The Mimics model is first tested against data acquired on various cultivated canopies, using an airborne 
scatterometers in C and X bands, at various incidence angles. It is then shown that “water-cloud” models, 
using LAI or water content to characterize the vegetation and surface moisture to characterize the soil 
adequately represent the canopy backscattering coefficients simulated by Mimics. This allows the complex 
outputs of Mimics to be summarized with a limitated set of parameters, being related to the albedo and 
the optical depth of the vegetation on the one hand, and to the response of the underlying soil on the other 
hand. Finally, an inversion algorithm allowing the estimation of canopy biomass and surface soil moisture 
from multi-fequency or multi-angle radar data is tested on these simulated data. 
The method we propose in this paper is to fit the parameters of semi-empirical models on radar databases 
generated by a theoretical model of radar backscatter. Let us first present the two types of models used 
The theoretical model: MIMICS 
Developed at the University of Michigan, Mimics is a first order radiative transfer model [1]. The canopy 
is supposed to be horizontally homogeneous with two layers of discrete elements: a lower layer of resonant 
cylinders for the trunks and an upper layer for the crowns. The crown layer can be made up of different 
elements: flat disks for the leaves, cylinders for the needles and the branches. These elements are characterized 
by their density and their geometric (shape, size, orientation) and dielectric properties. The soil is modelized 
using the 3 classical models (small perturbations, physical optics or geometric optics). For a given radar 
configuration (frequency, incidence angle, polarization) the model predicts the total backscattering coefficient 
as well as the different components (direct canopy backscatter, direct soil backscatter attenuated by the 
canopy, soil-canopy interaction). 
The semi-empirical model: “water-cloud” 
The “water-cloud” models has been introduced by [3] and modified by several authors [4, 5]. In these models, 
the power backscatter by the whole canopy <r° is represented as the incoherent sum of the contribution of the 
vegetation <r° v and of the contribution of the underlying soil a ° t , the latter being attenuated by the vegetation 
layer, for a given incidence angle 9, “water-cloud” models take the general form: 
<r° = <t° v + 
<r° = AV\(l — i 2 ) 
t 2 = exp (—IBVil cos0) 
( 1 ) 
( 2 ) 
(3) 
where i 2 is the two-way attenuation through the vegetation layer, V\ and Vi are canopy descriptors such as 
LAI or total canopy water content (see [8] for a review). A and B are parameters depending on the canopy 
type. 
This formulation corresponds to the first-order solution of radiative transfer equation through a homo 
geneous medium, where multiple scattering effects are neglected. There is no theoretical background allowing 
the definition of the best set of canopy descriptors V) and Vi and the prediction of the values of the A and 
B parameters, which are always determined by fitting the model against experimental data. 
Test of MIMICS on experimental data 
Before using Mimics to generate a radar database, we verified it against experimental data. This was all the 
more important as Mimics was initially developed to simulate the backscatter of forests and not tested over 
crops. 
Experimental radar data 
The radar data used here were obtained with the airborne scatterometer Erasme during the European 
Agriscatt’88 campaign over the French site. Erasme was operating in C-band HH and X-band VV (5.35 
and 9.65 GHz respectively) with incidence angles ranging between 20° and 40°. It was flown at 4 dates 
on a 17 km axis along which 11 winter wheat fields were sampled. Surface soil moisture and vegetation 
measurements (LAI, water content) were done at the same time Erasme was operating. A comprehensive 
description of the experiment can be found in [9] and [10]. The ground measurements are summarized in 
Figure 1. Note the large range of values for both the soil and vegetation characteristics.