International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
  
3.1 Error due to forest cover 
Tundra 
  
| ait 
:| 
SWE error (%) 
© 
  
  
  
-40 
-50 
OCT NOV DEC JAN FEB MAR APR MAY 
Taiga 
50 
40 
s t1 
5 3 3 3 3 
-10 
go | + i 
-30 
SWE error (%) 
OCT NOV. DEC JAN FEB MAR APR MAY 
Prairie 
SWE error (%) 
e 
-10 + 
zs tt 
  
E -LELI 
OCT NOV DEC JAN FEB MAR APR MAY 
Alpine 
50 
40 | 
30 
= criidd 
Br $ 
SWE error (%) 
OCT NOV DEC JAN FEB MAR APR MAY 
Maritime 
e ; 2 3 55 
mes $ 
SWE error (%) 
© 
OCT NOV DEC JAN FEB MAR APR MAY 
Ephemeral 
SWE error (%) 
© 
»3 333833383 
OCT NOV DEC JAN FEB MAR APR MAY 
Figure 2. SWE overestimation or underestimation for the six Strum classes due to the assumption of constant grain size. 
The primary source of systematic error in SWE is the 
masking effect of vegetation, which reduces the 
brightness temperature difference term in (1). In the 
PM portion of the electromagnetic spectrum, the error 
due to forest cover is expected to be very high, 
upwards of 50%, since the emissivity of the overlying 
forest canopy can overwhelm the scattering signal 
from the snowpack (Chang et al., 1996; Brown at al., 
2003). Where forests are scant or absent PM estimates 
of SWE are more accurate. 
For each forested pixel, a fractional forest cover fi is 
calculated using the International Geosphere- 
Biosphere Program (IGBP) Land Cover Data Set 
described by Loveland et al. (2000). These data, at ! 
km x 1 km, are averaged to the 19 x IP 
latitude/longitude grid used in this study. The 
   
     
     
   
   
     
    
    
    
   
   
   
    
    
     
    
    
Inter 
— 
pe 
ca 
cl: