523 
2.4 Emission behavior of forest 
The emissivity of forest e forest ( fp ) will be a basic parameter in the mixed signature algorithm, 
where we assume that the temporal variability of this parameter is very low and well known even after snow 
events. Additionaly, since forest can form large and homogenous areas, this surface type is well suited for 
investigations using passive microwave data from satellites without a mixed signature approach. Using (6) as 
an estimator of T, we can calculate e forest(f p) from the SSM/I brightness temperatures Tb f p applying 
®forest(f,p) (7) 
at frequencies f = 19 and 37 GHz. Figure 4 shows the temporal variation of the emissivity at 19 and 37 GHz of 
a large forest with an area of around 150 km 2 located in the Rhine valley near Mulhouse (France), the so called 
,jFor6t Domaniale“. The emissivity at 19 GHz shows temporally very constant values at a mean of 0.938 with a 
standard deviation of 0.00321. At 37 GHz the standard deviation is higher (0.00888) and the mean value is 
0.927. The higher standard deviation at 37 GHz is mainly due to the fact that the microwave brightness 
temperatures at 37 Ghz are more affected by atmospheric effects (especially rain) than at 19 Ghz. Additionaly, 
since T is calculated from data at 19 Ghz, it may show a higher correlation with the estimated emissivity at 19 
GHz than with 37 GHz. 
For Februray 13 we observe a decrease in estimated emissivity, which is far more pronounced 
at 37 GHz, where we can find values down to 0.88. As mentioned above, our algorithm overestimates physical 
temperatures below the freezing point An assumed overestimation of 5K gives a 37 GHz emissivity value erf 
0.92 instead of 0.88. The polarisation difference of the 19 GHz emissivity (Fig. 5) shows a small increase to 
0.05. The nearest ANETZ-station (Basel Binningen), which is located about 15km south of the forest reports 
for the period in question grass-temperatures constantly below 273K and a total snow depth of 8 cm (Fig. 6). 
These data lead to the conclusion that the decrease in emissivity at 37 GHz can be attributed to homogenous, 
comparatively shallow snow within the observed area. Because of the only low increase in polarisation 
difference, the snow is homogenous and has no internal structure, which would have lead to higher polarisation 
differences. The snow events during December 1991 could not be monitored due to missing SSM/I data 
3 CONCLUSIONS 
We presented three methods for estimating the physical land surface temperature in a 
sufficient spatial and temporal resolution. Comparisons between NOAA/Meteosat thermal infrared brightness 
temperatures visually classified as cloud-free and ground measurements revealed differences up to 17°C. These 
errors are mainly due to atmospheric effects, infrared surface emissivities below 1.0 and local topographic 
conditions. The statistical interpolation of ground-measurements resulted in a temperature field which is 
strongly dependent on the spatial distribution of the used ground-stations and which veils the true variations erf 
the temperatures of the different surface types. As a third approach, we discussed a method for estimating the 
physical temperature of land surfaces from a linear combination of SSM/I brightness temperatures at 19 GHz. 
RMS-errors of up to 3.53K are observed. No preclassification of the SSM/I brightness temperatures is necessary 
with this algorithm. The influence of the presence of extended snow-covered areas within the footprints needs 
further investigations. 
Together with the interpolated SSM/I brightness temperatures, the physical temperature data 
will be incorporated into a mixed signature model using a-priori knowledge of the composition of the earth 
surface within a SSM/I footprint in order to extract and analyse time series of microwave emissivity values erf 
the snow-free and the snow-covered alpine and lower regions. To make full use of the temporal and spectral 
resolution of the SSM/I channels, the atmospheric correction of the brightness temperatures at the higher 
SSM/I frequencies (37 and 85 GHz) has to be performed.