400 
L BACKSCATTERING COEFFICIENT IMAGES OF THE AFRICAN CONTINENT 
Scatterometer data delivered by ESA consist of a serie of products including 19 x 19 values corresponding to 
measurements acquired over a surface of about 500 km x 500 km. For each value or node, the following 
information are given : 
. o°(dB) 
. longitude / latitude (degrees) 
. local incidence angle (degrees) 
. coefficient of variation of the measure Kp (%) 
The procedure adopted to elaborate a backscattering coefficient image from the ERS-1 original data and an 
illustration of a global map can be found in Mougin et al (1993). To built a cr° image, the region under 
consideration is divided into elementary cells equal to 0.25° x 0.25° wide. Backscattering coefficient data are 
attributed to their corresponding cells as a function of their associated latitude/longitude coordinates. Each 
antenna is treated separately. Ascending or descending passes can also be considered separately. One important 
characteristic of the ERS-1 scatterometer lies in its capability to acquire measurements at different incidence 
angles. ct° angular signatures are characteristic of surface parameters including topography at different scales, 
soil surface roughness and vegetation density. Furthermore, successive acquisitions are not time coincident and 
are not exactly performed at the same location. Angular signatures are also expected to vary throughout the 
year. Accordingly, angular signatures, although interesting are difficult to interpret. Illustration of angular 
signatures for different vegetation surfaces can be found in Mougin et al (1993). For tropical forests, the 
angular signature remain very flat (1-2 dB variation within the range 25° to 55° of incidence angle). On the 
opposite, important variations are observed for sparse vegetation or for bare surfaces. Over the desert, a 
maximum of 12 dB can be observed. 
Within each cell and following Kennett and Li (1989b), a linear fit is retained to model the variation of o° (in 
dB) versus incidence angle. A polynomial fit up to the 2 nd order does not improve significantly the final results. 
This procedure allows global maps to be generated at any incidence angle between 20° and 55° and for each 
antenna. About one month of data is necessary to obtain a global image of land surfaces with only a few 
missing values. Following this procedure, twelve images can be produced per year at a given incidence angle 
and for each antenna. Illustrations of the resulting images are given in figures 1 and 2 showing backscattering 
images of the whole African continent for may 1992 and September 1992, respectively. These images, given 
here at a 30° incidence angle, correspond to data acquired by the midbeam antenna. Several observations can 
be made. Firstly, it is surprising to note that these images present a strong correspondence with existing global 
vegetation maps (for instance, the FAO map, Lavenu et al, 1986). Desert, tropical forests (Congo bassin, 
Guinean forests,...), savannas, woodlands can be easily discriminated. Furthermore, tropical forests are 
particularly well identified indicating the high capability of the ERS-1 scatterometer to provide a global map of 
the intertropical belt. On desertic areas (Sahara, Kalahari), important spatial variations are observed. These 
variations in backscattering (up to 10 dB) can be attributed to the local topography and relief; the slopes 
oriented to the radar beam giving the strongest responses. On the whole, with the exception of desertic areas, 
the highest backscattering coefficient values correspond to the tropical forests. 
Important variations can be also be noted between these different maps, especially in West Africa where 
vegetation of savannas is known to exhibit a well pronounced seasonnality between may and September. In 
order to analyse more precisely the spatial and temporal variations of the backscattering coefficients, a North- 
South transect is chosen along a vegetation gradient from desert to tropical forests. The location of the transect 
is indicated on the image by a black line, starting from latitude 25°N to 25°S at a 15° longitude. Results are 
given in figure 3 for may 1992 and September 1992 corresponding to important vegetation development 
periods. As already noticed on the previous o° maps, the lowest values are found in desertic areas (down to -18 
dB) whereas the highest values are encountered on tropical forests (around -7 dB). Savannas and woodlands 
exhibit typical temporal signatures which can be related to rainy periods and to the associated growing season 
of vegetation. Tropical forests present a stable backscattering coefficient indicating also that the scatterometer 
is well calibrated. Similar observations can be made on deserts where the bascattering coefficients do not 
present significant variations throughout the year.