meteorological forecasts. It produces a hourly diagnostic of air temperature, wind, humidity, cloudiness, 
rainfall, snowfall, and surface radiations 
w grains morphology, 
zrains. This parameter 
irs of about 30 to 50 
:tive snow grains size 
i convex diameter and 
r these measurements, 
mal boxes to the cold 
itical parameters were 
finite assumption. The 
i pm. 
with a DEM (Digital 
CROCUS results, we 
All the images Nvere 
ground control points, 
h cartesian projection, 
graphic and the image 
ations on a 30 m x 30 
e difficulty is to locate 
available. 
ite are computed from 
xlel (Simulation of the 
rent ground reflectance 
er reflectance on lakes 
still are processing the 
i in laboratory but also 
eorological conditions, 
h different layer. They 
umerical model, called 
». It derives a complete 
age and stratigraphy of 
orological observations 
'¡cal objective analysis 
y input data. It aims to 
it points of the Alpine 
s as well as numerical 
5 ■ SNOW REFLECTANCE MODELING 
The bidirectional reflectance of the snow was computed with a model based on the radiative transfer 
(Stamnes et al„ 1988). It was found indeed that, if the snow is assumed to be iambertian, the measured 
reflectance was much too low. The phase function is computed from the Mie theory, i.e. assuming that the 
snow grains are spherical. All the calculations were done for pure snow without any pollution and for a 
uniform grain size. 
6 • PRELIMINARY RESULTS 
Two examples are given hereafter about the 24 April 1992 Landsat TM data. One example is about the 
large scale comparison which can be done between the satellite data and the CROCUS output on the snow 
surface temperature. The other one is a study of the dépendance between ratios of reflectances at different 
wavelenghts on the snow grain size: the measurements (satellite and in situ data) are compared to the 
theoretical curves. 
6.1. Snow surface temperature 
The thermal infrared channel 6 of Landsat TM is used for the surface snow temperature determination. 
After calibration of the data, the Planck function is inverted, assuming a snow emissivity e-1, to get the 
apparent temperature of the snow at the top of the atmosphere. Instead of using a model for atmospheric 
corrections, which would be difficult in this Alpine context, we assume a linear relationship between the 
apparent and the ground temperature. In situ measurements are used to get the coefficients of the linear 
regression (Fig. 1). A water surface temperature on a lake is also used because the range of in situ 
measured snow temperatures is small. Some ground measurements were rejected because the large pixel 
size in channel 6 (120 m) enhances the environment effects. 
Therefore a map of surface snow temperature can be derived and, with the DEM, the variation of 
temperature against elevation and slope orientation. Comparisons will then be made between this set of data 
and CROCUS output which also gives the variation of snow surface temperature against elevation and slope 
orientation. Preliminary results were made on a few sites and are satisfactory. 
F 'g- I: Linear regression between TM6 apparent temperatures and measured surface temperatures