415 
COMBINED USE OF THEORETICAL AND SEMI-EMPIRICAL 
MODELS OF RADAR BACKSCATTER TO ESTIMATE 
CHARACTERISTICS OF CANOPIES. 
Laurent PRÉVOT and Thomas SCHMUGGE 1 
Inra Bioclimatologie, BP 91, 84143 Montfavet, France 
U) Usda-Ars Hydrology Laboratory, Beltsville, Md 20705, Usa 
Abstract 
Semi-empirical “water-cloud” models are convenient tools for inversion of radar data over canopies, but suf 
fer several limitations: their parameters cannot be infered physically and their stability cannot generally be 
assumed. To improve their use for estimating vegetation characteristics, we propose in this paper to couple 
them to a theoretical model of canopy backscatter, Mimics, used as a radar data generator. We first ver 
ify Mimics against data obtained in C and X bands over wheat fields. We show that Mimics adequately 
represents the data in C-band, at all incidence angles. A strong discrepancy is observed in X-band at high 
incidence angle, for high levels of biomass but the agreement is fairly good otherwise. Simplified expressions 
for the attenuation by the canopy and for the canopy direct backscatter are then proposed. These expres 
sions are similar to the “water-cloud” models and only involves LAI, canopy water content and surface soil 
moisture as variables. We also demonstrated that the outputs of Mimics in C-band can be represented by 
a simplified model, involving a limited number of variables and parameters. This provides a way to invert 
Mimics. 
Keywords: radar, modelling, inversion, LAI, canopy water content. 
1 Introduction 
Remote sensing techniques using the optical domain have demonstrated their capabilities to monitor vegeta 
tion canopies. Unfortunately, their sensitivity to the cloud cover constitutes a serious draw back, particularly 
when a good temporal repetition is desired, as it is the case for agronomic and hydrologic applications. 
Because of their greater penetration depth in natural media, atmosphere and vegetation, microwave 
remote sensing techniques can overcome the limitations of the optical domain. As several imaging radars are 
already or to be launched during the coming decade, it is the most important to develop inversion algorithms 
allowing the use of radar data for estimating canopy characteristics such as biomass and leaf area index, as 
well as to correct estimations of surface soil moisture for the effect of vegetation. 
Existing direct backscattering models from a canopy can be divided in two general classes: the theo 
retical models, based on the field or intensity approaches, and the semi-empirical “water-cloud” models. The 
theoretical models [1, 2] are excellent tools for understanding the mechanisms of volume scattering, since one 
can study the effect of varying the vegetation and soil parameters. But they are usually rather complex sets 
of equations, with many variables and parameters, so they cannot be inverted easily at the present time. 
Furthermore, the variables they involve are difficult to relate to the bulk parameters (LAI, canopy water con 
tent) needed in agronomic applications, and finally, they are probably too much computer-time consumming 
to be used in operational procedures. 
On the other hand, semi-empirical models [3, 4, 5] are very simple sets of equations, involving few 
variables and parameters. The variables they used are usually global ones, such as total canopy water 
content or leaf area index for vegetation and surface moisture for the soil, and it has been shown by various 
authors [6, 7, 8, 9] that they can be inverted with respect to these variables. Therefore they are good 
candidates for being used in inversion algorithms, but because of the relative weakness of their physical 
background, their parameters cannot be theoretically predicted, so they are usually fitted on experimental 
datasets. 
To improve inversion algorithms of radar data, it is proposed in this paper to combine these two 
modelling approaches in the following way. A theoretical model is used to simulate backscattering coefficients 
of agricultural canopies. Semi-empirical “water-cloud” models are then fitted on the outputs of this theoretical 
model, allowing the evaluation of inversion schemes. The main advantages of this approach are first to allow 
extensive study of the sensitivity of the parameters in “water-cloud” models to the variations of target 
parameters, and secondly to compensate for both the lack of generality of “water-cloud” models and the 
complexity of the theoretical ones.