e x = ke 19,v-( k -1)^19,h- 
( 4 ) 
522 
By empirical optimization we can reduce the uncertainty of the microwave surface emissivity in (1) and then 
estimate T from the SSM/I brightness temperatures at 19 GHz: 
^ kTb 19 v - (k - l)Tb 19 h 
(5) 
with e x = average value of e x of all surface types for which the emissivity was determined. Mhtzler (1994) 
originally developed this method from data at 4.9 and 35 GHz. We linearly extrapolated the c x values to the 
SSM/I frequency (19 Ghz). Setting k=1.5 gave the best correspondence of the SSM/I derived physical 
temperature with in-situ measurements. Thus, we can use 
f=(L5Tb 19v -05Tb 19h )/0548 (6) 
as the estimator for land surface temperature from SSM/I brightness temperatures. 
In order to validate (6) values of T where compared against grass temperature measurements 
of six ANETZ-stations. The selected stations are located in the Central Plains of Switzerland (altitude £ 700 
m.a.s) and are representative for their surroundings. Figure 3 shows the comparison of the estimated phyiscal 
temperature of all available evening overpasses (-18:00 UTC) of winter 1990/91 for all six meteorological 
stations against the ground-measured grass temperature. The rms for all observations is 3.53K, the coefficient 
of determination (ri) is 0.822. The method performance is better for grass temperatures above 273K (rms= 2.63, 
ri= 0.839), temperatures below this value may be overestimated by 10K. 
ANETZ Subset 
Fig. 3: Comparison of ground-measured grass-temperature and es timate d physical temperature (using (6)) for 
all coincident evening overpasses of winter 1990/91 and four ANETZ-stations (Ziirich-Kloten, Bem-Liebefeld, 
Tanikon, Reckenholz).