TEST OF SURFACE BIDIRECTIONAL REFLECTANCE MODELS WITH SURFACE 
MEASUREMENTS: RESULTS AND CONSEQUENCES FOR THE USE OF REMOTELY 
SENSED DATA 
Cabot F. (1) , Qi I.< a , Moran M. S. (2) and G. Dedieu 11 ’ 
(1) LERTS, 18, Avenue E. Belin, 31055 TOULOUSE CEDEX, FRANCE 
(2) USDA/ARS, USWCL, 4331 E. Broadway, PHOENIX, AZ 85040, USA 
Abstract 
The use of remote sensing measurements of the surface reflectance, especially when applied to satellites, does not allow 
us to choose, most of the time, the geometric conditions of the acquisition. The effect of this limitation is important 
because the value that we measure strongly depends on these conditions. The aim of this paper is, first, to give an idea of 
the amplitude of the variations due to bidirectional effects. Then, it gives a panorama of various methods aimed at 
minimizing these effects, or extracting some information from their presence. 
1. Introduction 
Our capacity to use remotely sensed data for vegetation temporal monitoring and for the retrieval of surface 
characteristics such as albedo is strongly dependent on our knowledge of the surface bidirectional reflectance distribution 
function (BRDF). Although a number of BRDF models exist, few studies have dealt with the comparison of these models 
and the definition of a strategy to use them with satellite data. In this study, we used measurements of surface 
bidirectional reflectance factors (BRF) to fit the parameters of several published BRDF models. BRF measurements were 
made using aircraft-based sensors in the principal and orthogonal planes, and in the viewing plane of satellite-based 
sensor. Ground targets included smooth bare soils, full-cover agricultural crops and orchards. During each field 
campaign, remotely-sensed measurements were supplemented with measurements of vegetation, soils and atmospheric 
characteristics. 
In a first step, BRF measurements were used to fit model parameters, and the output of the various models was then 
compared on the basis of residual runs, errors, systematic defaults, and retrieval of land surface parameters when they 
were explicitly included in the model. In a second step, we analyzed the capacity of these models to infer the surface 
BRDF from reflectance factor measurements along a single plane. This test was conducted to assess the feasibility of 
deriving surface albedo from current sensors like SPOT/HRV and NOAA/AVHRR, which acquire bidirectional data 
along a single plane. We addressed this issue by fitting the models with BRF data acquired in one plane, and testing the 
ability of the models to predict reflectance factors in a different plane. Finally, the models were used to simulate the 
signal at the satellite level. These simulations provided guidelines for the definition of methods aimed at the 
normalization of off-nadir viewing effects and the subsequent retrieval of surface albedo. 
2. Data 
The data sets used in this study are part of what was acquired during the MAC-BRDF experiment, held at Maricopa farm 
near Phoenix (USA) in September 1991. As far as this paper is concerned, only three data sets have been extensively 
used: i) Aircraft borne radiometer acquiring surface l uminanc e measurements in 4 spectral bands (equivalent to 
LANDSAT-TM1, SP0T-XS1, SPOT-XS2 and SPOT-XS3). The radiometer was flown at an altitude of about 150 m 
above ground level, in the solar principal plane, with pointing direction from -40° (solar direction) to 40° (anti solar 
direction) with a 10° step, ii) Advanced Solid state Array Spectroradiometer (ASAS) was flown at two altitudes (5300 m 
and 2300 m), also in solar principal plane. This instrument provides images of the ground with pointing directions from - 
45° to 45° with a 15° step, in 29 spectral bands from 462.3 nm to 865.5 nm with an average bandwidth of 20 nm. iii) 
SPOT imagery data were also used. Two scenes were acquired, on the 7 and 8 of September, simultaneously with ASAS 
and aircraft measurements, both operated in the same acquisition plane. A full description of this data set can be found in 
Moran etal., 1990. 
All these measurements were calibrated in reflectance and atmospheric corrections were applied to SPOT and ASAS 
data, using respectively SMAC (Rahman and Dedieu, 1994) and a combination of 5S (Tanrti et al„ 1990) and a new 
version of the Hermann Browning (1965) algorithm. These models were ran with aerosol optical thickness, water vapor 
content and pressure measured at the time of acquisition.