(a) (b; (c) 
(d) 
(e) 
(Í) 
Figure 9. The relative contributions, according to the assumptions of the multi-level model, from lay 
ers in the forest canopy and the ground. Each bar represents a 50% confidence interval for the corres 
ponding estimates. In each figure 2, 3 or 4 results are grouped for each layer. 
rences between relevant scatterers are not accounted 
for is hard to predict. The error becomes negligible 
when all contributions originate from a small 
layer. This situation might occur when (a) the 
forest is very low or (b) the forest has a closed 
and smooth canopy and attenuation is very strong (at 
the higher frequencies). 
But when the canopy is not closed, or has a rough 
surface (emergent trees), or the attenuation is 
not strong (lower frequencies) the effect is far from 
predictable. In these situations only a worst case 
error can be indicated for a given height of flight, 
canopy height and incidence angle. 
7. CONCLUSIONS 
By means of large corner reflectors the Dutch 
X-band SLAR can be used for the acquisition of data 
on the attenuation properties of forest canopies. The 
attenuation was found to be primarily determined by 
canopy architecture (i.e. factors like crown closure, 
crown shape etc.) and aspect of measurement. The 
attenuation of crowns of deciduous trees is strong. 
For crowns of coniferous trees it may be lower but 
still is in the order of 20 dB (two-way). 
The multi-band scatterometer DUTSCAT was used to 
conduct the same experiment in the L-band but the 
acquired data have not been analysed yet. 
The attenuation properties are of direct impor 
tance when the influence of the soil, soil moisture 
content, standing water or undergrowth under a fo 
rest canopy has to be modelled. 
The DUTSCAT when operated from relatively low 
altitudes can provide information on the vertical 
distribution of backscattering. This may be done 
through inversion of the multi-level model introdu 
ced in section 4. The y value of a forest stand 
could be divided in contributions from a number of 
layers (3 or 4) of the forest. In C-band at low 
incidence angles, it was found that a substantial 
amount of the backscattering from poplar stands 
originates from the ground and understory. In L-band 
at low incidence angles ~75% of the backscattering 
originates from the ground. Apparently the poplar 
crowns are very transparent at this frequency. As 
expected the relative contributions from the lower 
layers decreased with increasing incidence angle. 
The accuracy of these measurements can be influenced 
by measurement geometry. When more detailed infor 
mation is desired the DUTSCAT has to be modified. 
The range resolution has to be increased and the 
beam width in range direction decreased. 
Measurements of additional physical parameters 
of forest canopies, other then y or 0°, can simpli 
fy the model-making effort. This was illustrated 
by the cloud model. Some parameters of the cloud 
model could be determined directly through inversion 
of the appropriate form of the multi-level model. 
The multi-level model is also a useful tool in the 
pre-processing of scatterometer data of forests. In 
this model the radar equation is applied for each 
level individually and as a result the calculation 
of y is more accurate. In fact when the model is not 
used for the calculation of y or o° gross errors may 
occur. 
8. ACKNOWLEDGEMENTS 
The author would like to acknowledge the support of 
the working group ROVE-Forestry. Financial support 
for the radar flights was provided by the Netherlands 
Remote Sensing Board (BCRS). 
9. REFERENCES 
Attema, E.P.W. and F.T. Ulaby, 1978, Vegetation 
modelled as a water cloud, Radio Science, Vol. 13, 
Figure 10. : 
as a fune tii 
incidence ai 
in the secoi 
of 1800 metí 
No. 2, pp 
Attema, E.P 
a 6-frequ( 
EARSeL Woi 
to vegetai 
ESA SP-22; 
Hoekman, D.l 
A multila) 
vegetatior 
Remote Ser 
Voi. 2, T/ 
Hoekman, D.E 
backscatte 
classificò 
Colloquiun 
remote sen 
SP-247, pp 
Ulaby, F.T., 
Microwave 
Addison-We