stability 
were used to calculate values of z Q , and these latter data were found to be uniform between 3 to 5 cm at the main 
site. Hot-wire anemometer data were taken at 1.5 m above the surface on three other sites for both morning and 
temoons. A proportionality constant of 0.75 was assumed to relate zero-plane displacement to the measured crop